- Shopping Bag ( 0 items )
Of the myriads of developmental processes that define plant form and function, flowering is of exceptional interest to horticulturalists. The vast majority of horticulturally important crops are in some way dependent upon flowering, whether the flower is the primary goal of production, or is simply required for a crop to be produced. Much effort is currently being put into regulating the timing of flowering. In floriculture crops, the interest is in abbreviating or extending the vegetative phase in order to create an aesthetically pleasing balance between leaves and flowers, or to conveniently induce or repress flowering to take advantage of market potential. In ornamental foliage plants, and agronomically important plants that are grown for their leaf tissues (e. g., lettuce, spinach, and other greens), it is highly desirable to suppress flowering as long as possible. Also, in woody plants, there is a great deal of interest in finding means to abbreviate the vegetative phase, which in some species can last ten or more years and is probably the single most limiting factor for germplasm improvement through traditional breeding techniques.
Most efforts at controlling flowering time have involved manipulation of environmental conditions or the application of synthetic growth regulators. However, these approaches can increase production costs and labor requirements. In addition, the use of many traditionally utilized chemical compounds is becoming restricted. Alternative approaches to manipulate flowering-- including biotechnology-- will require a better understanding of the associated molecular mechanisms.
The physiology and phenomenology of the developmental transition from vegetative growth to reproductive growth-- flowering-- has been studied for many years, but only in approximately the last 10 years have the molecular mechanisms begun to be addressed. Flowering is ultimately determined by genes that govern the identity of the meristem, promoting or repressing floral fate versus shoot fate. When and how these genes are activated, in response to environmental cues and/ or developmental progression, is a fascinating question. As might be expected from the incredible diversity of flowering strategies employed in nature, it is now becoming apparent that flowering at the molecular level involves an extraordinarily complex web of interactive pathways. Here we review the current knowledge about the genetics and molecular biology of flowering in Arabidopsis thaliana, the only plant in which these aspects of flowering have been extensively studied.II. ARABIDOPSIS AS A MODEL FOR
Arabidopsis thaliana is an herbaceous weed of the mustard family with a natural distribution throughout the Northern Hemisphere (Meyerowitz 1989; Meinke et al. 1998). In addition to its many qualities that make it a superior model for plant biology in general (i. e., small size, rapid life cycle, and well-characterized genome), Arabidopsis is especially attractive as a subject for flowering-time studies because the timing of flowering can be strongly influenced by environmental conditions (e. g., light and temperature), thus permitting the molecular analyses of the associated input pathways.
In this species, at least fifty genes have been identified that act directly or indirectly either to promote or to repress flowering (Levy and Dean 1998). Many of these genes have been identified through a traditional genetic approach. Delayed flowering can result from loss of function of genes that presumably act to promote flowering, whereas accelerated flowering can result from loss of function of flowering-repressor genes. Although several repressor genes have been identified, most mutagenic approaches have targeted genes that act to promote flowering (Redei 1962; Koornneef et al. 1991). This is in part because the genotypes commonly used in the laboratory flower soon after germination in photoperiodically inductive conditions (long-day photoperiods), and mutants that flower even earlier, are difficult to discern in large populations. Screens designed to find early-flowering mutants among late-flowering genetic backgrounds, or employing photoperiodically noninductive conditions, should result in the identification of additional repressor genes.III. FLORAL INDUCTIVE PATHWAYS
An interesting finding coming from genetic analyses is that no single mutation completely eliminates flowering. This was an early indication that flowering is promoted by at least two pathways that can operate in a parallel, or partially redundant, manner. That such redundancy should have evolved makes sense, given the crucial importance of flowering in maintaining the species. The promotive genes identified through genetic analyses have traditionally been assigned into distinct groups based on the sensitivity of the mutant phenotype to environmental conditions, and these groups have formerly been considered to define the pathways (Martinez-Zapater et al. 1994; Coupland 1995). Mutations in a subset of flowering-time genes predominately affect the photoperiodic control of flowering, such that the flowering habit of the corresponding mutant tends toward day-neutrality. Mutations in another subset of flowering-time genes result in delayed flowering without a significant loss of photoperiodic sensitivity-- i. e., these mutants flower later than wild-type plants under both photoperiodically inductive and noninductive conditions. Because mechanisms for sensing daylength are evidently intact in the latter mutants, the corresponding genes are supposed to function in an environmentally "autonomous" pathway that acts in parallel with the "photoperiodic" pathway to eventually initiate flowering, even under unfavorable conditions (Martinez-Zapater et al. 1994; Coupland 1995; Amasino 1996). Another characteristic of mutants in the autonomous pathway is that they exhibit a significant vernalization response-- i. e., the late-flowering phenotype can be fully "rescued" by a long-term cold treatment given to the imbibed seed or young plant. In contrast, cold is largely ineffective to accelerate flowering of the photoperiodic pathway mutants (Martinez-Zapater et al. 1994).
Koornneef et al. (1991, 1998a) used double-mutant analysis to examine the epistatic relationships between the commonly studied flowering-promoting genes in an attempt to better define such pathways. The rationale for this type of approach is as follows: if two genes operate in a more-or-less linear pathway, then loss of both genes' function should confer a phenotype that is similar to that of the single mutant (i. e., the double mutant should flower no later than either single mutant). However, if genes operate in parallel pathways, a significant enhancement of the late flowering might be conferred by combining the mutant alleles. A caveat to this type of genetic approach is that it is only valid when using complete loss-of-function alleles, as enhancement of the pheno-type should be expected when partially functional alleles operating in the same pathway are combined. In general, the results of these experiments were inconsistent with the simple assignment of flowering-time genes to independent pathways. This suggests that there is significant interaction (" crosstalk") between pathways. Another finding from these studies was that flowering was not prevented even when combining mutations in genes considered to act in the photoperiodic and autonomous pathway. Thus, the redundancy of flowering pathways is more extensive than was previously thought.A. Photoperiodic Induction
1. Light Effects on Flowering in Arabidopsis. As in many other plants, both light quality (wavelength) and photoperiod strongly influence flowering time in Arabidopsis. In general, flowering in this species is delayed by red light and accelerated by blue light (Brown and Klein 1971; Eskins 1992). The molecular biology of the major photoreceptors in plants, the red/ blue-sensitive phytochromes and green/ blue/ UV-A-sensitive cryptochromes, has been extensively reviewed and will not be discussed here (Barnes et al. 1997; Cashmore 1998; Whitelam and Devlin 1998; Ahmad 1999; Cashmore et al. 1999; Deng and Quail 1999; Briggs and Huala 1999). Mutations that abrogate synthesis of the phytochrome chromophore and therefore result in an absence of functional phytochrome, or mutations that specifically result in loss of the major light-stable phytochrome, PHYB, confer early flowering, suggesting that the negative effect of red light is mediated by PHYB. Mutants lacking function of the CRYPTOCHROME1 (CRY1) gene exhibit delayed flowering that is evident in both long and short days (King and Bagnall 1996; Coupland 1997). This phenotype is especially striking when plants are grown under blue light (Bagnall et al. 1996), suggesting that CRY1 mediates blue-light promotion of flowering. In contrast, mutants lacking function of CRY2 (allelic to the previously described flowering time gene FHA) exhibit a much-reduced photoperiodic response, flowering much later than wild type in long days and slightly earlier than wild type in short days (Koornneef et al. 1991; Guo et al. 1998). In addition, constitutive expression of the CRY2 gene in transgenic plants accelerates flowering in short days, but not long days. Unlike in cry1 mutants, flowering in plants lacking CRY2 is accelerated by blue light (Guo et al. 1998). Given the delay in flowering in white light conferred by loss of CRY2 activity, one interpretation of this data is that CRY2 normally acts not as a direct positive regulator under blue, but as a negative regulator of the repression of flowering imposed by PHYB (Guo et al. 1998).
2. The Endogenous Clock. In plants, as in other organisms, one or more molecular mechanisms sustain oscillations with periods of approximately 24 h. The circadian rhythms generated by these endogenous "clocks" allow plants to anticipate daily variations in environmental conditions and thereby optimize their responses to them. One example is the family of LHC genes encoding light-harvesting chlorophyll a/ b-binding (CAB) proteins, which are upregulated in a diurnal manner before the expected onset of illumination (Piechulla 1988; Nagy et al. 1988).
A large body of physiological evidence implicates the clock in mediating the effects of photoperiod on flowering. Evidence is also accumulating that light quality as well influences flowering time by virtue of its effects on the clock. Thus, the clock has a central and very important role in flowering. How might a self-sustaining oscillatory mechanism in plants be composed at the molecular level? Some clues come from research on the Drosophila (fruit fly) clock mechanism that controls eclosion (emergence from the pupae) and locomotor activity. This clock is essentially comprised of an oscillatory mechanism set up through the interactions between two proteins, TIMELESS (TIM) and PERIOD (PER). Transcription of both the PER and TIM genes increases during the subjective day, from a minimum rate near the onset of illumination (referred to as Zeitgeiber time 0, or Zt0) and reaching a maximum rate at approximately Zt12 (Hardin et al. 1992). Maximal accumulation of PER and TIM mRNA is offset 2 to 4 hours, whereas maximal accumulation of the proteins is offset another 2 to 4 hours (So and Rosbash 1997). Thus, PER and TIM protein levels reach a maximum at Zt16-20, a point during which transcription of the genes is rapidly decreasing. In fact, transcriptional repression of the PER and TIM genes is a direct result of the increase in protein levels. Heterodimerization between PER and TIM allow the proteins to gain entrance into the nucleus, where they block the transcription of their own genes by the CLOCK and BMAL1 transcription factors (Vosshall et al. 1994; Gekakis et al. 1995; Darlington et al. 1998). The inhibition of transcription by PER/ TIM allows the circadian cycle to begin anew. Constant turnover of the mRNAs and proteins is necessary for the oscillations to continue. The TIM protein is thought to be destabilized through phosphorylation by the product of the DOU-BLETIME gene, which is structurally related to the kinase domain of human casein kinase I (Kloss et al. 1998; Price et al. 1998). In addition, in the absence of TIM, PER protein fails to accumulate, suggesting that TIM functions directly or indirectly to stabilize PER (Price et al. 1995).
Although great progress has been made in understanding the basics of this Drosophila clock mechanism, how the clock operates in plants is mostly unknown. PER, TIM, and other components of the fly clock were discovered through traditional genetic analysis. Arabidopsis displays numerous visible phenotypes that cycle in a circadian manner [e. g., movements of cotyledons and primary leaves (Engelmann et al. 1992), alterations in the rate of hypocotyl elongation (Dowson-Day and Millar 1999), and changes in stomatal aperture (Somers et al. 1998b)], but in all cases these phenotypes are subtle and thus not useful for mutant screening. Millar et al. (1995) generated a synthetic circadian phenotype by expressing the firefly luciferase gene under the control of an LHC gene promoter in transgenic plants. Screens using this LHC: LUC genetic background yielded numerous mutants. The best-characterized, designated toc1-1, exhibits a slightly shorter period length of LHC mRNA expression in both constant light and constant darkness (Millar et al. 1995; Somers et al. 1998b). In addition, the mRNA expression of members of at least one other circadian-cycling nuclear gene family, GRP7/ 8 (see below), is altered in a similar manner (Kreps and Simon 1997). Although toc1-1 plants were originally reported to be phenotypically indistinguishable from wild-type plants, more careful observations revealed that toc1-1 plants were disrupted in multiple circadian cycling phenotypes. In addition, toc1-1 plants exhibited aberrant floral initiation, flowering earlier than wild-type plants under short photoperiods and later than wild-type plants under long photoperiods (Somers et al. 1998b). These findings suggest that the multiple circadian processes and the timing of flowering are controlled either by a single clock, or by multiple related clocks sharing the TOC1 component. The TOC1 gene was recently cloned and found to encode a protein with homology to the receiver domain of response regulators from two-component signal transduction systems (Strayer et al. 2000).
A dominant mutation in the LATE ELONGATED HYPOCOTYL (LHY) gene leads to loss of rhythmic mRNA expression of clock-regulated genes and defects in multiple clock-influenced phenotypes, including flowering time and circadian leaf movements (Schaffer et al. 1998). In wild-type plants, LHY mRNA levels oscillate in a circadian manner, whereas LHY mRNA is expressed at a constitutive high level in the lhy mutant (Schaffer et al. 1998). In addition, in transgenic plants containing a singly copy of a lhy mutant allele, cycling of the endogenous wild-type LHY mRNA is suppressed. These findings indicate that LHY is part of a feedback circuit that regulates its own mRNA expression. The LHY gene product is a member of a large family of proteins structurally related to the vertebrate proto-oncogenic transcription factor c-Myb (Martin and Paz-Ares 1997). In Arabidopsis, this family also includes the product of the CCA1 (CIRCADIAN CLOCK ASSOCIATED 1) gene, orig-inally identified as a factor that bound a LHC promoter element essential for its regulation by light and the clock (Wang et al. 1997). Like LHY mRNA, CCA1 mRNA and protein also cycle with a circadian rhythm (Wang and Tobin 1998). Constitutive expression of CCA1 mRNA under control of the strong, viral CaMV 35S promoter (35S: CCA1) in transgenic Arabidopsis, like constitutive expression of LHY mRNA from the mutant lhy allele, leads to the disruption of the circadian mRNA expression patterns of various clock-regulated genes, including LHY, and such plants exhibit delayed flowering in long-day conditions (Wang and Tobin 1998). These findings suggest that both LHY and CCA1 are potential key components of a central clock mechanism. That the clock defect conferred by loss of CCA1 function is apparent even in the presence of LHY activity indicates that, despite their structural similarities and similar effects by constitutive expression, LHY and CCA1 do not have strictly redundant roles (Green and Tobin 1999).
