KNOX2 mutant phenotypes and expression patterns

KNOX2 mutant phenotypes were characterized using null alleles (S5 Fig.). As reported previously [20], single mutants lack conspicuous aberrant phenotypes. Amongst double mutants, knat3 knat5 seedlings are distinguishable from wild type by a longer petiole and narrower lamina of cotyledons, and more deeply serrated leaf margins (Fig. 2A-B). Venation pattern is also affected in knat3 knat5 cotyledons (S6 Fig.). knat3 knat4 plants also have serrated leaves (Fig. 2I) and are sporophytically female sterile with abnormal integument development. While knat3 knat4/+ knat5 plants are also female sterile, knat3/+ knat4 knat5 plants are phenotypically wild type and produce viable seeds, facilitating characterization of segregating triple mutant plants.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. KNOX2 mutant phenotypes and KNAT5 expression patterns.

(A-C) 10-day-old seedlings of wild-type (WT, A), knat3 knat5 (designated as knat35, B), and knat345 (C). (D-E) Wild-type (D) and knat345 (E) plants after bolting; 5 weeks-old plants are shown. (F-G) Representative wild-type (F) and knat345 (G) rosette leaves. (H-L) First cauline (stem) leaves of wild-type (H), knat3 knat4 (designated as knat34, I), knat3 knat4 knat5/+ (designated as knat34, 5/+, J), knat3 knat5 (K), and knat345 (L) plants; progressive loss of the KNOX2 activity results in increasingly serrated leaves. (M-N) Wild-type (M) and knat345 (N) flowers. (O-R) Mature wild-type (O, Q) and knat345 (P, R) ovules and embryo sacs. An arrow marks ectopic formation of tracheary elements. (S-V) _proKNAT5:KNAT5-GUS expression in developing embryos (S-T), vegetative shoot apex (U), leaf (V), and ovules (W). _proKNAT5:KNAT5-GUS is not detected in the shoot apical meristem (marked by an arrowhead) and the youngest leaf primordium (marked with an arrow). An asterisk indicates a stipule. (X) _proKNAT4:GUS expression in vegetative shoot apex. For additional expression data, see S10 Fig. Plants are all in the Columbia (Col) background. Scale bars, 1 cm.

https://doi.org/10.1371/journal.pgen.1004980.g002

Selfed knat3/+ knat4 knat5 plants segregated small, dark-green plants with deeply lobed leaves (Fig. 2C-G and S14F Fig.), a phenotype reminiscent of gain-of-function KNOX1 alleles [12,13]. PCR-based genotyping indicated these plants were triple-mutant homozygotes (knat345). Since only a single mutant allele was available for each gene, we designed an artificial miRNA (amiRNA, [27]) targeting only KNAT4, amiR¹⁵⁹-KNAT4 (S7A Fig.), and generated knat3 knat5 plants constitutively expressing amiR¹⁵⁹-KNAT4 under the control of the Cauliflower Mosaic Virus 35S promoter (_pro35S). _pro35S:amiR¹⁵⁹-KNAT4 knat3 knat5 lines closely resembled the identified knat345 plants (S8C-D Fig.). Another amiRNA, amiR¹⁵⁹-KNAT345–1, was designed to target KNAT3, KNAT4, and KNAT5 (S7B Fig.). _pro35S:amiR¹⁵⁹-KNAT345–1 plants also show a deeply serrated leaf phenotype (S8E and S14H Figs.). We thus conclude that this is the triple mutant phenotype. Consistent with functional redundancy among these genes, dosage-dependent enhancement of the leaf serration phenotype was observed (Fig. 2H-L). Likewise, venation pattern in cotyledons is more severely affected in knat345 plants (S6 Fig.). Floral organs homologous with leaves are also affected. Sepals and petals are narrower and partially dissected in the knat345 mutant, and integument development is defective as seen in knat3 knat4 plants (Fig. 2M-N and S9 Fig.). Ectopic formation of tracheary elements is observed in knat345 embryo sacs (Fig. 2R). Although these genes are expressed in roots [15,16,20], the morphology of primary roots in knat345 plants appeared normal (S10 Fig.).