Like that of TIM, the activity of the CCA1 and LHY proteins may be negatively regulated by phosphorylation. Both proteins are substrates for the protein kinase CK2 in vitro (Sugano et al. 1998, 1999). Constitutive expression of CKB3 mRNA, encoding a regulatory subunit of CK2, in Arabidopsis mimics the effects of loss of CCA1 function, substantially shortening the rhythm periods of multiple clock-regulated genes. However, in contrast to the delay of flowering conferred by loss of CCA1, flowering is accelerated in 35S: CKB3 plants (Sugano et al. 1999).
GIGANTEA (GI) is another member of the set of genes involved in the promotion of flowering by long days. In plants grown in a regular light-dark photoperiod, mRNA levels for GI oscillate in a diurnal pattern, and studies in which plants were kept in constant light or darkness indicate that GI is under clock control (Fowler et al. 1999; Park et al. 1999). In 35S: CCA1 or 35S: LHY plants, this rhythmic expression pattern is disrupted, indicating that GI is regulated by these genes (Fowler et al. 1999). However, disruption of GI function also affects expression of CCA1 and LHY mRNA. In some gi mutant backgrounds, the amplitude of CCA1 and LHY oscillations are diminished and periodicity becomes less obvious (Fowler et al. 1999; Park et al. 1999). The finding that GI may act both upstream and downstream of the clock genes CCA1 and LHY suggests that GI is intimately associated with the clock. Mutants lacking GI function also exhibit reduced seedling deetiolation under red light, suggesting that GI could be involved in PHYB signaling (Huq et al. 2000).
How GI might carry out its function has not been determined. GI was recently cloned and encodes a large protein that is predicted by computer modeling to contain transmembrane domains (Fowler et al. 1999; Park et al. 1999). However, more recent evidence indicates that the GI protein is localized to the nucleus (Huq et al. 2000). The GI transcript was detected in both very young seedlings and mature plants, and is apparently not restricted to any specific tissue type (Fowler et al. 1999). Open reading frames have been identified from both rice and maize that would encode proteins with significant amino acid sequence identity to the GI protein (Fowler et al. 1999; J. Liu and S. van Nocker, unpublished data). Because structural homology is often associated with functional homology, it is possible that these monocot GI orthologs are also involved in flowering. The current efforts to better characterize the rice genome, and determine gene function in maize through reversed genetics approaches, will allow this idea to be tested (Goff 1999; Martienssen 1998; Walbot, 2000).
Other potential clock genes include members of the GRP7/ 8 (also called CCR1/ 2) family. These genes encode small proteins containing an interesting bipartite structure (van Nocker and Vierstra 1993; Carpenter et al. 1994). The amino-terminal domain contains a specific RNA-binding consensus sequence termed the RRM motif (found also in the flowering-time genes FCA and FPA; below), whereas the carboxyl-terminal region is greatly enriched in glycine residues, a configuration seen in many plant cell wall proteins (Showalter 1993; Cassab 1998). These genes are expressed to high levels in meristematic tissues, and, in addition to being regulated in a circadian pattern, are upregulated by lowered temperatures (Heintzen et al. 1994; Kreps and Simon 1997). The protein products of these genes also oscillate with a circadian period, and are localized to the nucleus (Heintzen et al. 1994). Constitutive expression of the GRP7 gene in transgenic Arabidopsis suppresses the circadian oscillations of mRNAs for both the endogenous GRP7 gene and for GRP8, suggesting that the respective proteins are involved in a mutual, autoregulatory feedback loop. This effect of GRP7 on the oscillations of its own transcript is not mediated entirely through its promoter, suggesting that at least some regulation occurs at the posttranscriptional level (Staiger and Apel 1999). As previously mentioned, GRP7/ 8 expression is affected by impairment of TOC1 function. Interestingly, however, unlike in toc1 mutants, the rhythmic expression patterns of specific clock-regulated genes were not affected in 35S: GRP7 plants (Heintzen et al. 1997). This suggests both that the GRP7/ 8 clock acts downstream from TOC1, and that the output of the AtGRP7/ 8 oscillator is limited. The function of the so-called AtGRP7/ 8 "slave" oscillator is not known, as phenotypic abnormalities associated with constitutive expression of GRP7have not been reported. In light of the intimate relationship between circadian rhythms and flowering, it is reasonable to hypothesize that these genes are somehow involved in the regulation of flowering time. On the other hand, genes encoding small proteins exhibiting the RRM motif/ glycinerich bipartite structure have also been identified in mammals, amphibians, ascidians, and cyanobacteria (Nishiyama et al. 1997; Danno et al. 1997; Uochi and Asashima 1998; Tanaka et al. 2000; Maruyama et al. 1999). At least a subset of these genes cycle in a diurnal manner and/ or are inducible by lowered temperatures (Nishiyama et al. 1997, 1998; Sato and Maruyama 1997; Maruyama et al. 1999; Danno et al. 1997). Thus, these RRM-GRP proteins may carry out a function that is conserved among kingdoms.
3. Entrainment of the Clock by Light. A notable feature of clocks in all organisms yet studied is that the innate period is somewhat longer or shorter than 24 h. In order to cycle with a precise daily rhythm, the clock must be entrained, or synchronized, each day. In Drosophila, light serves to entrain the clock by initiating the phosphorylation and rapid degradation of the TIM protein (Hunter-Ensor et al. 1996; Myers et al. 1996; Zeng et al. 1996). This results in a phase delay in the evening, when TIM is being continually resynthesized, and a phase advance in the morning, when TIM is not effectively replaced. The light signals are perceived by cryptochrome, a protein that is similar in amino acid sequence to the CRY proteins in plants. Upon illumination, CRY undergoes a photochemical change that allows a physical interaction with the TIM protein and presumably initiates the degradation process (Ceriani et al. 1999).
As in other organisms, under constant illumination, the period of the clock can be modulated by light in an intensity-dependent manner (Somers et al. 1998a, b). Although this phenomenon may not be physiologically relevant, it has proven useful for determining the potential role of specific genes in the light entrainment of the clock. Millar et al. (1995) found that both red and blue wavelengths were effective in shortening the period of LHC: LUC expression, suggesting the involvement of phytochromes and potentially also cryptochromes. By examining the period length of LHC: LUC expression in phyA, phyB, cry1, or cry2 mutant plants, Somers et al. (1998a) concluded that red light signals are transmitted to the clock by phytochromes A and B, with PHYA acting under low intensities and PHYB acting under high intensities, whereas blue light inputs are provided by PHYA and cryptochromes. Interestingly, loss of CRY2 shortened the period under low-fluence blue light, but had little effect under the higher fluences where photoperiodic timing of flowering is affected.
Unlike in Drosophila, no direct interactions between photoreceptors and clock components have been demonstrated in plants, and clock function may depend on a mechanism to transmit information from photoreceptors to the clock. The EARLY FLOWERING 3 (ELF3) gene may play such a role. Plants lacking ELF3 activity display phenotypes that mimic phyB mutations (i. e., increased hypocotyl elongation in red light and petiole length). However, mutations in ELF3 and PHYB have additive effects when combined, suggesting that ELF3 is not simply a component of a PHYB signal transduction pathway (Reed et al. 2000). Unlike mutations in PHYB, which alter the periodicity of the clock (above), mutations in ELF3 abolish rhythmicity, and do so in a light-dependent manner (Hicks et al. 1996). The CCA1 gene is rapidly and transiently upregulated in response to light, specifically red light (Wang et al. 1997), suggesting that CCA1 could also be involved in transmitting signals to the clock from phytochrome.
Whether or not the phosphorylation of CCA1 or LHY is associated with their turnover, as is the case with TIM, has not been reported. However, as in the Drosophila clock, protein degradation is an important process of the plant clock mechanism. Two homologous genes have been identified in Arabidopsis that might play a role in the turnover of clock components. These genes, FKF1 and ZEITLUPE (ZTL), encode proteins containing an F-box motif (Nelson et al. 2000; Somers et al. 2000). Where studied in other organisms, F-box-containing proteins act in the recognition of degradation substrates for the ubiquitin proteolytic pathway (Patton et al. 1998; Kornitzer and Ciechanover 2000). FKF1 tran-scripts oscillate in a circadian manner, whereas ZTL mRNA expression is apparently not under clock control. However, mutations in both genes confer a similar phenotype, with flowering delayed primarily under photoinductive conditions. In addition to the F-box, the FKF1/ ZTL proteins contain a segment similar to the flavin-binding domain in the blue-light receptor NPH1 involved in phototropism. At least ztl mutants exhibit a period-lengthening phenotype that is strongly light-dependent, and at least the FKF1 promoter is selectively activated under white or blue light. Taken together, these observations indicate that these proteins may function as light-dependent clock regulators (Nelson et al. 2000; Somers et al. 2000).
4. Other Photoperiodic Pathway Genes. CONSTANS (CO), another of the promotive photoperiod pathway genes, was one of the first of the flowering-time genes to be cloned, and thus has been the most extensively studied (Putterill et al. 1995). The CO protein contains zinc-finger-type DNA-binding domains common to the GATA1 family of transcription factors, and thus likely acts as a component of the tran-scriptional apparatus. Known mutations in CO are semidominant. Where a mutation results in complete loss of function of the gene, semi-dominance is an indicator that the respective gene product is limiting for the respective process (i. e., that relative levels of the gene product are important). Consistent with this, increasing CO activity, either con-stitutively through adding extra copies of the gene in transgenic plants (Putterill et al. 1995), or transiently by activating the protein in an inducible system (Simon et al. 1996), is sufficient to trigger flowering. In addition, consistent with its role in promoting flowering under photoinductive conditions, the mRNA levels of CO are elevated in long-day grown Arabidopsis plants relative to short-day grown plants (Putterill et al. 1995), and this regulation is accomplished at least in part by transcriptional upregulation of the gene (Suarez-Lopez et al. 1998). CO was found to be expressed in both leaf and stem tissue, but the very low abundance of the mRNA complicated a more thorough analysis of spatial expression patterns (Putterill et al. 1995). Its likely function as a transcription factor and regulation at the transcriptional level suggests that CO is an intermediate in a cascade of transcriptional events.
What could be upstream regulators and downstream targets of CO? CRY2 and PHYB likely act upstream of CO as indirect positive and negative regulators, respectively, because mutations in CRY2 decrease CO mRNA expression (Guo et al. 1998), and the early-flowering seen in phyB mutants in short days is alleviated in a genetic background compromised for CO activity (Putterill et al. 1995). Onouchi et al. (1998) clarified the relationship between CO and several other photoperiodic pathway genes by examining the effect of 35S: CO expression in pho-toperiodic pathway-mutant backgrounds. The premise of this experiment was that if CO acted in a genetic pathway downstream from GENE X, then removal of GENE X function should have no effect on the phenotype conferred by constitutive expression of CO (i. e., constitutive expression of CO would be epistatic to loss of GENE X function). In contrast, if CO acted upstream of GENE X, then adding CO activity should make no difference to the phenotype conferred by loss of GENE X function. In this case, the 35S: CO transgene was completely epistatic to gi and lhy, suggesting that CO acts downstream from these two genes. In contrast, 35S: CO had only a small effect in genetic backgrounds in which expression of the FT or FWA flowering-time genes was disrupted (see below), suggesting that these two genes function downstream from CO. Extending this experimental approach, Onouchi et al. (2000) discovered a target of CO by searching for mutations that suppressed the early-flowering phenotype conferred by constitutive CO expression. This gene, designated SUPPRESSOR OF CONSTITUTIVE EXPRESSION OF CO1 (SOC1), encodes a protein containing a domain designated the MADS box. This motif is present in other proteins known to bind DNA as homo-or heterodimeric complexes (Trobner et al. 1992; Riechmann and Meyerowitz 1997). SOC1 is expressed in the shoot and inflorescence apical meristem as well as the leaf primordia in response to inductive photoperiods (Samach et al. 2000).
Consistent with the results of Onouchi et al. (1998), Samach et al. (2000) identified the FT gene (see below) as a very early downstream target of CO activity. In the approach used here, the CO coding sequence was translationally fused to the ligand-binding domain of the rat glucocorticoid receptor. This CO-GR fusion protein was expressed constitutively in transgenic plants, and could be directed to the nucleus and thus "activated" by application of the synthetic glucocorticoid hormone dexamethasone. In this case, FT transcript accumulation was seen within two hours of dexamethasone application. This experimental approach also resulted in the identification of two other early downstream targets of CO, AtP5CS2 involved in proline biosynthesis, and ACS10, encoding a potential 1-aminocyclopropane-1-carboxylic acid (ACC) synthase involved in the production of ethylene. AtP5CS2 is apparently essential for the elongation of the internodes that occurs upon flowering in Arabidopsis (bolting), as reduction in AtP5CS2 expression in transgenic plants eliminated this response (Nanjo et al. 1999). Although ethylene plays an obvious role in flowering in other species (e. g., Bromeliads), the precise role of ethylene in flowering in Arabidopsis is not known (Bernier et al. 1981). Mutants insensitive to ethylene exhibit slightly delayed flowering, but the molecular mechanism of this effect has not been explored (Guzman and Ecker 1990).
CO exists as a member of a gene "family," or group of genes that encode structurally related proteins (Ledger et al. 1996). It is possible that these CO-like (COL) genes also have a role in flowering, however Putterill et al. (1997) have noted that at least one of these genes, COL1, does not seem to be expressed at higher levels in inductive photoperiods, and to date has not been identified as important in flowering by traditional genetic analyses. In other species studied [i. e., apple (Hoon et al. 1999) and Brassica napus (Robert et al. 1998)], families of CO homologs also exist. It has been suggested that the function of CO in apple could be different from that in Arabidopsis based on the apparent abundance of mRNAs of two of the apple genes in the developing flower and fruit (Hoon et al. 1999), but such studies are complicated by the ambiguity of the evolutionary relationships between the identified apple genes and the Arabidopsis CO and COL genes, and the fact that the spatial pattern of CO expression in Arabidopsis has not been fully established. In Brassica napus, CO homologs are found at genomic positions corresponding to quantitative trait loci (QTL) affecting flowering time, and at least one of the Brassica CO homologs is functionally homologous to its Arabidopsis counterpart, because it is able to complement the flowering-time defect conferred by the co-2 mutation when expressed in transgenic Arabidopsis (Robert et al. 1998). A gene encoding a protein closely related to CO has recently been cloned from the short-day plant Pharbitis nil through an assay designed to identify genes that are upregulated in response to inductive photoperiods (i. e., short days; J. Liu and H. Kende, pers. commun). This is an important finding because the fact that CO is upregulated in both species in response to inductive photoperiods, even though the plants are of opposite flowering habits, suggests that the molecular mechanisms that are distinct between plants of varying photoperiodic responses lie genetically upstream of CO.