A _proKNAT5:KNAT5-GUS translational fusion line was generated to monitor expression patterns (Fig. 2S-W and S11 Fig.). In line with the mutant phenotypes, GUS activity was observed in developing leaves but excluded from the shoot apical meristem (SAM) (Fig. 2S-V and S11C Fig.). During early stages of leaf development, GUS activity was not detected in youngest leaf primordia but was observed in older leaf primordia (Fig. 2U). Reduced signal levels were observed in older leaves (S11B Fig.). Prolonged incubation detected GUS signal along cotyledon and leaf veins and in ovules (Fig. 2T and S11H-I Fig.). _proKNAT5:KNAT5-GUS expression is nuclear in trichomes, supporting a role for KNAT5 in transcriptional regulation (S11 Fig.). A transcriptional fusion line, _proKNAT4:GUS, was generated to examine KNAT4 expression patterns. Ten independent T1 plants were examined, all of which exhibited KNAT4 promoter activity in leaves but not in the SAM (Fig. 2X). A similar expression pattern has been described for KNAT3 using either a GUS reporter line or RNA in situ hybridization [15]. Exclusion of KNOX2 expression from the SAM is also supported by cell-type specific expression analyses of the inflorescence SAM (S3 Fig.).

Genetic evidence for KNOX2/BELL heterodimarization

KNOX and BELL heterodimerization plays a pivotal role in regulating their activities as transcription factors [13]. We speculated that the lack of BELL partners may explain why no conspicuous phenotype has been described to date upon ectopic expression of KNOX2 genes [28] (S12 Fig.). The founding BELL gene, BELL1 (BEL1), and closely related paralogs, SAWTOOTH1 (SAW1) and SAW2, represent candidates for KNOX2 partners since loss-of-function phenotypes in ovules and leaf margins resemble those of KNOX2 mutants [29,30,31]. Physical interactions have been previously proposed between these BELL and KNOX2 proteins [22,29,32]. SAW1 and SAW2 are expressed in leaves but not in meristems [29] (S3 and S4 Figs.). Thus, we co-expressed SAW2 and KNAT3 throughout the SAM by trans-activating SAW2 under the control of SHOOT MERISTEMLESS (STM) regulatory sequences (_proSTM>>SAW2; >> denotes the use of transactivation system hereafter) in _pro35S:KNAT3 plants [33]. _pro35S:KNAT3 _proSTM>>SAW2 plants lack an embryonic SAM and resemble loss-of-function stm or stm knat6 mutant plants [34,35,36] (Fig. 3A-C, 3E). Combined expression of KNAT5 and SAW2 in the _proSTM region resulted in a similar phenotype (Fig. 3D), confirming that the presence of both SAW2 and KNOX2 proteins simultaneously accounts for the phenotype. Collectively, these data indicate that concurrent expression, and by proxy, heterodimerization with BELL proteins, is important for KNOX2 function and that KNOX2 activity may thus be constrained by limited access to corresponding BELL partners.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Genetic interactions between BELL and KNOX2 genes.

(A-E) 10 days-old seedlings of _pro35S:KNAT3 (A), _proSTM>>SAW2 (B), _pro35S:KNAT3 _proSTM>>SAW2 (C), _proSTM>>KNAT5 _proSTM>>SAW2 (D), and stm-11 (E), showing resemblance between stm (E) and plants expressing both KNOX2 and SAW2 proteins in the meristem (C, D). Defective meristems are marked with arrows. In (B-D), expression of SAW2 or KNAT5 alone, or both genes together is transactivated by the STM regulatory sequence. (F) Introducing the _pro35S:amiR¹⁵⁹-KNAT345–1 construct did not alter the seedling lethal phenotype of stm-11 plants. Shown is a 25 days-old plant. (G-J) Gynoecia from wild-type (G, emasculated and unpollinated), bel1–154 (H), knat345 (I), and bel1–154 blh1–114 (J) flowers. (I-J) Unfertilized gynoecia turn yellow. See S9 Fig. for more data. This phenotype was not observed in emasculated wild-type (G), bel1 (H), and blh1–114 plants. Plants in (B, D, E, F) are in the Landsberg erecta (Ler) background. A plant in (C) is in the Col/Ler mixed background. All other plants are in the Col background. Scale bars, 0.5 mm.

https://doi.org/10.1371/journal.pgen.1004980.g003

A mutation in KNAT3 suppresses the gain-of-function phenotype caused by ectopic expression of another BELL gene, BLH1, suggesting that BLH1 is likely a functional partner for KNOX2 proteins [37]. This prompted us to examine genetic interactions between BLH1 and BEL1-related BELL genes, and we found that bel1 blh1 double mutants show color changes in unfertilized gynoecia as seen in knat3 knat4 and knat345 plants (Fig. 3G-J and S9 Fig.). Thus, BEL1 and BLH1 play a redundant role in gynoecium development and perhaps act in association with KNOX2 genes. More comprehensive genetic analyses as well as expression analyses are required to assign specific roles to functionally redundant BELL genes.

KNOX2 exhibits selectivity for BELL

To further dissect BELL-KNOX interactions, plants expressing BELL and/or KNOX genes in the _proSTM region were characterized. Among BELL proteins, PENNYWISE (PNY) and POUND-FOOLISH (PNF) are expressed in the SAM and act in conjunction with KNOX1 proteins to promote SAM activity [12,13]. As expected, plants expressing KNOX1 genes (STM or KNAT2) or PNY in the STM domain appeared wild type (S13B-D Fig.). In contrast, _proSTM>>SAW2 plants displayed abnormal floral morphologies, such as fused sepals, reduced petals, and misshapen fruits (S13E, H Fig.), phenotypes often associated with reduced KNOX1 activity, e.g. weak stm mutants [38]. Flower development was not impacted in _proSTM>>KNAT5 plants, but fused sepals are also observed in strong _pro35S:KNAT3 lines (S12F Fig.). Concomitant expression of KNOX2 with PNY or KNOX1 with SAW2 did not enhance the KNOX2 or SAW2 overexpression phenotypes. We therefore conclude KNOX2 shows selectivity for BELL proteins in vivo.

KNOX2 mutant phenotype is independent of KNOX1 activity

Loss-of-function and gain-of-function KNOX2 phenotypes are reminiscent of gain-of-function and loss-of-function KNOX1 phenotypes, respectively [12,13]. To characterize the relationship between the two gene classes, loss-of-function alleles for KNOX1 and KNOX2 were combined. Plants constitutively expressing an amiRNA targeting KNAT3, KNAT4, and KNAT5, _pro35S:amiR¹⁵⁹-KNAT345–1, in KNOX1 loss-of-function (stm or bp knat2 knat6) backgrounds were examined. Neither the meristem failure of stm mutants nor the KNOX2 loss-of-function mutant leaf phenotype was suppressed in these plants (Fig. 3F and Fig. 4A-B). Similarly, neither knat2 knat3 knat5 knat6 nor bp knat345 showed significant suppression of the KNOX2 loss-of-function mutant leaf phenotype and the bp inflorescence phenotype (Fig. 4F-J). Thus, loss-of-function phenotypes of KNOX1 and KNOX2 mutants are not due to ectopic activation of KNOX2 and KNOX1, respectively. Furthermore, BP, STM, and KNAT2 expression was not altered in knat3 knat5 plants (Fig. 4K-P), arguing against mutual repression between KNOX1 and KNOX2 genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Genetic interactions between KNOX1 and KNOX2 genes.

(A-B) Morphology of _pro35S:amiR¹⁵⁹-KNAT345–1 plants in wild-type (A) and bp-9 knat2–5 knat6–1 mutant (B) backgrounds. 6 weeks-old plants are shown. (C-E) F1 plants expressing both _proBLS:STM and _pro35S:amiR¹⁵⁹-KNAT345–1 constructs (D) show a stronger serration phenotype than either parental line (C, E). All plants shown are hemizygous for the transgene(s) and 2 weeks old. (F-G) Morphology of knat3 knat5 (F) and knat2–5 knat3 knat5 knat6–1 (G) plants grown for 5 weeks. (H) A representative bp knat345 rosette leaf with deeply lobed margins, as seen in knat345. (I-J) The bp inflorescence phenotype (I) is observed in bp knat345 infloresences (J), indicating additive effects of mutations in these genes. (H-J) Plants were grown for 2 months. (K-P) KNOX1 reporter expression in wild-type (K-M) and knat3 knat5 (N-P) plants. Shown are _proBP:GUS expression in 17 days-old plants (K, N), _proKNAT2:GUS expression in 20 days-old plants (L, O), and _proSTM:GUS expression in 21 days-old plants (M, P). No ectopic expression of _proBP:GUS was detected during stages of leaf development when lobes are forming in knat3 knat5 plants (K, N), but ectopic expression was observed at leaf serration tips after their development. _proKNAT2:GUS and _proSTM:GUS expression patterns in knat3 knat5 plants are similar to those in wild-type plants (L-M, O-P). Occasionally, longer incubation detected _proSTM:GUS activity in the sinus of wild-type and mutant leaves (M). _proKNAT2:GUS and _proSTM:GUS were analyzed in the mixed genetic background (refer to S1 Table for details), and other plants are in the Col background. Scale bars in I, J, 3 mm, K, L, N, 100 μm and in M, O, P, 200 μm.

https://doi.org/10.1371/journal.pgen.1004980.g004

Deeply lobed leaves, a phenotype characteristic of gain-of-function KNOX1 alleles, occur in Arabidopsis plants where STM is driven by the leaf specific promoter, _proBLS, _proBLS:STM ([39]; S14B, C Fig.). These were crossed with loss-of-function KNOX2 plants (_pro35S:amiR¹⁵⁹-KNAT345–1) to generate plants with ectopic KNOX1 and reduced KNOX2 activities in the leaves. Compared to the parental lines, F1 plants harboring both transgenes displayed more extreme leaf margin elaboration (Fig. 4C-E). The additive effects, rather than epistatic interactions, suggest it is unlikely that the two subclasses negatively regulate one another.