In maize, ancestrally a short-day plant, the INDETERMINATE1 (ID1) gene promotes flowering in response to inductive photoperiods (Singleton 1946; Galinat and Naylor 1951). This is the only gene cloned to date that has unequivocally been demonstrated to be involved in flowering time in species other than Arabidopsis. ID1 encodes a protein containing zinc-finger motifs, suggesting that it binds DNA. It is expressed predominately in the leaf and influences flowering in a non-cell-autonomous manner, and thus possibly regulates the production of a transmissible signal (Colasanti et al. 1998). There is no strong structural homology between ID1 and any of the Arabidopsis flowering-time genes that have been cloned to date, and the family of ID1-like genes that do exist in Arabidopsis (Colasanti et al. 1998) have not yet been reported to be involved in flowering. This could indicate a significant divergence in flowering mechanisms between Arabidopsis and maize.B. Non-photoperiodic Induction: The Autonomous Pathway
As mentioned previously, loss of function of photoperiodic pathway genes does not prevent flowering but merely delays it, suggesting that at least one redundant pathway exists. Numerous flowering-time genes have been identified that are presumed to work outside of the photoperiodic control of flowering in the so-called autonomous pathway. LUMINIDEPENDENS (LD) was one of the first flowering-time genes identified (Redei 1962), and one of the first plant genes cloned through T-DNA mutagenesis (Lee et al. 1994a). In this technique, segments of DNA of known sequence are transferred into a plant by Agrobacterium, where they integrate into the genome at random locations. Interruption of a gene by the T-DNA often results in loss of gene function, and the corresponding gene sequence can be easily cloned by simple molecular techniques (Azpiroz-Leehan and Feldman 1997). The LD gene encodes a large protein containing two interesting structural features. First, a homeodomain-- a nucleic acid-binding motif found in developmentally important proteins from yeast, plants, and animals-- is found near the amino terminus. The homeodomain in LD is highly homologous to that found within the Drosophila Distal-less protein, which functions as a developmental switch to initiate limb formation (Cohen 1990). It also closely resembles the homeodomain found in Mata1, a yeast protein that acts as one component of a heterodimeric factor that represses expression of haploid-specific genes (Johnson and Herskowitz 1985). The other interesting structural feature is an acidic carboxyl-terminal region enriched in glutamine residues and containing short, homopolymeric glutamine stretches. These structural features are common to the activation domains of known transcriptional activators such as Drosophila Antennapedia and herpes virus VP16 (Gerber et al. 1994; Triezenberg 1995). Thus, it is possible that LD acts as a transcriptional regulator. Consistent with this proposed role, the LD protein contains nuclear localization signals and is localized to the nucleus. LD is expressed ubiquitously throughout the plant, with a concentration of mRNA expression in proliferating tissues, including the shoot, root, and floral apices (Aukerman et al. 1999).
The function of LD may have diverged through evolution. An orthologous gene has been characterized from maize (van Nocker et al. 2000). The maize LD gene is highly homologous to its Arabidopsis counterpart, containing both the homeodomain and the potential transcriptional activator region, and exhibits an analogous mRNA expression pattern in the maize plant. However, when expressed in transgenic Arabidopsis containing an ld mutation, it does not complement the flowering-time defect, but instead causes developmental abnormalities associated with the shoot and floral meristems (van Nocker et al. 2000). What function this gene has been recruited for in maize is not presently known, but should be revealed by analyses of transposon-tagged lines.
Although the activity of most plant genes studied to date seems to be controlled predominately at the transcriptional level, recent evidence suggests that posttranscriptional control may be an important factor in the regulation of flowering. The FCA gene was cloned and found to encode a large protein containing RRM motifs thought to mediate binding to RNA (Macknight et al. 1997). In support of the structural suggestion of function, the FCA protein binds to RNA in vitro, with a preference for G-and U-rich sequences (Macknight et al. 1997). The FCA gene produces multiple transcripts as a result of alternative splicing and transcriptional termination. Only one of these, which is a minority of FCA transcripts, would encode the presumed, full-length protein, and splicing to produce the full-length "active" FCA mRNA is likely to be regulated, as high-level expression of the genomic FCA sequence in transgenic plants resulted in only a minor increase in the amount of active mRNA (Macknight et al. 1997). Interestingly, it appears that FCA is able to promote flowering in a cell non-autonomous manner, because flowering is not delayed in periclinal chimaeras that express FCA only in the epidermal cell layers (Furner et al. 1996).
In addition to the RNA-binding motifs, the presumed active form of the FCA protein contains a region designated the WW motif that contains two closely spaced tryptophan residues (Bork and Sudol 1994). This is potentially an essential component of the FCA protein, as it is excluded from the protein encoded by the strong fca-1 allele (Macknight et al. 1997). In other systems, WW motifs mediate interactions with protein partners containing proline-rich regions (Kay et al. 2000).
Proteins or RNAs that interact with FCA have not been identified. One possibility is the protein encoded by FY. Mutations in FY do not further enhance the late flowering conferred by loss of FCA function (Koornneef et al. 1998a), suggesting that the two gene products operate in close proximity. In contrast, mutations in two other autonomous pathway genes, FPA and FVE, greatly enhance the lateness of fca (and fy) mutants. This suggests some redundancy in the mechanism of the autonomous pathway. However, as previously cautioned, this type of genetic analysis is contingent on mutations creating a complete loss of function, and even in cases where the gene has been cloned, this is difficult to demonstrate. FPA was recently cloned and, like FCA, encodes a protein containing RRM-type RNA-binding motifs (R. Amasino, pers. commun.). The cloning of FY, FVE, and another autonomous pathway gene, FLD, have not yet been reported.
Semidominant mutations in the SHORT VEGETATIVE PHASE (SVP) gene confer photoperiod-sensitive early flowering (Hartmann et al. 2000). SVP encodes a MADS-box transcription factor that is expressed in the apical meristem during the vegetative phase, but apparently not in the inflorescence apical meristem. This expression pattern is consistent with its role as a repressor of flowering. SVP mRNA is also expressed in the early floral meristem, suggesting a role in flower development. However, loss of SVP function confers no gross floral defects, indicating that a function at this stage is redundant or minor (Hartmann et al. 2000). The genetic relationships between SVP and other autonomous pathway genes have not been characterized.C. Vernalization
In many plants, flowering can be accelerated or induced by exposure to a long period of near-freezing temperatures. This is a commonly employed reproductive strategy that allows for flowering and seed production in the environmentally favorably period following natural winter. This phenomenon, termed vernalization, has been studied for decades at the physiological level but only recently at the molecular level. The lack of molecular work addressing vernalization is partly due to the fact that in Arabidopsis thaliana, the commonly utilized laboratory strains flower soon after germination, and extended cold treatments do little to further abbreviate the vegetative phase (Koornneef et al. 1998b). However, most natural ecotypes of Arabidopsis behave as winter annuals, flowering extremely late in the absence of cold, but very early when exposed to cold for extended periods. The flowering habit among natural ecotypes is largely determined by allelic variation at two loci, designated FRIGIDA (FRI) and FLOWERING LOCUS C (FLC); (Lee et al. 1993; Koornneef et al. 1998b). "Early" alleles at either loci behave similarly to presumed null alleles created by induced mutation, suggesting that natural early alleles have lost function (Michaels and Amasino 1999). FLC is expressed predominately in the vegetative apex and roots, but is absent from the inflorescence apex. Expression of FLC mRNA is apparently not significantly decreased as the plant proceeds through the vegetative phase, suggesting that repression of flowering by FLC can be overcome by developmental progression (Sheldon et al.1999). FLC encodes a MADS-box-containing protein (Michaels and Amasino 1999; Sheldon et al. 1999). Because other MADS-box proteins are known to work as heterodimers (Trobner et al. 1992), it is possible that FLC has a DNA-binding partner. One possibility is SVP, as both genes are expressed in the shoot apical meristem specifically during the vegetative phase. However, unlike those of FLC, mRNA levels of SVP are not diminished after extended cold treatments (Hartmann et al. 2000).
The activity of FLC is semidominant, and transgenic plants containing extra copies of the FLC genomic sequence never flower without cold, acting in essence as biennials (Michaels and Amasino 1999; Sheldon et al. 1999). Importantly, these findings suggest that the difference in flowering habit between winter-annual plants and biennial plants could be quantitative rather than qualitative. The cloning of FRI has recently been reported; this gene encodes a protein that does not exhibit significant sequence identity to any other protein of known function (Johanson et al. 2000).
In Arabidopsis, a genotype conferring the winter-annual habit can also be synthesized by impairing the function of the promotive autonomous pathway genes (Koornneef et al. 1991), and, like repression of flowering imposed by FRI, the block to flowering resulting from the loss of autonomous-pathway gene function is also dependent on FLC activity (Lee et al. 1994b; Koornneef et al. 1994; Sanda and Amasino 1996a, b). These data suggest that the flowering-repressive activity of FLC is both positively regulated by FRI and negatively regulated by autonomous pathway genes. Consistent with this idea, FLC mRNA expression is increased both in genotypes containing late FRI alleles, and in autonomous-pathway gene mutants (Michaels and Amasino 1999; Sheldon et al. 1999). FLC mRNA expression is decreased after extended cold exposures (Michaels and Amasino 1999), suggesting that vernalization involves molecular events "upstream" from FLC. The winter-annual habit conferred by loss of autonomous-pathway gene function is not dependent on FRI, indicating that neither the activity of FRI nor the autonomous pathway genes is necessary for the vernalization response. Thus, although these genes set up a requirement for cold for flowering, they are unlikely to be directly involved in the associated cold signal transduction.
Specific components involved in transmitting the signal from the cold stimulus to FLC expression have not yet been identified. Using a genetic approach, Chandler et al. (1996) identified at least two loci, des-ignated VRN1 and VRN2, that could play such a role. These mutants were isolated in an fca mutant background based on a lack of vernal-ization response. FLC mRNA levels are only partially decreased after cold treatment in these mutants, consistent with the idea that VRN1 and VRN2 function upstream to regulate FLC expression (Sheldon et al. 2000). Ishitani et al. (1998) identified a recessive mutation, hos1-1, that constitutively activated gene expression from a cold-responsive promoter. hos1-1 plants exhibited accelerated flowering, suggesting that HOS1 might normally act as a negative regulator of vernalization. However, these results were difficult to interpret because hos1-1 conferred pleiotropic effects on growth, and the specific genetic background utilized (C24) is normally early flowering due to an "early" FLC allele (Sanda and Amasino 1996a). The well characterized cold-regulated (COR) genes involved in the process of acclimation probably have little or no role in vernalization, as freezing tolerance is not affected in the vrn mutants (Chandler et al. 1996), and constitutive expression of members of the CBF family of transcriptional activators upregulates COR gene expression in a winter annual line in the absence of cold, but has no effect on flowering time (J. Liu and S. van Nocker, manuscript submitted).
Some characteristics of vernalization, including the requirement for cell division for the vernalized state to be attained and the stability of the vernalized state through mitosis, suggests an epigenetic mechanism (Wellensiek 1964). One possibility is the covalent modification of DNA through cytosine methylation. Evidence for the involvement of DNA methylation in the vernalization response has been presented by Burn et al. (1993) and Brock and Davidson (1994), who found that the promotion of flowering by extended cold in Arabidopsis and wheat, respectively, could be partially substituted for by exposure of plants to the ribonucleotide analog 5-azacytidine (5-azaC). Treatment with this compound results in demethylation of DNA. In the study by Burn et al. (1993), flowering was reportedly accelerated only in genotypes that are known to exhibit a strong vernalization response. Thus, the partial substitution for cold treatment conferred by 5-azaC apparently acted specifically upon the vernalization pathway. It was hypothesized that extended cold results in the selective demethylation and transcriptional activation of floral-promotive genes (Finnegan 1998).
A further possible link between DNA methylation and vernalization was hypothesized by Finnegan et al. (1996), who reported that antisense expression of the METHYLTRANSFERASE1 (MET1) gene in transgenic Arabidopsis resulted in both decreased genomic DNA methylation levels and early flowering. This early flowering was apparently associated with decreased FLC mRNA abundance (Sheldon et al. 1999), again suggesting specificity for the vernalization pathway. In contrast to these results, Ronemus et al. (1996) found that MET1 antisense expression conferred highly pleiotropic effects, including slightly delayed flowering, in transgenic Arabidopsis. The apparent contradictions between these two reports could reflect differences in environmental conditions or genetic backgrounds used. However, neither group utilized a genetic background that exhibits a strong vernalization response. Other aspects of these results should also be interpreted with caution. Goto and Hamada (1988) and Chandler and Dean (1994) demonstrated that growth of plants on the nucleotide analog 5-bromodeoxyuridine, which does not result in reduced DNA methylation, could also accelerate flowering.D. Induction by Gibberellins
The influence of GAs on flowering in many plants is well known (Lang 1965; Zeevaart 1983). In long-day rosette plants such as Arabidopsis, GAs generally have an inductive effect, and this is especially striking in Arabidopsis where flowering is delayed by growth in short days, or in winter-annual genotypes grown in the absence of cold. Consistent with this, flowering is delayed in the ga1 mutant that is defective in GA biosynthesis, and in the gai mutant, which is insensitive to GAs (Koornneef and van der Veen 1980; Koornneef et al. 1985). In addition, plants carrying mutations in the SPINDLY (SPY) gene, which exhibit a constitutive GA response, flower early (Jacobsen and Olszewki 1993).
Exogenously applied GA is able to promote flowering in all late mutants studied, and mutations in GA biosynthesis or perception are interactive with all flowering-promotive genes studied, especially those grouped into the photoperiodic pathway (Putterill et al. 1995; Simpson et al. 1999; our independent observations). Consistent with the strongly interactive effect with photoperiod pathway genes, ga mutant plants are apparently unable to flower when grown in short days, and gai plants flower extremely late under such conditions (Wilson et al. 1992). Thus, it appears that the production of GAs represents an additional pathway to flowering that operates in parallel with the photoperiodic pathway, and, to some extent, the autonomous pathway as well.
The role of GAs in vernalization is unclear. Although GAs are able to promote flowering in winter-annual genotypes (thereby bypassing the requirement for cold), winter-annual genotypes containing the ga or gai mutations still exhibit a normal vernalization response (R. Amasino, pers. commun.; Chandler et al. 2000). This would suggest that GAs are not necessary for vernalization. However, as GA production is not completely eliminated in the ga mutant (J. A. D. Zeevaart, pers. commun.), such results should be interpreted with caution. In addition, although GAI has an obvious role in GA signal transduction during vegetative growth, other yet unidentified GA-signaling components could be involved in the flowering response.IV. ROLE OF CARBOHYDRATES
Carbohydrates have long been known to play a key role in flowering (Bernier et al. 1993). The concentration of sucrose, the major translocated sugar in most plants, increases dramatically in phloem exudates upon photoinduction in both short-day and long-day plants, even when the photoinductive treatment does not result in a net increase in photosynthesis (Bodson and Outlaw 1985; Houssa et al. 1991; Corbesier et al. 1998). One of the earliest biochemically detectable changes in the shoot meristem upon photoinduction is the accumulation of sucrose (Bodson and Outlaw 1985) and labeling experiments suggest that this sucrose originates not from increased photosynthesis, but from mobilization of sugars from reserve carbohydrates such as starch in the leaves and stem (Bodson et al. 1977).
Arabidopsis will flower in complete darkness if the aerial portion of the plant is supplied with sucrose or glucose (Redei et al. 1974; Goto 1982; Araki and Komeda 1993). Under such conditions, the late-flowering phenotype conferred by mutations in GI, CO, FCA, FPA, and FVE was complemented or nearly complemented. In contrast, flowering was not promoted by these conditions in plants carrying mutations in FWA or FT (Araki and Komeda 1993; Roldan et al. 1999). These surprising results suggest that the fundamental mechanism of both the photoperiodic and autonomous pathways could be the delivery of sugars to the shoot apex! Sucrose is synthesized in the cytosol from the products of photosynthesis or starch degradation, transported to and loaded into the phloem, translocated throughout the plant, unloaded from the phloem, and then transported from cell to cell. This complicated routing provides many opportunities for control of sugar transport, and thus it is likely that many genes are involved.
Interestingly, mutations in the GI gene are pleiotropic in that mutants accumulate excess levels of starch in the leaves and stem (Eimert et al. 1995). At least one other Arabidopsis mutant that was originally identified as a starch accumulator, carbohydrate accumulation mutant1 (cam1), was found to flower late relative to wild-type plants, especially when grown under continuous light (Eimert et al. 1995). High starch con-tent per se does not seem to be the direct cause of the flowering-time defect, because the flowering-time defect conferred by gi and cam1 was not rescued in genetic backgrounds where starch synthesis was disrupted (Eimert et al. 1995). Other mutants that lack starch, ADP-glucose pyrophosphorylase1 (adg1) and phosphoglucomutase1 (pgm1), and at least one other mutant that accumulates starch, starch-in-excess1 (sex1), also exhibit delayed flowering, but only under photoperiods of less than 16 h (Lin et al. 1988; Caspar et al. 1985; Caspar et al. 1991; Corbesier et al. 1998). The observation that both the overabundance of starch, and lack of starch, can affect flowering in a similar manner further suggests that flowering is not directly affected by starch content. In fact, the lack or excess of starch in the pgm1 and sex1 mutants, respectively, seems to disrupt carbohydrate metabolism in a similar manner, as in both mutants soluble sugars (including sucrose) accumulate to abnormally high levels (Caspar et al. 1985; Caspar et al. 1991). Thus, it seems probable that the predominant effector of flowering in these mutants is the levels of sugars. Arabidopsis plants grown at low temperature also accumulate soluble sugars, and this may be related to the delayed flowering seen under these conditions.V. CONTROL OF MERISTEM IDENTITY
The shoot and flower are, in spite of their radical difference in morphology, essentially analogous structures produced by the meristem. The fate of meristems-- to generate flowers rather than shoots-- is gov-erned by a group of meristem identity genes, which are activated during the transition to flowering. This group of genes in turn controls expression both of the floral organ identity genes, which control the development of the floral organs, and cadastral genes, which regulate the boundaries of expression of the organ identity genes. The molecular biology of flower development is beyond the scope of this review, and the reader is referred to recent excellent discussions on this topic (Bowman 1997; Sessions et al. 1998).A. Meristem Identity Genes
As with genes influencing the timing of flowering, genes involved in influencing meristem identity have been identified by screening for mutants in which meristem identity is altered. Such screens have identified genes that both positively and negatively regulate the shoot-to-flower transition. In Arabidopsis plants homozygous for the recessive terminal flower 1 (tfl1) mutation, the normally indeterminate inflorescence terminates in a single flower, and lateral shoots develop as soli-tary flowers (Shannon and Meeks-Wagner 1991; Alvarez et al. 1992).
Thus, a presumed function of TFL1 is to keep the inflorescence meristem in an indeterminate state. Plants lacking TFL1 activity also flower slightly early, suggesting that TFL1 functions during the vegetative phase as a repressor of the shoot-to-inflorescence transition (Shannon and Meeks-Wagner 1991; Schultz and Haughn 1993). TFL encodes a member of a small protein family exhibiting limited homology to mammalian Raf kinase inhibitor protein (RKIP). RKIP is a membrane-associated protein that regulates Raf-1 kinase, which is intimately involved in signal transduction cascades controlling cell proliferation and differentiation in mammals (Ferrell 1996). The amino-terminus of RKIP is cleaved off to form a small peptide hormone, leading to the speculation that TFL may in a similar manner be the progenitor of a small signaling peptide involved in flowering (Bradley et al. 1997). That intercellular signaling should be involved in flowering is expected, as the meristem must function as a unit to organize flower primordia even though it is composed of clonally unrelated cells (see below). Constitutive expression of the TFL1 gene in transgenic Arabidopsis confers a phenotype that is essentially opposite to that seen in tfl1 mutants-- such plants exhibit delayed flowering, and produce secondary inflorescences that are not subtended by cauline leaves (Ratcliffe et al. 1998). Because Arabidopsis flowers are not normally found in association with leaves (bracts), such structures can be interpreted as a conversion of flowers to inflorescence shoots.
Conversion of flowers to shoots is also seen in plants carrying loss-of-function mutations in a group of genes best typified by LEAFY (LFY). In plants carrying strong lfy alleles, early-arising (basal) flowers are completely transformed into shoots, whereas those that develop in more apical positions exhibit partial floral character. In plants carrying very weak lfy alleles, secondary shoots subtended by cauline leaves develop at the first few positions normally occupied by flowers (Schultz and Haughn 1991).
That flowers eventually do develop even in the absence of LFY activity indicates that other genes function in a partially redundant manner to promote the inflorescence-to-floral switch. One of these is APETALA1 (AP1). Loss of AP1 function phenocopies very weak lfy alleles with respect to inflorescence structure, but dramatically enhances the phenotype of lfy plants, such that in lfy/ ap1 double mutants, even the most apical nodes produce structures with strong shoot characteristics (Bowman et al. 1993; Huala and Sussex 1992; Weigel et al. 1992). Strong ap1 alleles also confer a striking floral phenotype-- sepals found in the outer whorl of the flower exhibit leaf-like characteristics, and often subtend secondary flowers (Irish and Sussex 1990). This phenotype can be inter-preted as a partial reversion of the flower into a shoot, further implicating AP1 in meristem identity (Mandel et al. 1992). Strong constitutive expression of both LFY and AP1 in transgenic plants results in premature transformation of the shoot into a flower, mimicking loss of TFL function (Weigel and Nilsson 1995; Mandel and Yanofsky 1995). Incredibly, Arabidopsis LFY is able to accomplish this even in a divergent tree species, aspen, suggesting a high degree of conservation of meristem identity function during evolution (Weigel and Nilsson 1995).
That loss-of-function mutations in genes such as LFY and AP1 exhibit additive phenotypic effects when combined is evidence that at least two pathways are normally involved in establishing the floral meristem (Shannon and Meeks-Wagner 1993). The genetic evidence indicating that these pathways are partially redundant is reinforced by experiments showing that the shoot-to-flower conversion conferred by 35S: AP1 is not dependent on LFY, and that mutations in AP1 cannot fully suppress this effect in 35S: LFY plants (Mandel and Yanofsky 1995; Weigel and Nilsson 1995). However, these two pathways are strongly interactive. In primordia destined to become flowers, LFY mRNA expression precedes that of AP1, and AP1 upregulation is delayed in lfy plants (Simon et al. 1996; Hempel et al. 1997; Liljegren et al. 1999) . In addition, ectopic activation of LFY activity results in premature AP1 expression (Parcy et al. 1998). These findings suggest that LFY acts as a positive reg-ulator of AP1. Conversely, LFY is expressed prematurely in primordia of 35S: AP1 plants, suggesting a reciprocal positive regulation between the two genes (Liljegren et al. 1999).
The protein encoded by LFY does not resemble any other known protein (Weigel et al. 1992), but numerous lines of evidence indicate that it acts as a transcription factor. LFY protein is localized to the nucleus and is able to mediate transcriptional activation in yeast when fused to a suit-able activation domain (Parcy et al. 1998). Consistent with the genetic evidence that LFY positively regulates AP1 expression, the LFY protein binds to a potential AP1 promoter element in vitro (Parcy et al. 1998). Moreover, Wagner et al. (1999) demonstrated through the use of an inducible LFY-GR fusion protein that LFY is able to rapidly activate AP1 expression even in the presence of cycloheximide, suggesting a direct interaction.
Another gene with a promotive role in floral identity is CAULIFLOWER (CAL). The function of this gene is apparently completely redundant with that of AP1, such that in an otherwise wild-type genetic background, plants lacking CAL activity appear normal. However, in plants lacking both CAL and AP1 activity, meristems never develop determinate floral character and continue to proliferate, and inflorescences develop into structures that resemble tiny versions of the garden vegetable for which the gene is named (Bowman et al. 1993). The phenotypic similarity between Arabidopsis cal/ ap1 double mutants and cauliflower led Kempin et al. (1995) to investigate functional conservation of CAL genes in the two species. In both Brassica oleracea, and its cauliflower derivative (Brassica oleracea, var. botrytis) CAL is expressed in a spatial and temporal pattern similar to that seen with CAL in Arabidopsis. However, the gene from var. botrytis encodes a protein that is significantly truncated and is probably nonfunctional (Kempin et al. 1995), suggesting a molecular explanation for what is probably the most popularly recognized natural inflorescence variation. Both AP1 and CAL encode proteins containing a MADS-box domain, consistent with roles as transcription factors (Mandel et al. 1992; Kempin et al. 1995).
The essentially opposite phenotypes conferred by loss of TFL andLFY/ AP1/ CAL function indicates that these genes operate antagonistically, and molecular studies support this conclusion. Expression of both TFL and LFY mRNAs is maintained at a low level in the shoot apex during the vegetative phase, and is upregulated upon the transition to flowering. However, their expression is spatially separated, with TFL mRNA present in the center of the meristem, and LFY mRNA present only in emerging primordia (Bradley et al. 1997). In plants lacking LFY, AP1, or CAL activity, TFL expression extends into the lateral primordia (Ratcliffe et al. 1999), whereas in plants constitutively expressing LFY, TFL expression is greatly decreased (Ratcliffe et al. 1999; Liljegren et al. 1999).
Like AP1 and CAL1, at least two other genes are known to have redundant roles in promoting floral identity. Mutations in APETALA2 (AP2), for example, do not confer strong flower-to-shoot conversion, but instead enhance both lfy and ap1 phenotypes (Shannon and Meeks-Wagner 1993). Plants carrying mutations in the UNIDENTIFIED FLORAL ORGANS (UFO) gene resemble weak lfy mutants in that basal inflorescence nodes exhibit some shoot identity (Levin and Meyerowitz 1995; Wilkinson and Haughn 1995). In addition, in short-day-grown ufo plants, the transition to an inflorescence meristem is incomplete, and reversion to a vegetative meristem can occur (Wilkinson and Haughn 1995). Like the aforementioned FKF1 and ZTL, UFO encodes an F-box-containing protein, implicating this factor in the selective elimination of other, possibly regulatory proteins (Ingram et al. 1995).
The AGAMOUS gene, better known for its role in governing floral organ identity in the inner whorls, also functions to promote floral meristem identity by regulating meristem determinacy. The normally determinate floral meristem becomes indeterminate in ag mutant plants, and under short-day conditions can even completely revert to an inflorescence meristem (Yanofsky et al. 1990; Mizukami and Ma 1995, 1997). B. Integration of Flowering Pathways and
Genes affecting the timing of flowering are obvious candidates for direct regulators of meristem identity genes, and recent genetic and molecular studies have demonstrated numerous interactions between the two classes of genes. Simon et al. (1996) utilized the inducible CO-GR expression system described above to examine the response of the LFY, AP1, and TFL1 genes to increased CO activity. Within 24 h of CO: GR activation in plants grown in inductive photoperiods, LFY and TFL1 mRNAs accumulated to detectable levels, and detectable AP1 mRNA upregulation occurred after ~72 h. The delayed expression of AP1 relative to that of LFY is in accordance with genetic evidence that LFY acts upstream of AP1 (above). These kinetics were similar to those seen upon the trans-fer of short-day-grown plants to inductive photoperiods (Simon et al. 1996). In contrast, when CO-GR was activated in short-day-grown plants, LFY and TFL1 were again activated within 24 h, but the delay in the upregulation of expression of AP1 was extended to ~120 h. The authors concluded that since inductive photoperiods were more effective than CO to activate AP1, an additional, unknown mechanism operating in inductive photoperiods is involved in the upregulation of AP1 (Simon et al. 1996). In fact, the flowering time genes FT and FWA (see below) may play a role that is redundant with that of LFY in activating AP1. Evidence for this was presented by Ruiz-Garcia et al. (1997), who showed that the phenotype conferred by a strong allele of lfy is greatly enhanced in an ft or fwa mutant background.