Antagonistic relationship between KNOX1 and KNOX2

An attractive hypothesis for the antagonism between KNOX1 and KNOX2 is that they regulate shared downstream events in an opposite manner. The complex leaf of gain-of-function KNOX1 alelles is suppressed by reduction in CUP SHAPED COTYLEDON (CUC) transcription factor activity [40] (S14C Fig.). Two CUC genes are targeted by the miR164 family of miRNAs, and expression of miR164b in young leaves using regulatory sequences of the FILAMENTOUS FLOWER (FIL) gene (designated as _proFIL), _proFIL:miR164b, flattens the leaf margin in wild-type plants (S14D-E Fig.). Thus, the miRNA-mediated CUC regulation plays a key role in leaf margin elaboration [41]. Introduction of _proFIL:miR164b also suppressed the leaf dissection phenotype in a knat345 mutant background (S14F-G Fig.). Among miR164 targets, CUC2 plays a major role in leaf serration development [41]. We find leaf serration is largely suppressed in the cuc2 knat345 and _pro35S:amiR¹⁵⁹-KNAT345–1 cuc2 backgrounds (Fig. 5A-C and S14H-I Fig.). In addition, constitutive expression of KNOX2 (_pro35S:KNAT3) can partially suppress the _proBLS:STM leaf phenotype (Fig. 5D-F). Thus, a common developmental program mediates both gain-of-function KNOX1 and loss-of-function KNOX2 leaf phenotypes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. KNOX1 and KNOX2 converge on CUC activity.

(A) One-month old cuc2 knat345 quadruple mutant. A mutation in the CUC2 gene largely suppresses the leaf serration phenotype of knat345 plants. Compare with a knat345 plant in Fig. 2E. See S13 Fig. for additional data. (B-C) Representative knat345 (B) and cuc2 knat345 (C) rosette leaves, demonstrating marginal leaf lobing is suppressed by a cuc2 mutation. Shown are the 10th leaves from 2 month-old plants of each genotype. Note that other mutant phenotypes, including leaf size and female sterility, are not suppressed by the cuc2 mutation. (D-F) _pro35S:KNAT3 partially suppresses the leaf lobing phenotype of _proBLS:STM plants. 12 days-old plants are shown. (G-H) Second leaves of wild-type (G) and _pro35S:amiR¹⁵⁹-KNAT345–2 (H) Cardamine hirsuta plants grown for four weeks. (G) In wild-type plants, the first and second leaves always consist of a single, undivided, lamina. (H) Reducing KNOX2 activity by consitutively expressing an amiRNA that targets Cardamine hirsuta orthologues of KNAT3, KNAT4, and KNAT5 genes (_pro35S:amiR¹⁵⁹-KNAT345–2) results in plants with an extra lateral leaflet (marked with an arrow) on the second leaf. (I-J) Third leaves removed from one month-old wild-type (I) and _pro35S:KNAT3 (J) Cardamine hirsuta plants. (I) In wild-type plants, the third leaf typically consists of three leaflets. (J) Introduction of gain-of-function KNOX2 alleles (constitutive expression of the KNAT3 gene from Arabidopsis; _pro35S:KNAT3) results in an undivided third leaf in strong lines. Plants in (A-F) are all in the Col background.