It is known that GAs promote flowering at least in part through upregulation of LFY, because in the ga1 mutant, LFY promoter activity is reduced, and its upregulation in response to inductive photoperiods is delayed relative to wild-type plants. In addition, 35S: LFY expression can partially complement the flowering defect conferred by ga1 in short day conditions (Blazquez et al. 1998). Recently, Blazquez and Weigel (2000) demonstrated that distinct cis elements in the LFY promoter mediate the induction of LFY in response to GAs or inductive photoperiods. Thus, it appears that LFY represents an integration point of at least the photoperiod pathway and the GA pathway.VI. COMPETENCY
A multitude of physiological studies has indicated that flowering is dependent not only on the ability of the leaf to produce the floral stimulus, but also on the ability of the shoot apical meristem to respond to it (Lang 1965). The shoot apical meristem is a group of specialized cells found within the apex at the growing tip of shoots. In Arabidopsis, as in other plants, the meristem displays typical tunicacorpus organization, recognizable as early as the torpedo stage of embryogenesis (Long et al. 1996). The tunica layers (L1 and L2) are propogated through anti-clinal cell divisions, and thus the L1, L2, and L3 tend to be clonally unrelated. The meristem can display an additional level of organization that is superimposed on the tunica-corpus structure. This consists of radially symmetric "zones" that are often distinguished by mitotic activity and cell size and density (Vaughan 1952; Steeves and Sussex 1989). The central initiation zone, at the summit of the meristem, is characterized by a group of large cells with prominent vacuoles that apparently divide very slowly. Flanking the central initiation zone is a ring of smaller, more densely cytoplasmic, proliferative cells termed the peripheral zone. In the peripheral zone, groups of cells are recruited into leaf or flower pri-mordia where they may soon assume specialized roles. Immediately subtending the central initiation zone is a group of proliferative cells referred to as the rib, or file, meristem that produces the internal tissues of the plant stem.
When given inductive photoperiods, the commonly studied annual genotypes of Arabidopsis flower soon after germination. Before the tran-sition to flowering in such young plants, the apical meristem is small and without easily recognized cytological zonation. However, when the vegetative phase is extended (i. e., through growth in short-day photoperiods, or in winter-annual genotypes lacking cold treatment) the zonal pattern described above becomes more apparent (Vaughan 1952; Besnard-Wibaut 1981). The appearance of well-defined zonal organization in Arabidopsis has been correlated with the ability of the plant to exhibit a significant flowering response to applied GAs or long days (Besnard-Wibaut 1981), suggesting that appropriate meristem structure could constitute the morphological basis of competency.
The molecular determinants of competency remain unknown. It is important to note that although transgenic Arabidopsis expressing LFY in a constitutive manner flower very early, they still progress through a short vegetative phase (Weigel and Nilsson 1995). That LFY is insufficient to force early-arising primordia into a floral fate suggests that genes controlling meristem competence act genetically downstream, or in a separate pathway, from LFY. Two candidates are FWA and FT. Mutations in these genes cause delayed flowering, and are epistatic to a 35S: LFY transgene [i. e., constitutive expression of LFY is unable to rescue that late-flowering phenotype conferred by fwa and ft mutations (Nilsson et al. 1998)].
All known mutations in FWA are dominant. Dominance of mutations often indicates that the mutant gene product has gained activity. This could happen if the gene product were increased in abundance, or if it usually existed in the "off" position and were turned "on" by the mutation in a manner similar to the action of an upstream signaling molecule. Indeed, FWA mRNA levels were found to be increased in fwa mutants relative to wild-type plants, suggesting that the dominance of the muta-tion indeed results from increased FWA activity. Interestingly, it appears that the known mutant alleles of fwa result from epimutation, a class of mutation that does not disrupt the DNA sequence. In the case of fwa mutations, constitutive expression of the gene is associated with a reduc-tion in methylation of DNA residues found in the promoter region of the gene (W. Soppe, pers. commun.). The FWA gene was recently cloned and found to encode a transcription-factor-like protein (M. Koornneef and W. Soppe, pers. commun,).
Because a gain in function of FWA results in later flowering, the function of the wild-type FWA gene is likely to repress flowering. Alleles of FT conferring late flowering are recessive and likely result from decreased function, suggesting that FWA acts in a manner opposite that of FT. The Arabidopsis FT gene was recently cloned by activation tagging (Kobayashi et al. 1999; Kardailsky et al. 1999). In this approach, ran-dom plant genes are transcriptionally activated by the nearby insertion of T-DNAs containing strong enhancer sequences, and function of the activated gene is surmised based on the resulting phenotype. Activation tagging has become a powerful technique for the identification of genes whose products are normally limiting in a pathway affecting a given phenotype (Lindsey et al. 1998). The product of the FT gene is structurally related to that of TFL, suggesting that like TFL, the FT protein may function in cell-to-cell signaling (Kardailsky et al. 1999). FT mRNA is expressed throughout the aerial tissues of the plant, and is not localized to any specific domain within the shoot apex (Kobayashi et al. 1999; Kar-dailsky et al. 1999). This expression becomes evident only near the floral transition, and, consistent with being regulated by CO (see above), this upregulation of expression is delayed in the co mutant and in short-day conditions. However, upregulation of expression still occurs in plants lacking CO activity, indicating that FT is also regulated by another mechanism. As might be expected from a gene controlling meristem competence, constitutive expression of FT in transgenic plants results in a nearly complete elimination of the vegetative phase, as such plants form a flower after only ~2 leaves, which are embryonic in origin.
Other genes that have been postulated to play a role in competence include EMBRYONIC FLOWER (EMF) 1 and 2. Plants carrying mutant alleles of these genes produce a modified flower upon germination (Sung et al. 1992; Bai and Sung 1995; Yang et al. 1995; Chen et al. 1997). Plants carrying strong emf1 alleles develop no leaves at all, indicating that the vegetative phase has been completely bypassed. Mutations in EMF2 confer a milder but similar phenotype, with a few small leaves produced on a modified inflorescence. The very early flowering associated with loss of EMF function suggests that the EMF genes normally act as strong repressors of reproductive development. It is possible that the two genes operate in distinct genetic pathways, as combining strong emf1 and emf2 alleles leads to severe developmental defects that could be considered an additive phenotype (Yang et al. 1995).
Although cloning of the EMF genes has not been reported, some indication of their mode of action comes from examining interactions of emf mutations with other mutations affecting the timing of flowering. Mutations in GI or CO, for example, have no effect on emf1 or emf2 phenotype; that CO and GI are not required for the expression of the emf mutant phenotype suggests that CO and GI act as upstream regulators of the EMF genes. However, mutations in other flowering-time genes, including FCA, LD, FVE, FY, FHA, FPA, FWA, and FT resulted in a par-tial rescue of the early flowering emf phenotype, with mutations in FWA and FT having the greatest effect (Huang and Yang 1998; Page et al. 1999). These results suggest that the EMF genes, rather than having a general repressive effect on flowering, act specifically as downstream negative regulators of the photoperiodic pathway.
Finally, it is interesting to note that constitutive expression of LFY was not sufficient to fully rescue the flowering defect conferred by the ga mutation in short days (Blazquez et al. 1998). This suggests that in addition to playing a role in LFY activation, gibberellins play a role in promoting competency.VII. CONCLUSION AND PERSPECTIVES
Research into the molecular mechanisms of flowering is now entering its second decade. Many flowering genes have been identified in the model plant Arabidopsis thaliana, and most of these have now been cloned. In addition, the functions of most of the known flowering-time genes have been assigned within one of the multiple parallel pathways that promote flowering in this plant (Fig. 1.1). However, the research accomplished to date should be considered merely as a foundation for future work, as many aspects of flowering in this plant remain unexplored. For example, in many cases, the relationships and interactions among the genes in these pathways have been surmised based solely on genetic evidence, and need to be confirmed with molecular and biochemical data. Some more general questions remain to be answered as well. For example, in spite of the wealth of physiological data suggesting that flowering is the result of a transmissible signaling substance (" florigen"), produced in the leaves and acting at the shoot apex, the molecular identity of this signal is still unknown.
The possibility that flowering can be manipulated in species other than Arabidopsis through molecular methods is largely dependent on the degree to which flowering-time mechanisms have been conserved through evolution. In order to prove their utility in solving horticultural problems, the models proposed to describe floral initiation based on genetic and molecular studies in Arabidopsis will likely need to be evaluated in plants with dissimilar flowering habits. With very few excep-tions (see above), this is an area of research that has remained largely unexplored. Probably the most attractive opportunity in this regard is maize, which has traditionally been the most popular monocot model for developmental studies, and has diverged significantly from Arabidopsis in terms of genetics, physiology, and anatomy. Current efforts underway to discover maize gene function by high-throughput, reversed-genetics approaches should greatly simplify this task (Martienssen 1998; Walbot, 2000). The apparent conservation of function of meristem identity genes among species offers some indication that the function of flowering-time genes may be similarly conserved. As the intriguingly complex pathways that constitute flowering become more completely characterized in Arabidopsis, this area holds much promise for future research.
|Dedication : Margaret Sedgley|
|1||Analyzing fruit tree architecture : implications for tree management and fruit production||1|
|2||Peach orchard systems||63|
|3||Irrigation scheduling and evaluation of tree water status in deciduous orchards||111|
|4||Leucadendron : a major proteaceous floricultural crop||167|
|5||Chinese Jujube : botany and horticulture||229|
|6||Taxus spp. : botany, horticulture, and source of anti-cancer compounds||299|
|7||The genus Allium : a developmental and horticultural analysis||329|
|8||The invasive plant debate : a horticultural perspective||379|
Note: The Figures and/or Tables mentioned in this sample chapter do not appear on the Web.Chapter 1
In vitro propagation is based on enhanced axillary bud proliferation and on the ability of differentiated, often mature plant cells, to redifferentiate, and develop new meristematic centers that are capable of regenerating fully normal plants. Regeneration is potentiated through two morphogenic pathways: organogenesis-the formation of unipolar organs, and somatic embryogenesis-the production of bipolar structures, somatic embryos with a root and a shoot meristem (Steward et al. 1970; Ammirato 1985).
All plant somatic cells once isolated and cultured in vitro are capable of expressing totipotency. The injured cells in the outer layers of the isolated explant evolve ethylene that induces the inner layers of cells to undergo dedifferentiation. After the loss of coordinated control, which is ensued by cell division, the formation of new gradients of endogenous phytohormones in the dedifferentiated cells enhance cell divisions in response to the various growth regulators added to the medium. Cell division can take place in an unorganized pattern with callus formation, or in an organized pattern with the formation of meristematic centers directly in the explant tissues. Redifferentiation in the callus tissue takes place when new physiological gradients are formed in the non-organized pa-renchymatous tissue along with the formation of meristemoids (in vitro meristems), which can further differentiate into organized structures. In several plant species, meristematic centers form directly on the explant with very little or no callus formation and can develop into either shoots or somatic embryos. Plant regeneration can follow after the initial stages of controlled redifferentiation through either the organogenic or embryogenic pathways (Ammirato 1983, 1985; Ziv 1999).
Plant regeneration in vitro is dependent on the manipulation of the inorganic and organic constituents in the medium, as well as the type of explant and the species. In most plants, successful regeneration from the callus or directly from the explants takes place after a series of sub-cultures in various media, in a sequence which is often specific to the species, variety, or the newly introduced genotype. The determining factors are the combination of the concentration in relation to medium-volume and the composition of growth promoting and retarding regulators in the medium, the physiological status and competence of the cells and their capability for morphogenetic expression (Christianson 1985, 1987).
Micropropagation (in vitro propagation of axillary and/ or adventitious buds as well as somatic embryos) is presently used as an advanced biotechnological system for the production of identical pathogen-free plants for agriculture and forestry (George 1996). The technique, however, is still costly due to intensive hand manipulation of the various culture phases and is not used commercially for all plant species. In addition, in some plants, the initial stage of establishment and response is slow and the survival of the plants in the final stage ex vitro is often poor, which further reduces the micropropagation production potential (Ziv 1995a; Aitken-Christie et al. 1995).
Efficient commercial micropropagation depends on rapid and extensive proliferation along with the use of large-scale cultures for the multiplication phase. Furthermore, normal plant development during the acclimatization and hardening stage is mandatory to ensure a high percent of survival after transplanting to the greenhouse (Preece and Sutter 1991; Ziv 1991a, b, 1995b). Mechanization and automation of the micropropagation process can greatly contribute to overcoming the limitation imposed by existing conventional labor-intensive methods. Considerable attention has been directed toward automation of the repeated cutting, separation, subculture, and transfer of buds, shoots, or plantlets during the multiplication and transplanting phases (Levin et al. 1988; Aitken-Christie 1991, Aitken-Christie et al. 1995; Vasil 1994). Progress in tissue culture automation will depend on the use of liquid cultures in bioreactors, which will allow fast proliferation, mechanized cutting, separation, and automated dispensing (Alper et al. 1993; Sakamoto et al. 1995). These techniques were reported for some plants and were shown to reduce hand manipulation and thus reduce in vitro plant production costs (Levin et al. 1988; Ziv 1990b, 1992a, b, 1995b; Ziv and Hadar 1991; Vasil 1994; Aitken-Christie et al. 1995). The various propagation aspects of several plant species in bioreactors and some of the problems associated with the operation of bioreactors were recently reviewed by Takayama and Akita (1998). One major problem addressed was microbial contamination as affected by both the introduced plant material and the operation procedures of large-scale bioreactors.