https://doi.org/10.1371/journal.pgen.1004980.g005

As observed in _proBLS:STM plants, elevated levels of KNOX1 activity are often associated with increased leaf complexity (reviewed in [11]). In Cardamine hirsuta, a close relative of Arabidopsis, dissected leaf development requires KNOX1 expression in leaves, and additional KNOX1 expression leads to ectopic leaflet initiation [42]. We investigated the outcome of reduction in the level of KNOX2 activity in this species. In Cardamine, leaf shape exhibits heteroblasty with leaflet number increasing in later produced leaves. Although leaflet number can vary for a particular leaf position, the first and second leaves always consist of a single, undivided, lamina, and the third leaf typically consisting of three leaflets (S15A Fig.). An amiRNA, amiR¹⁵⁹-KNAT345–2, was designed to target three Cardamine genes homologous to Arabidopsis KNAT3, KNAT4, and KNAT5. Constitutive amiR¹⁵⁹-KNAT345–2 expression (_pro35S:amiR¹⁵⁹-KNAT345–2; S7C Fig.) results in plants with an extra lateral leaflet on the second leaf, observed in approximately 15% of individuals (27 of 188 plants derived from 6 independent lines), indicating KNOX2 activity influences complexity of dissected leaves in Cardamine (Fig. 5G-H). Furthermore, gain-of-function KNOX2 alleles (_pro35S:KNAT3) simplify leaf shape, a phenotype particularly obvious in third leaves, which are undivided in strong lines (Fig. 5I-J and S15B Fig.). Thus, reduction or increase in KNOX2 activity leads to increase or decrease in leaf complexity, respectively, in Cardamine (Fig. 5 and S15 Fig.). This observation and the deduced KNOX1/KNOX2 antagonism are in consistent with the results in Arabidopsis.

Evolution of leaf complexity

Expression of KNOX1 genes in leaves is correlated with increased leaf complexity and has been hypothesized to be influential in the evolution of leaf shape [42,43,44]. Given that seed plants leaves evolved from ancestral shoot systems, the ancestral seed plant leaf was likely complex, but fossil evidence and phylognetic analyses suggest that the ancestral angiosperm leaf may have been simple [45]. Regardless of the ancestral state, transitions from simple to more complex and vice versa have occurred repeatedly during angiosperm evolution [43,44,46]. In angiosperms, increase in leaf complexity is associated with increased KNOX1 activity while loss of KNOX1 activity in leaves results in decreasing complexity. While KNOX1 activity has been shown to play a pivotal role, other loci, such as REDUCED COMPLEXITY (RCO) in Cardamine and LEAFY (LFY) orthologues in legumes either contribute directly to modifying leaf shape or influence sensitivity to KNOX1 activity [11,47,48]. The lability of angiosperm leaf architecture may reflect that addition or loss of enhancer modules directing KNOX1 activity in leaves does not affect general plant viability.

The present study demonstrates that KNOX2 activities can also influence leaf shape—leaf dissection increases with decreasing KNOX2 activity (Fig. 2) in a dose dependent manner—raising the possibility of whether changes in KNOX2 activity could also have contributed to the evolution of leaf morphology. Just as KNOX1 gain-of-function alleles result in increases in leaf complexity, novel gain-of-function KNOX2 alleles that alter temporal or spatial expression patterns within developing leaves could contribute to the evolution from complex towards simple leaf morphology, as suggested by our experimental results in Cardamine, via acquisition of leaf specific enhancers. Alleles resulting in loss of KNOX2 activity could also contribute to increases in leaf complexity as suggested by the dose dependent changes to leaf shape in Arabidopsis, however, this may be less likely due to pleiotropic effects of loss-of-function KNOX2 alleles.

Intriguingly, in monilophytes KNOX1 gene expression is broadly similar to that of seed plants, with expression limited to less differentiated tissues including the shoot apical meristem, developing leaves, and procambial tissues [43,49,50]. KNOX2 gene expression has not been studied in detail, but similar to the situation in angiosperms, is reported to be throughout the sporophyte body [50]. In parallel with seed plants, simple leaves have evolved from more complex ancestral leaves within monilophytes [51]. Whether changes in KNOX1 or KNOX2 gene expression may be related to evolution of leaf form in monilophytes is presently unknown.

Nature of the KNOX1/KNOX2 antagonistic relationship

One plausible explanation for the opposing action of KNOX1 and KNOX2 genes is an epistatic relationship between the gene classes. While non-overlapping expression patterns have been observed between KNOX1 and KNOX2 genes, we found no evidence for mutual repression. Alternatively, KNOX1 and KNOX2 proteins may interfere one another’s activity. Such a mode of action was proposed for KNATM in Arabidopsis and PETROSELINUM (PTS)/TKD1 in tomato, both of which are KNOX-related proteins that lack a DNA-binding homeodomain [52,53]. It is suggested that these mini KNOX proteins act as passive repressors and interfere with formation of a functional complex composed of canonical KNOX and BELL proteins. That KNOX2 function depends on the availability of appropriate BELL partners to be active, argues against a similar mechanism for the KNOX1/KNOX2 antagonism. Instead, our data favor a model whereby the antagonistic roles of KNOX1 and KNOX2 are at the level of opposing modes of transcriptional regulation.