Liquid media have been used for plant cells, somatic embryos, and organ cultures in both agitated flasks or various types of bioreactors (Smart and Fowler 1984; Ammirato and Styer 1985; Stuart et al. 1987; Chen et al. 1987; Preil 1991; Paque et al. 1992; Scragg 1992; Attree et al. 1994; Ziv 1995b; Archambault et al. 1995; Tautorus and Dunstan 1995). Although the use of bioreactors has been directed mainly for cell suspension cultures and secondary metabolites production, research directed at improving bioreactors for somatic embryogenesis has been reported for several plant species (Styer 1985; Preil et al. 1988; Nadel et al. 1990; Scragg 1990, 1992; Hale et al. 1992; Archambault et al. 1994; Tautorus et al. 1994; Attree et al. 1994; Ziv et al. 1994; Takayama and Akita 1998; Hvoslef-Eide and Munster 1998). The cultivation in liquid media using a temporary immersion system with different frequencies of immersion was reported to improve plant quality and multiplication rates of banana, coffee, and rubber (Alvard et al. 1993; Teisson and Alvard 1995; Etienne et al. 1997). Bioreactors were used also for the cultivation of hairy roots mainly as a system for secondary metabolite production (Rodriguez-Mendiola et al. 1991; Flores 1995; Weathers et al. 1989). The use of a simplified acoustic window mist bioreactor for transformed roots and for carnation shoots was reported to improve biomass growth (Chatterjee et al. 1997).
Information on the use of bioreactors as a system for plant propagation (Table 1.1) through the organogenic pathway, although limited to a small number of plant species, is presently being applied to several ornamental and some vegetable and fruit crop plants (Takayama et al. 1991; Takahashi et al. 1992; Ziv 1992a, b; Akita and Takayama 1994; Ilan et al. 1995; Ziv and Shemesh 1996; Ziv et al. 1998).
However, liquid cultures confer several problems associated with abnormal plant development, a phenomenon described as hyperhydricity, previously termed "vitrification" (Debergh et al. 1992), which causes poor plant development in vitro and later in ex vitro. Attempts were carried out to overcome hyperhydricity in liquid cultures by the use of growth retardants, inhibitors of gibberellin biosynthesis, which decreased hyperhydricity, reduced shoot elongation, and induced bud or meristem cluster formation (Ziv 1990a, b; Ziv and Ariel 1991; Ziv 1992b; Ziv and Shemesh 1996; Ziv et al. 1998). The clusters were shown to be an alternative propagation system for bioreactor cultures, providing a biomass with limited leaf elongation. Recently micropropagation of pineapple was scaled up using clusters in a temporary immersion system (ebb and flow) in bioreactors (C. G. Borroto 1998, pers. commun.).
Bioreactors are presently being used for commercial micropropagation in the USA, Japan, Taiwan, Korea, Cuba, Costa Rica, Holland, Spain, Belgium, and France for ornamental and bulbous plants, pineapple, potato, and forest trees. The cost per propagule unit of foliage plants was estimated to be reduced from 17 to 6- 7 cents (US) (R. Levin, pers. commun.). A recent review by Takayama and Akita (1998) describes several bioreactor techniques using plant propagules for large-scale propagation of potato, gladiolus, lilies, strawberry, Hyacinth, Amaryllis, and several Araceae. The present paper reviews various aspects of large-scale liquid culture in bioreactors for plant propagation, emphasizing the physical and chemical growth factors that control development and proliferation in liquid media for efficient plant production.
II. PLANT DEVELOPMENTAL PATHWAYS IN BIOREACTORS
A. Somatic Embryogenesis
The use of liquid cultures for the cultivation of somatic cells in recapitulating embryogeny was reported by Steward et al. and Reinert as early as 1958 in carrot (Daucus carrota L.). The underlying concept for cell totipotency and the ability to renew growth and express morphogenesis was based on the assumption that isolation and bathing of the cells in a specific medium will simulate the zygotic embryo conditions in the ovary. The bathing liquid medium that contains various nutrients and growth regulators was assumed to be a source for similar stimuli that were present in the zygotic embryos' immediate environment. In addition, the segregation of embryogenic from non-embryogenic cells amplified the expression of cell totipotency (Steward et al. 1970). Somatic embryogenesis was achieved by the initial use of an induction medium containing an auxin, usually 2,4-dichlorophenoxyacetic acid (2,4-D), and coconut water; the latter was mainly a source of cytokinins, inositol, and reduced nitrogen compounds. Once the pro-embryonic clusters formed, the removal or lowering of the auxin concentration initiated a sequence of events similar to zygotic embryogeny (Halperin 1967; Ammirato 1985). Somatic embryogenesis was observed at first mainly in members of the Umbelliferae cultured in liquid media (Steward et al. 1970) and has since been reported in gymnosperms and angiosperms, including several ornamentals, and vegetable and field crops, as well as perennial woody plants.
The embryogenic pathway, as opposed to the organogenic pathway, can be a more efficient and productive system for large-scale clonal propagation. Somatic embryogenesis may result in less variation through chimerism. However, somaclonal variation will be less likely to take place in direct regeneration, but might be more pronounced in somatic embryos developing from continuously cultured callus tissue. Since the embryo contains both a root and an apical shoot meristem, the rooting stage required in conventional in vitro bud or shoot propagation technology is obviated. Somatic embryos are small and can be adequately handled in scaled-up procedures. They are amenable to sorting and separation by image analysis, dispensing by automated systems, and can be encapsulated and either stored or planted directly, with the aid of mechanized systems (Ammirato and Styer 1985; Cazzulino et al. 1991; Cervelli and Senaratna 1995; Sakamoto et al. 1995).
Somatic embryogenesis in liquid shake or bioreactor cultures was reported in carrot (Ammirato and Styer 1985; Jay et al. 1992, 1994; Archambault et al. 1995), in caraway (Ammirato 1983), poinsettia (Preil 1991), alfalfa (Stuart et al. 1985, 1987; Denchev et al. 1992; Kuklin et al. 1994), celery (Nadel et al. 1990; Saranga and Janick 1991), in Eschcholtzia californica (Archambault et al. 1994), in Nerine (Ziv et al. 1994), in Octea catharinensis (Moura Costa 1992), in sweet potato (Harrell et al. 1994), in rubber (Etienne et al. 1997), and in spruce (Tautorus and Dunstan 1995). However, the ultimate goal of production of synthetic seeds through successful encapsulation of somatic embryos from liquid shake or bioreactor cultures was reported only for carrot (Kitto and Janick 1985; Redenbaugh et al. 1991; Sakamoto et al. 1995), alfalfa (Senaratna 1992; Fujii et al. 1992), celery (Kim and Janick 1989; Onishi et al. 1992, 1994), and white spruce (Attree et al. 1994) and is currently under continuous investigation for several other species.
B. Organogenic Pathway
The propagation of most plants is presently carried out commercially through the organogenic pathway in agar-gelled cultures, even though the protocols are long and costly. The advantages for the mechanization of the process are mainly for achieving a reduction in hand manipulation and labor costs. The process has to be scaled up using liquid cultures in bioreactors to amend it to automation (Levin et al. 1988; Kurata 1995). The information on the use of bioreactors for unipolar structures such as protocorms, buds, or shoots is limited, mainly because of the problems of hyperhydricity of the leaves and shoots (Debergh et al. 1992). One of the first attempts in using liquid cultures for micropropagation of buds was reported for orchids that formed protocorms with minimal shoot elongation during the liquid culture stage (Morel 1974). The protocorms could be separated and induced to form new plants after subculture to agar-gelled medium. Several other ornamental plants also have been propagated through the organogenic pathway in liquid shake cultures and bioreactors. By controlling shoot growth and providing culture conditions that reduced abnormal leaf growth and enhanced the formation of bud or meristematic clusters, as was shown for potato, gladiolus, and Ornithogalum dubium (Fig. 1.1A, B, C), a high proliferation rate was achieved without the phenomenon of hyperhydricity (Levin et al. 1988; Ziv 1990a, 1991a, 1992a; Ziv and Hadar 1991; Ziv and Ariel 1991; Takayama 1991; Takayama et al. 1991; Takayama and Akita 1994; Ilan et al. 1995; Ziv et al. 1998).
C. Bud or Meristem Clusters
The development of spherical meristematic or bud clusters in liquid cultures was shown to provide a highly proliferative and rapid growing system amenable to automated inoculation, control of the medium components, mechanical separation, and efficient delivery to the final stage for plant growth and development (Levin et al. 1988; Ziv 1991b; Ziv et al. 1998). Cluster formation appears to be associated in most species with the continuous submergence, circulation, and agitation of the plant biomass in the medium, as well as with a balanced ratio of growth-promoting and growth-retarding regulators. Clusters can form from axillary buds and/or adventitious buds as well as from meristemoids in pro-embryogenic callus that later differentiate to somatic embryos (Fig. 1.1A, 1.2A, B).
The formation of condensed organized structures in which the shoots are reduced to buds or meristematic tissue was reported in several plant species. The clusters were made up of densely packed meristematic cells, actively dividing and forming new meristematic centers on the outer surface. The meristemoids surrounded loosely packed cells in the center and exhibited some vascularization, as was shown in liquid-cultured poplar clusters (McCown et al. 1988). In banana, on the other hand, the clusters were made up of condensed buds surrounding a central core with a cavity (Ziv et al. 1998).
Halperin in 1967 described the formation of pro-embryogenic masses in carrot root explants which were induced by the addition of 2,4-D to the liquid medium, resembling orchids protocorm-like bodies that formed when isolated buds were cultured in liquid medium in shake cultures (Morel 1974). Protocorm-like clusters also were induced in liquid-cultured gladiolus buds (Ziv 1989, 1990a) and in several species of the complex Brodiaea (Ilan et al. 1995) by the addition of paclobutrazol or ancymidol (gibberellin biosynthesis inhibitors) to the medium. In chicory, nodules were induced on leaves by IBA and BA in the medium (Pieron et al. 1992). In potato and banana, bud clusters were induced by a balanced ratio between kinetin and ancymidol in the liquid medium (Ziv et al. 1998). In woody species, McCown et al. (1988) and Aitken-Christie et al. (1988) described nodules in poplar in liquid medium and in radiata-pine in agar-gelled medium respectively, induced by the use of a balanced ratio of growth regulators. Levin et al. (1988), working with several ornamental species, described a several-fold increase of an organogenic biomass of clusters which proliferated in bioreactors and were separated mechanically prior to dispensing to agar-gelled cultures for further growth. The production of clusters in Philodendron cultured in liquid medium required the presence of benzylaminopurine (BA) and an inductive treatment for 24- 48 h with ancymidol. The inductive treatment prevented a carry-over dwarfing effect of ancymidol on leaf and shoot development after transplanting the clusters to agar-gelled medium for further plant growth (Ziv and Ariel 1991). Gladiolus clusters cultured in a disposable 2-L bior eactor proliferated and made up to 60% of the vessel volume after 4- 5 weeks. The aggregates which reached 0.5- 1.5 cm in diameter tended to sink to the bottom of the bioreactor as the biomass increased and could not be resuspended unless the rate of aeration was increased. An increase from 0.5 to 1.5 vvm (volume air/ volume medium per minute) provided the required aeration for recirculation of the cluster biomass (Ziv et al. 1998). Potato and banana bud clusters proliferated in bubble and disposable plastic bioreactors at a faster rate than on agar-solidified medium (Levin et al. 1997b)
Growth regulators, when given in a specific balanced ratio of promoting and retarding substances, apparently act as morphogenic signals and control the development of the spherical meristematic or bud clusters. The spherical meristematic or bud clusters were described by various terms: pro-embryogenic mass (PEM) in carrot (Halperin 1967), protocorms in orchids (Morel 1974), nodules in poplar (McCown et al. 1988) and radiata-pine (Aitken-Christie et al. 1988; Carmi et al. 1997), nubbins in daylilly (Krikorian and Kann 1981), and meristematic or bud clusters in gladiolus, Brodiaea, Nerine sarniensis, fern, Philoden-dron, potato, banana, and Ornithogalum dubium (Ziv 1989; Ziv and Hadar 1991; Ziv and Ariel 1991; Ziv et al. 1994; Ilan et al. 1995; Ziv and Shemesh 1996; Ziv et al. 1998; Ziv and Lilien-Kipnis 1997). It is suggested that the term cluster should be used, as it is more appropriate for describing these proliferative, usually rounded structures, which develop from buds or other proliferative tissue in liquid culture, through either the organogenic or somatic embryogenesis pathways.
D. Anomalous Plant Morphogenesis
Scaling-up in bioreactors requires the use of liquid instead of agar-gelled media during the proliferation and biomass production stages. The culture of plants in liquid medium is known to cause anomalous morphogenesis, resulting in plant hyperhydricity (Debergh et al. 1992). The plants that develop in liquid media are fragile, have a glassy appearance, with succulent leaves or shoots and a poor root system (Paque and Boxus 1987; Werker and Leshem 1987; Ziv 1991a, c, 1995a). The leaves are the organs affected most severely in liquid cultures. They develop an unorganized mesophyll tissue that is made up mainly of spongy parenchyma tissue with large intercellular spaces (Werker and Leshem 1987; Gaspar et al. 1987), a deformed vascular tissue, and an abnormal epidermis. The epidermal tissue in hyperhydric leaves lacks a well-developed cuticle and possesses malfunctioning guard cells which can-not respond to closure signals (Sutter 1985; Ziv et al. 1987; Ziv 1991a; Ziv 1995a). Hyperhydricity affects plant survival after transplanting and causes loss of the in vitro developed leaves, or even whole plants, which often wilt and die. The two major processes carried out by the leaves, photosynthesis and transpiration, are not fully functional in hyperhydric leaves and thus cause the poor performance of the transplanted plants ex vitro (Preece and Sutter 1991; Ziv 1991c, 1995a; Ziv and Ariel 1994).