Since addition of a repressor domain causes a dominant negative phenotype, KNOX1 proteins can act as activators [39,54]. Conversely a KNOX2 protein, KNAT7, can repress transcription in a transient protoplast system [18,19], and a motif similar to known repression domains is found in the ELK domain of all KNOX2 proteins [55] (S16 Fig.). Comparison of KNOX1 and KNOX2 homeodomains reveals that the third helices, an important determinant of DNA binding specificity, are highly conserved, indicating similar DNA binding properties, at least in vitro (S16 Fig.). Concurrently expressed KNOX1 and KNOX2 proteins could thus conceivably compete with each other at some target genes. Indeed, a putative KNOX2-SAW2 complex can overcome endogenous KNOX1 activities in the meristem, as does a dominant-negative form of KNOX1 (e.g., TKN2-SRDX [39] and en²⁹⁸-STM [54]). However, as KNOX1 proteins have also been reported to act to repress gene expression, the activity of KNOX proteins may be modified by either BELL partners, or third parties, such as OVATE proteins that interact with KNOX/BELL heterodimers and influence both their cellular localization and transcriptional activity [32,37,56]. In a related scenario, KNOX1 and KNOX2 could act on different sets of paralogs of downstream targets. These hypotheses are not mutually exclusive, and depending on the cellular contexts, different modes of action could operate, as is the case for the yeast TALE protein, Matα2, which has different partners in different cell types (reviewed in [23]).

Phylogenetic analyses indicate land plant KNOX1 and KNOX2 genes are derived from a single, ancestral KNOX gene. We hypothesize that subsequent to the KNOX1/KNOX2 gene duplication, accumulating structural differences endowed a new mode of action to at least one paralog. Therefore a possible evolutionary scenario could have an ancestral KNOX protein acting primarily as a transcriptional activator, with the evolution of a transcriptional repressor following gene duplication and neofunctionalization. The evolution of a repressor from an ancestral activator may be a common event, with several instances documented in plant transcription factor families [52,53,57,58,59,60]. Thus, within the context of land plant KNOX genes two types of negative regulators, in which the modes of repressor action are mechanistically different, may have evolved. Mini KNOX proteins act to inhibit KNOX activity by interacting with and sequestering BELL proteins [52,53], as opposed to antagonistic action at the level of downstream gene expression as we propose for KNOX2. The latter provides more flexibility due to the potential to act independently. Accompanying divergence in protein functionality, our data provides additional evidence for nearly complementary expression patterns of KNOX1 and KNOX2 genes in Arabidopsis thaliana. In contrast, in P. patens KNOX1 and KNOX2 genes exhibit both overlapping and distinctive expression patterns [14,26]. Changes in cis-regulatory sequences must have contributed to the establishment of complementary expression patterns during land plant evolution. Flexibility in gene regulatory networks governing meristematic maintenance and differentition engendered by the combination of changes in protein functionality and expression pattern could provide plasticity enabling morphological evolution.

Diversification of KNOX/BELL modules

Heterodimerization between BELL and KNOX proteins is important for translocation of the complex into the nucleus [13]. BELL-KNOX2 heterodimerization may also be critical for providing specificity or increasing affinity of DNA binding (e.g. [61]). Although studies based on the yeast two-hybrid technique suggest physical interactions between BELL and KNOX proteins in a rather nonspecific manner [29,32], our genetic data suggest KNOX2 proteins interact in planta with a subset of BELL proteins, including those of the BEL1/SAW1/SAW2 clade. KNOX1 proteins rely on a distinct set of BELL proteins, e.g. PNY and PNF (reviewed in [12,13]). Due to an obligate heterodimerization requirement, the activity of a KNOX/BELL pair may be limited by the protein with the more restricted expression domain. In Arabidopsis KNOX2 functions appear to be regulated by restricted availability of corresponding BELL partners [29] (Fig. 3).

Similar to KNOX genes, land plant BELL genes evolved from a single gene in the algal ancestor [9]. However, the diversification of paralogs followed a different trajectory in the two families since BELL genes do not fall into discrete functional clades (S17 Fig.). For instance, KNOX1-interacting BELL genes (PNY and PNF) form a sister clade with KNOX2-interacting BELL genes (BEL1 and SAW1/2). Moreover, genetic interactions implicate BLH1, from a phylogenetically distinct clade, as a KNOX2 partner since knat3 alleles suppress the phenotype induced by ectopic BLH1 embryo sac expression [37]. These phylogenetic relationships might be expected if the genome of the land plant common ancestor encoded a single BELL protein that interacted with both KNOX1 and KNOX2 proteins. As the BELL gene family diversified, subfunctionalization would have restricted interactions of BELL paralogs to specific KNOX1 or KNOX2 partners.