In many plants propagated in vitro, the anomalous morphology and anatomical aberration often observed in plants result from deviation from the normal course of morphogenic events that are manifested in plants in vivo (Ammirato 1985; Ziv 1999). Some of the deformation, such as hyperhydric, malformed leaves and shoots, abnormal embryos, recurrent embryogenesis, and several other disorders (Ziv 1999), are apparently the result of interruption or faulty timing of the signals involved in the normal sequence of organizational events known to exist in vivo. These are problems that are manifested more severely in liquid medium and await further studies in order to understand and to help control plant morphogenic events in bioreactor cultures.
III. PLANT CELL AND TISSUE GROWTH IN BIOREACTORS
The advantages provided by aerated liquid cultures in bioreactors include better contact between the plant biomass and the medium, no restrictions of gas exchange, the control of the composition of both the medium and the gaseous atmosphere, and the ability to manipulate the plant biomass in relation to the medium volume (Cazzulino et al. 1991; Heyerdahl et al. 1995; Leathers et al. 1995). The need for efficient circulation and mixing of the plant biomass, especially for cluster and embryogenic tissue, is essential to prevent sedimentation and allow optimal growth (Scragg 1992; Doran 1993).
Various types of bioreactors with mechanical or gas-sparged mixing were used for plant cell cultures, to provide stirring, circulation, and aeration (Margaritis and Wallace 1984; Takayama 1991; Scragg 1992; Doran 1993; Takayama and Akita 1994). Mechanically stirred bioreactors depend on impellers, including a helical ribbon impeller (Archambault et al. 1994), magnetic stirrers, or vibrating perforated plates (Styer 1985; Preil 1991; Cazzulino et al. 1991). Aeration, mixing, and circulation in bubble column or airlift bioreactors is provided by air entering the vessel from a side or bottom opening through a sparger; as the air bubbles rise, they lift the plant biomass and provide the oxygen gas required (Merchuk 1990). Mechanically stirred bioreactors are used for effective mixing, aeration, dispersion of air bubbles, and prevention of large cell aggregates formation in cell suspension cultures (Denchev et al. 1992; Scragg 1992; Doran 1993).
In recent years, large-scale cultivation of plant cells, embryos, or organs has made use of airlift or bubble column bioreactors and, to a lesser extent, of mechanically stirred tank bioreactors, due to the former lower shearing force properties. Plant cells are relatively large in size, have sensitive cell walls which are highly susceptible to shearing forces, and are easily damaged. Furthermore, since plant cells, unlike microbial cultures, do not require high O2 , the use of bubble column (Fig. 1.3A) or airlift bioreactors was found to be adequate and advantageous for plant tissue cultured in liquid media (Margaritis and Wallace 1984; Smart and Fowler 1984; Merchuk 1990). In addition, it was shown that mixing by gas sparging in bubble column or airlift bioreactors lacking impellers or blades is far less damaging for clusters than mechanical stirring, since they were shown to have a lower shearing stress (Ziv and Hadar 1991; Ilan et al. 1995; Ziv and Shemesh 1996).
The main advantage of airlift bioreactors is their relatively simple construction, the lack of regions of high shear, reasonably high mass and heat transfer, and relatively high yields at low input rates (Kawase 1989; Denchev et al. 1992). A bubble free oxygen supply bioreactor with silicone tubing was found suitable for embryogenic cell suspensions and provided foam-free cultures (Luttman et al. 1994). For hairy root culture, an acoustic mist bioreactor was found to increase root biomass significantly (Chatterjee et al. 1997).
For efficient large-scale cultures of both somatic embryos and organogenic plant tissue, the bioreactor configuration and volume must be determined according to the mixing and aeration requirements of the specific plant or tissue propagated, as well as for minimizing the intensity of the shear stress (Doran 1993; Hvoslef-Eide and Munster 1998).
One major problem encountered in large-scale liquid cultures is contamination. Fungi, bacteria, yeast, and insects are a source of serious contaminations, causing heavy losses of plant material in commercial laboratories. In scaled-up liquid cultures, the losses are even greater, as the source of contamination due to manipulation of the bioreactor apparatus is dependent on the various stages of preparing and maintaining the equipment. In several laboratories, the operation area is kept sterile by a positive pressure airflow, which decreases contamination risks. In addition, constant screening of the plant tissue for contaminants and continuous indexing is another safeguard (Cassells 1991).
IV. PHYSICAL AND CHEMICAL FACTORS IN LIQUID CULTURES
In order to control plant morphogenesis and biomass growth in bioreactors, various culture conditions must be manipulated, i. e., the gaseous atmosphere, oxygen supply and CO2 exchange, pH, minerals, carbohydrates, growth regulators, and the liquid medium rheology and cell density (Heyerdahl et al. 1995).
A. The Gaseous Atmosphere
The atmosphere of the culture vessel is made up mainly of nitrogen (78%), oxygen (21%), and carbon dioxide (0.036%). The culture vessel gas composition is influenced by the volume of the vessel and the extent of ventilation. Plants evolve CO2 and consume O2 during respiration, while during photosynthesis CO2 is used and O2 is produced. During the dark period, CO2 levels were found to increase in cultures, and if photoautotrophic conditions prevail, its level decreased in the light (Fujiwara et al. 1987). Ethylene, ethanol, acetaldehyde, and other hydrocarbons are additional components of the gaseous atmosphere in vitro.
Most of the effects of CO2 , O2 , and C2 H4 on plant growth in vitro were reported for agar-gelled or cell suspension cultures (Buddendorf-Joosten and Woltering 1994). In bioreactors, the control of the gaseous phase depends on the gas flow and can be easily manipulated to provide the required levels of O2 , CO2 , and C2 H4 . In airlift or bubble column bioreactors, the air supplied is used for both mixing and aeration (Scragg 1992; Doran 1993). The importance of aeration and the gaseous phase were shown in potato cultured in airlift bioreactors. Induction of tubers was inhibited under continuously submerged conditions. Microtubers developed only after the shoots elongated and reached the gaseous phase at the top of the bioreactor. Enrichment with O2 and manipulation of the hormonal and osmotic conditions had no effect and could not explain the phenomenon. A two-phase culture substituting the growth medium with a tuber induction with 9% sucrose in the medium enhanced tuber formation from shoots that developed above the medium and were exposed to the gaseous phase (Akita and Takayama 1994). These results emphasize the importance of the gaseous phase in the bioreactors for specific developmental phases.
1. Oxygen Level. Oxygen levels in liquid cultures depend on the presence of O2 in the gas phase above and in the air bubbles inside the medium, as well as in the dissolved O2 in the medium. Air is released through a sparger located at the base of the bioreactor. The available oxygen for plant cells in liquid cultures determined by oxygen transfer coefficient (kL a values) is the part that dissolves in water. Its depletion as a function of the metabolic activity of the growing cell biomass can affect the culture yield. Plant cells have a lower metabolic rate than microbial cells and a slow doubling time and therefore require a lower O2 supply. In general, high aeration rates appear to reduce the biomass growth (Cazzulino et al. 1991).
The requirements for O2 may vary from one species to another, and must be supplied continuously to provide adequate aeration, since it affects metabolic activity and energy supply as well as anaerobic conditions. The level of O2 in liquid cultures in bioreactors can be regulated by agitation or stirring and through aeration, gas flow, and air bubble size. The use of a porous irrigation tube as a sparger generated fine bubbles, high kL a values, low mechanical stress and provided a high growth rate (Takayama and Akita 1998). Growth of poinsettia cell suspension in bioreactors was inhibited when the level of O2 dropped below 10%. When the level of O2 was elevated to 80%, cell number increased to 4.9 × 10 5 /ml as compared to 3.1 × 10 5 /ml at 40% O2 (Preil et al. 1988). Somatic embryo development in alfalfa and poinsettia suspension cultures was enhanced at 78% and 60% O2 levels, respectively (Stuart et al. 1987; Preil 1991). High aeration rates were found to inhibit cell growth in cell suspensions cultured in airlift bioreactors. This result was explained to be due to an effect of "stripping" of the volatiles produced by the plant cells, which are apparently necessary for cell growth (Smart and Fowler 1984). Increasing O2 levels from 21% to 80%, in bioreactor cultures of Boston fern clusters, enhanced growth values (final FW - initial FW/ initial FW) from 0.61 to 0.92. Reducing O2 levels to 10% (v/ v) affected cell differentiation in bioreactor cultures of carrot embryogenic tissue. Under these conditions embryo production was severely inhibited (Jay et al. 1992). The best producti on of somatic embryos from embryogenic cultures of Eschcholtzia californica and carrot in a helical ribbon impeller bioreactor was achieved under 20% oxygen concentration. When O2 was low (5- 10%), it inhibited biomass and somatic embryo production, while high O2 (60%) favored undifferentiated biomass production (Archambault et al. 1994, 1995). Additional studies are required to provide information on optimal dissolved O2 requirements in large-scale liquid cultures.
2. CO2 Effects. The effects of CO2 were reported mainly for plants in agar-gelled cultures or for cell suspension cultures used for secondary metabolite production (Scragg 1992; Buddendorf-Joosten and Woltering 1994). The reports on the effects of CO2 enrichment in sugar-free agar-gelled cultures suggest beneficial promotion of plant growth during plant acclimatization and transplanting ex vivo (Kozai et al. 1992). The contribution of CO2 supply during the proliferation and multiplication stage in media supplied with sucrose in bioreactors is debatable. It is implicit that if photoautotrophic conditions do not prevail, CO2 enrichment beyond the 0.36% in the air supply is unnecessary. There are reports that high aeration rates rather than excessive oxygen inhibit growth and that reduced growth could be due to depletion of CO2 or to the removal of various culture volatiles including CO2 (Hegarty et al. 1986; Kim et al. 1991). The requirement for CO2 was not related to photosynthesis but to some other metabolic pathways involved in amino acid biosynthesis. In poinsettia, the growth of embryos in Erlenmeyer flasks was considerably higher than in bioreactors and it was suggested that the growth difference was related to the differences in the gaseous atmosphere in the headspace (Preil 1991). CO2 enrichment in an illuminated bioreactor culture of Brodiaea clusters did not affect biomass growth. An increase from 0.3% to 1% gave a similar growth value under the two CO2 levels and supply of 135 µ mols m -2 s -1 photosynthetic photon-flux (Ilan et al. 1995). In Cyclamen persicum Mill, high CO2 levels correlated with increased production of pro-embryogenic masses (Hvoslef-Eide and Munster 1998).
3. Ethylene. Ethylene level in the headspace in liquid cultures in flasks differs from that in continuously aerated bioreactor cultures. Most of the reports on ethylene relate to agar-gelled or cell suspension cultures (Buddendorf-Joosten and Woltering 1994). In various plant species cultured in bioreactors, the rate of aeration will affect the level of ethylene. High rates of aeration, which are often required at high biomass densities, can cause "stripping" of volatiles that are apparently important for some plants grown in culture. In clusters of Brodiaea cultured in liquid medium, ethylene had no effect on growth, although its level was reduced in the presence of silver thiosulfate (an inhibitor of ethylene action). The level of ethylene was reduced from 0.38 to 0.12 ml/ L in highly aerated bioreactor cultures without affecting biomass growth, which was similar under both levels (Ilan et al. 1995).
B. Mineral Nutrients Consumption
Revised MS medium (Murashige and Skoog 1962) or media with various partial modifications in the inorganic and organic constituents of MS are used for most plant species in agar-gelled or liquid cultures in vitro. The availability of mineral nutrients depends on the type of culture, whether agar-gelled or liquid, the type and size of the plant biomass, and the physical properties of the culture. Factors such as pH, temperature, light, aeration, the concentration of minerals, the medium volume, and the viscosity of the medium will determine the rate of absorption of the various nutritional constituents (Williams 1992; Debergh et al. 1994). Plant cells growing in liquid cultures are better exposed to the medium components and the uptake and consumption are faster.
Agar-gelled and liquid shaken cultures often dehydrate by water evaporation to the headspace and out of the vessel and such a water loss concentrates the medium. The content of a certain component in the medium is a product of its concentration and the medium volume. The effects of the medium volume and the initial strength on potato were discussed by Kozai et al. (1995). In bioreactors, in which either humidified air or condensers are used to prevent dehydration, the level of the nutrients in the medium is affected mainly by the absorption rate and by cell lysis (Archambault et al. 1994). Differentiation and proliferation of micropropagated fern, gladiolus, and Nerine nodular clusters in bioreactors was better on half-strength than on full-strength MS minerals (M. Ziv, unpubl.). This was also true for Lilium bulblets differentiating on bulb scales that were cultured in bioreactors (Takayama 1991). A drop in pH to 4.5 and lower values and the subsequent increase to pH 5.5 was attributed to the initial utilization of ammonium and to the uptake at a later stage of nitrate. In several species the depletion of NH + 4 is the first limiting factor of biomass growth and somatic embryos development. Increasing the concentration of NH + 4 to 15 mM resulted in maximum somatic embryo production in carrot cultures in a helical ribbon impeller bioreactor (Archambault et al. 1995). A thorough and detailed study of nutrient uptake in Eschcholtzia californica embryonic cultures was achieved in a helical ribbon impeller bioreactor. Following a lag phase of about 140 h, first NH + 4 and then NO - 3 , PO - 4 , K + and SO - 4 uptake was observed to coincide with biomass increase and somatic embryo production. After 300 h, PO - 4 was exhausted, while after 600 h NH + 4 was depleted and Mg ++ and Ca ++ composition continued to deplete after about 800 h. The K ++ ions were not depleted at that stage (Archambault et al. 1994). The composition of minerals, monitored during the culture of Begonia and Brodiaea in liquid media changed, with phosphate, ammonia, nitrate, and potassium depleting faster and prior to the depletion of Ca ++ and Mg ++ (Tormala et al. 1987; Ilan et al. 1995). In somatic embryos of spruce cultured in bioreactors, 80% of the ammonium was consumed by the growing biomass (Tautorus et al. 1994). In general, biomass growth is limited by the availability of phosphate, nitrogen, and carbohydrates and to a lesser extent by the availability of calcium, magnesium, and other ions.