KNOX1/KNOX2 gene duplication and land plant evolution

The defining feature of land plants is the formation of an embryo—a multicellular diploid generation. One prominent feature within land plant evolution is the transition from a gametophyte-dominant life cycle to a sporophyte-dominant life cycle [62,63]. This process is regarded as progressive sterilization and elaboration of vegetative organs [62], and in flowering plants, the gametophyte is reduced to a ephemeral structure of only a few cells that is dependent on a sporophyte body that can live up to thousands of years. If the ancestral KNOX-BELL genetic program regulated gene expression in a single celled zygote [25], it follows that during the course of land plant evolution, the KNOX/BELL module has been recruited to control numerous aspects of sporophyte development, with KNOX1/BELL modules promoting meristematic maintenance and continued growth and KNOX2/BELL modules promoting differentiation. In some cases, there is resemblance to a presumed ancestral function, such as in P. patens where KNOX2 genes regulate the gametophyte-to-sporophyte morphological transition [14,26]. In other cases, however, KNOX/BELL modules direct the development of novel structures, such as sporophyte shoot meristems and leaves (Fig. 6), that evolved later in land plant evolution, suggesting the duplication and diversification of the KNOX/BELL genetic module is linked with the evolution of morphological diversity in the land plant sporophyte. Neofunctionalization, exemplified by opposing activities between KNOX1 and KNOX2 genes in Arabidopsis, may underlie the molecular mechanism of key innovations and modification of body plans in the land plant history, through elaboration of transcriptional networks.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Proposed KNOX functions during land plant evolution.

Along the phylogeny of plants, the primary functions for BELL (depicted as B) and KNOX (K) proteins, as well as the gene copy number, are presented for Chlamydomonas (a unicellular Chlorophyte alga), Physcomitrella (a moss), and Arabidopsis (a flowering plant). The ancestral conditions at branches were deduced from our phylogenetic analyses (S1 and S17 Figs.). In each life cycle, a red arrow indicates meiosis, and haploid (grey) and diploid (green) stages are color-coded. In Chlamydomonas the plus gamete expresses a BELL (depicted as B) protein while the minus gamete expresses a KNOX (K) protein; upon gamete fusion the KNOX and BELL proteins heterodimerize and regulate zygotic gene expression. Prior to the origin of land plants, a gene duplication in an ancestral KNOX gene generated two subclasses, KNOX1 (K1) and KNOX2 (K2) genes. In Physcomitrella, KNOX1 activity maintains tissue proliferation during sporophyte (diploid) development while KNOX2 represses the haploid genetic program during the diploid generation. In Arabidopsis, KNOX1 activity promotes meristem maintenance, and our study demonstrates that KNOX2 activity promotes tissue differentiation, perhaps via repression of meristematic functions, in the diploid generation. We propose that (1) the gene duplication producing KNOX1 and KNOX2 paralogs and ensuing neofunctionalization was instrumental in the evolution of a complex multicellular diploid generations in land plants and (2) the diversification of KNOX/BELL modules during land plant evolution facilitated the evolution of ever more complex diploid sporophyte body plans.

https://doi.org/10.1371/journal.pgen.1004980.g006

The role of TALE genes in fungi and Chlamydomonas can be viewed as promotion of cellular specialization in the diploid zygote and progression towards a meiotic state. The life cycle of land plants arose by an interpolation of mitotic divisions between fertilization and meiosis. Thus there is cell proliferation and a delay in meiosis in the diploid generation. KNOX1 genes prevent differentiation and maintain an undifferentiated state of the cells, enabling the cells to proliferate and develop a multicellular body in the sporophyte generation. In organisms with two heteromorphic multicellular generations, such as land plants, the developmental programs for each must be tightly controlled—a role suggested for KNOX2 genes in preventing the haploid gametophyte genetic program to be active during the diploid sporophyte generation in Physcomitrella. We hypothesize the duplication and diversification of the KNOX/BELL genetic module was instrumental in the evolution of a diploid embryo such that multicellular bodies develop in both haploid gametophyte and diploid sporophyte generations known as alternations of generations [25,26]. Alternations of generations have evolved independently in phylogenetically diverse eukaryotic lineages [64,65], prompting the question of whether similar TALE class genetic diversification may be found in these lineages.