C. Carbohydrate Supply and Utilization
Cultured plants require a constant supply of carbohydrates as their source of energy. Sucrose and to a lesser extent glucose, fructose, or sorbitol are the most commonly used carbohydrates in vitro. In general, sucrose is removed rather rapidly from the medium and after 10- 15 days the sucrose can be completely depleted or reduced to 5- 10 g/ L from an initial level of 30 g/ L in both agar-gelled and liquid cultures. At the same time, glucose and fructose that appear in the medium due to sucrose hydrolysis increase in the presence of invertase in the culture medium, and can reach levels of 5- 10 g/ L. Cell suspensions of Catharanthus roseus cultured in a column airlift bioreactor showed a lag phase of 5 days, during which there was a total hydrolysis of the sucrose to glucose and fructose (Smart and Fowler 1984). In suspension cultures of alfalfa, sucrose also was hydrolyzed during the first 5 days. Most of the sugars uptake occurs after day 5 and glucose is taken up preferentially over fructose (McDonald and Jackman 1989). A higher yield of alfalfa embryos was obtained when 30 g/ L maltose combined with NH + 4 was used instead of 30 g/ L sucrose that was used in combination with various nitrogen sources (Stuart et al. 1987). In embryonic suspension cultures of celery, the addition of mannitol reduced cell lysis and enhanced somatic embryogenesis. When 40 g/ L mannitol was added, a higher number of embryos was produced and the frequency of singulated normal embryos was increased (Nadel et al. 1989). In embryogenic cultures of Eschcholtzia californica grown in a helical ribbon impeller bioreactor, sucrose uptake started after 100 h and the sucrose was depleted afte r 600- 800 h in culture (Archambault et al. 1994).
In Picea species cultured in shaken flasks in a medium with 30 mM glucose, the carbohydrates depleted after 10- 14 days (Lulsdorf et al. 1993). In mechanically stirred bioreactors, a level of 60 mM sucrose resulted in the highest cell biomass and somatic embryo number. The effects of various carbohydrates on the growth of spruce somatic embryos revealed that the response was species dependent (Tautorus et al. 1994).
The biomass of fern meristematic clusters in a bubble column bioreactor was increased with the increase in sucrose concentrations from 7.5 g/ L to 30 g/ L, while higher levels caused a decrease in cluster growth. Elevated sucrose concentrations caused a decrease also in the chlorophyll content of the clusters and leaves (Ziv and Hadar 1991).
Increasing sucrose concentrations from 30 g/ L to 60 g/ L in bioreactor cultured gladiolus clusters decreased the biomass FW by more than 50%. When 30 g/ L or 60 g/ L were combined with paclobutrazol, a further decrease in FW was observed. On the other hand, 60 g/ L sucrose induced a higher DW increment, which could be attributed to an osmotic effect and the subsequent water status of the clusters. Gladiolus clusters cultured in the presence of growth retardants had a higher level of starch, 845 as compared to 585 mg/ g DW in the control (Ziv 1992b). A higher number of bulblets were produced in bioreactor-cultured bulb scales of Lilium at 30 g/ L than at higher sucrose levels. However, larger size microbulbs were produced at 90 g/ L than at 30 g/ L sucrose in the medium (Takayama 1991).
D. pH Effects
The initial pH in most plant cell cultures ranges between 5.5- 5.9. Since most media are not buffered, changes during autoclaving and during the biomass growth in culture occurs. A rapid drop in pH to 4.0- 4.5 took place within 24- 48 h in cell suspension, organ, and embryogenic cultures (Stuart et al. 1987; Preil 1991; Ziv and Hadar 1991; Lulsdorf et al. 1993; Jay et al. 1994). These changes were related to an initial ammonium uptake and acidification due to cell lysis. However, the pH in-creased after a few days and reached a stable level around pH 5.0- 5.5, which was related to the uptake of nitrates. In spruce species cultured in liquid medium, the pH levels were shown to increase to 6.5- 6.8 after 14 days in culture (Tautorus and Dunstan 1995).
In embryogenic cultures of poinsettia, various initial pH levels stabilized after 20 days in batch bioreactor cultures, between pH 5.3- 5.7. Oxygen levels affected pH changes by 0.2 pH units (Preil 1991). In fern cultured in an airlift bubble column bioreactor, the pH dropped to 4.2 after 24 h and subsequently increased to pH 5.4. Keeping a constant pH by titration with KOH did not affect the growth response (Ziv and Hadar 1991). In alfalfa, the development of somatic embryos was affected by the pH. A higher rate of embryo production was observed at a constant pH of 5.5 than at a non-buffered medium or at lower pH levels (Stuart et al. 1987).
Carrot cell differentiation in a controlled bioreactor was affected by the pH; the highest rate of embryo production was observed at a pH of 4.3. However, embryo development was arrested before the embryos reached the torpedo stage and continued only at pH 5.8. The changes in carrot embryo development were associated with sugar uptake and ammonium depletion and can be attributed to enzyme and metabolic activity at an optimal pH (Jay et al. 1994). It appears that pH requirements are species and developmental stage dependent.
E. Growth Regulator Effects
The use of growth regulators in liquid cultures can be more effective in controlling the proliferation and regeneration potential than in agar-gelled medium due to the direct contact of plant cells and aggregates with the medium. The limited information available on regulators seems to indicate that similar levels of growth regulators were used in both agar-gelled and liquid cultures even though availability appears to be better in liquid medium.
Somatic embryogenesis in many species was first induced in an auxin-containing medium that promoted rapid cell division. Expression of the embryogenic potential was achieved in auxin-free media, or with low levels of auxin in the medium (Halperin 1967; Steward et al. 1970). Further embryo conversion and maturation into plants was promoted by the addition of abscisic acid (ABA), which was found to control abnormal embryo and plant growth (Ammirato and Styer 1985).
Bulblets development from scales in Lilium cultured in bioreactors was higher in the presence of BA than kinetin. However, high BA had an inhibiting effect on further bulblet growth (Takayama et al. 1991). High kinetin also stimulated bulblets differentiation, but inhibited further growth of the bulblets. However, the addition of 0.1 mg/ L naphthalenacetic acid (NAA) was found to further enhance kinetin activity, which was, however, inhibited again in the presence of high (90 g/ L) sucrose levels (Takayama 1991; Takayama et al. 1991).
In embryogenic cultures of Nerine, auxin and cytokinins were used to induce proembryogenic clusters. Embryogenic expression was achieved, however, only after a short exposure to 2-iso-pentenyladenine (2iP) and further subculture in a growth regulator-free medium (Lilien-Kipnis et al. 1992, 1994).
Since one of the major problems in liquid-cultured plants is malformation of shoots, hyperhydricity, the induction of meristematic or bud clusters with arrested leaf growth (McCown et al. 1988; Ziv 1991b; Ziv and Shemesh 1996) was one of the solutions to hyperhydricity. The use of relatively high cytokinin levels and growth retardants which inhibit gibberellin biosynthesis was the most effective method to reduce shoot and leaf growth and promote the formation of meristematic clusters (Ziv 1990a, b). The information on abscisic acid is limited and was found effectual mainly in the later stages of somatic embryo development, promoting normal embryo growth and maturation in carrot and alfalfa embryogenic cultures (Ammirato and Styer 1985; Denchev et al. 1990). Growth of callus and leaf development in bulblets regenerated on bulb scales of Lilium in bioreactor cultures was inhibited by abscisic acid, thus providing singulated propagules for easier handling and storage (Takayama et al. 1991). Ethylene effects appear to be species dependent (see IVA3). Information on the effects of growth regulators in bioreactor cultures is inadequate and further research is needed to optimize culture conditions.
F. Temperature Effects
The control of the temperature in the liquid medium inside the bioreactor can be easily manipulated by a heating element in the vessel or by circulating water in an enveloping jacket outside the vessel. There is, however, limited information on the effects of temperature in bioreactor cultures, which is usually kept constant at 25 C, with small day and night fluctuations. Temperature effects on potato tuber formation in an airlift bioreactor were studied by Akita and Takayama (1994). A higher number and larger-size tubers developed at 25 C than at 17 C; the lower temperature caused a decrease in tuber size. Ziv and Shemesh (1996), working with potato internode explants in liquid cultures, found that tuber formation was best at a 16 h photoperiod and 18/ 15 C day and night temperature. When bulblets of Nerine were cultured in liquid medium, the FW increase was higher at 25 C than at 17 C. However, when subcultured from liquid to an agar-gelled bulb induction medium, rooting was better at 17 C (J. Vishinevetzky, M. Ziv unpublished).
V. CELL AND AGGREGATE DENSITY, FOAMING AND MEDIUM RHEOLOGY IN BIOREACTORS
Growth and proliferation of the biomass in bioreactors depends on air-flow supply for the aeration and mixing, and for the prevention of the plant biomass sedimentation. In many plants cultivated in bioreactors, continuous aeration, mixing, and circulation cause shearing damage, cell wall breakdown, and accumulation of cell debris, which is made up mainly of polysaccharides. Cell debris accumulation results in foaming, adhesion of cells and aggregates to the culture vessel walls, and the development of a "crust" at the upper part of the bioreactor vessel. This layer prevents adequate circulation, causing additional cell debris formation and a demand for higher aeration rates that intensify the clogging problem (Scragg 1992). As the biomass increases and the cultures become viscous, higher rates of aeration are required to allow for oxygen supply and circulation. Medium viscosity and foaming were reduced by the use of half the concentration of MS minerals (Ziv 1992a) and by lowering the level of calcium in the medium (Takayama et al. 1991). In bioreactor cultures in which the plant biomass used was of an aggregate or cluster nature, the foaming and viscosity were not as severe as that observed in cell suspension cultures (Ziv 1991b). Surprisingly, in cluster cultures in bioreactors, even when the mixing was poor and the clusters sedimented, biomass growth was observed, suggesting that perhaps the continuous circulation accompanied by aeration was secondary in importance to oxygen supply (Ziv 1995b).
Shearing stress and cell wall damage were greatly reduced in a Vibra-mix bubble-free oxygenated bioreactor, in which silicone tubing was used for air supply (Preil 1991; Luttman et al. 1994). Embryogenic cells cultured in a helical ribbon impeller bioreactor produced poor-quality embryos when the mixing speed was increased from 60 to 100 rpm. This was attributed to shearing stress and cell damage (Archambault et al. 1994).
The introduction of polyethylene glycol 6000 changed the rheology of the medium in alfalfa liquid cultures and improved somatic embryo development beyond the globular stage while it was arrested in a less viscous medium (Denchev et al. 1990).
Disposable plastic presterilized bioreactors (Osmotek 'LifeReactor ' ª, Fig. 1.3B), with a volume capacity of 2 and 5 L, were used for the micro-propagation of various plant species through bud or meristematic clusters. The plastic bioreactors were found to provide good circulation with reduced shearing damage and foaming (Ziv et al. 1998). The problems of cell damage, foaming, and culture viscosity can be better controlled by developing bioreactors with an optimal shape suitable for micropropagation through meristematic or bud clusters. Clusters that are compact and consist mainly of small meristematic cells are apparently less shear sensitive than highly vacuolated cells in parenchymatous cell suspension cultures.
VI. SUMMARY AND CONCLUSIONS
The application of liquid cultures for micropropagation in bioreactors using the organogenic or the embryogenic pathway is becoming a more efficient alternative system for scale-up and automation in vitro (Aitken-Christie et al. 1995). The successful exploitation of bioreactors as a commercial micropropagation system will depend on careful studies of plant morphogenesis in liquid media and the understanding of the control mechanisms of organ and embryo development from meristematic or bud clusters. The chemical and physical environment, in relation to biomass growth and controlled regeneration, should be further investigated. The level of carbohydrates and specifically the levels and ratios of growth-promoting and -retarding regulators will need to be further studied in more detail.
A major aspect which will have to be addressed is the problem of contamination in large-scale liquid cultures, which can cause severe losses (Leifert and Waites 1992). Attempts to control contamination in liquid cultures of foliage plants were reported by Levin et al. (1997a), who used continuous filtration systems to control bacterial growth in the medium.
Harvest and quality assessment of bioreactor propagules is a major aspect in automation and was addressed by computerized image analysis (machine vision) to classify organogenic and embryogenic plant material (Harell et al. 1994; Kurata 1995). These are factors of major importance to the success of scaled-up automated in vitro propagation.
In addition, the understanding of the effects of aeration, mixing, consumption, and depletion of the various components present in the medium will provide information for establishing semi-continuous or continuous culture systems and thus provide optimal conditions for biomass growth, differentiation, and eventual production of quality plants.
Implementation of scale-up and mechanization are mandatory for the expansion of commercial micropropagation. New technologies in robotics and machine vision systems for cutting, sorting, and dispensing have been established (Aitkin-Christie et al. 1995). However, these technologies have actually increased the cost of plant micropropagation. Bioreactor cultures are being established in several commercial laboratories for ferns, spathiphylum, philodendron, banana, potato, lilies, poinsettia, sugar-cane, and some forest tree species such as eucalyptus, poplar, and early stages of conifer somatic embryos. The technique was estimated to render a 35% reduction in propagule unit costs when propagated through the organogenic pathway (R. Levin, pers. commun.). Using somatic embryogenesis as the propagation pathway in a semi-automated system (Cervelli and Senaratna 1995) reduced production costs by 24%. In conclusion, simple bioreactors combined with mechanized cutting, sorting, and delivery systems are the solution for low-cost micropropagation.