Plant material and growth conditions

Arabidopsis thaliana accessions Columbia and Landsberg erecta (Ler) were used as wild type in most experiments. _proKNAT2:GUS was generated in the C24 background and introgressed into Ler. Cardamine hirsuta ‘Oxford strain’ is a kind gift of A. Hay and M. Tsiantis. Plants were grown under long-day (18 hours light) or short-day (10 hours light) conditions at 20°C. knat3 and knat5 alleles are gift from V. Sundaresan and G. Pagnussat. bp-9 knat2–5 knat6–1 seeds are gift from V. Pautot. T-DNA insertion alleles for BELL and KNOX genes were obtained from the Arabidopsis Biological Resource Center (ABRC) or the Nottingham Arabidopsis Stock Center (NASC). Mutant and transgenic lines have been described previously: bp-9 knat2–5 knat6–1 [66]; stm-11 [67]; _proBP:GUS [68]; _proKNAT2:GUS [69]; Op:KNAT2 and Op:STM [39]; and _proSTM:LhG4 [70]. The mutant and transgenic lines used in this study are listed in S1 Table. Homozygous mutant lines were identified by polymerase chain reaction (PCR)-based genotyping. Sequences of genotyping primers are available in S2 Table. The details of the transactivation system was previously described [33].

Genetics

Multiple mutants combining knat3, knat4, and knat5 alleles were generated by crossing, and genotypes were confirmed by PCR-based genotyping. To generate bel1 blh1 double mutant, blh1 plants were crossed with bel1 plants, and the resulting F₂ plants were examined. bel1 plants were identified based on self-sterility, and among them, plants with yellow gynoecia segregated and were confirmed to be bel1 blh1 double mutant plants by PCR-based genotyping. knat2 knat3 knat5 knat6 and bp knat345 plants were identified among F₂ plants originating from a cross between bp knat2 knat6 and knat345 plants, and their genotypes were confirmed by PCR-based genotyping. cuc2 knat345 plants were identified in a F₂ population derived from a cross between cuc2 and knat345 plants. To generate _proFIL:miR164b lines in the knat345 mutant, self-fertile knat3 knat5 plants were transformed with the _proFIL:miR164b construct, and tranformants were selected by resistance to herbicide Basta. Single insertion lines were selected and crossed with knat3/+ knat4 knat5 plants. Among F₁ plants, self-fertile knat3/+ knat4/+ knat5 plants carrying the _proFIL:miR164b transgene were selected, and F₂ seeds were collected; _proFIL:miR164b knat345 plants were identified in the resultant F₂ population. To characterize the effects of the _pro35S:amiR¹⁵⁹-KNAT345–1 transgene in mutant backgrounds, the mutant plants were directly transformed with the _pro35S:amiR¹⁵⁹-KNAT345–1 construct, and transformants were selected by resistance to Basta. As stm null alleles are seedling lethal, heterozygous plants were used for transformation. More than twenty T₁ plants for each background were examined, and phenotypes consistently observed among independent lines were reported.

Semi-quantitative RT-PCR

RNA was extracted, using the RNeasy Plant Mini Kit (Qiagen), from 10-day-old seedlings grown on half-strength MS medium supplemented with 0.5% sucrose. RNA samples were treated with on-column DNaseI (Qiagen) and purified. SMARTScribe reverse transcriptase was used for cDNA synthesis (Clontech), and PCR reactions were performed using Ex Taq (Takara). Oligo sequences used for PCR reactions are described in S3 Table.

Plasmid construction and plant transformation

amiRNAs were designed using the Arabidopsis pre-miR159a backbone (S7 Fig.) and synthesized (GenScript). For construction of the _proKNAT5:KNAT5-GUS reporter construct, the genomic sequence spanning the KNAT5 locus (from the next upstream annotated gene [At4g32030] to the next downstream annotated gene [At4g32050]) was used, and the stop codon was replaced with the GUS coding sequence. For construction of the _proKNAT4:GUS reporter construct, an approximately 6.6-kb region of the sequence directly upstream of the KNAT4 coding sequence was amplified using BAC T5K6 as PCR template and cloned into pCRII-TOPO (Invitrogen). The KNAT4 upstream sequence was subcloned into the pRITA vector, which contains the GUS coding sequence and the terminator sequence from the nopaline synthase gene. For constitutive expression, the amiRNA sequences or the KNAT3 coding sequence were cloned into the ART7 vector, which contains the Cauliflower mosaic virus _pro35S sequence and the terminator sequence from the octopine synthase gene. KNAT5, SAW2, and PNY coding sequences were amplified from Ler cDNA and cloned downstream of an Lac Op array [33] to generate responder cassettes used in the transcription activation system. All constructs were subcloned into pMLBART or pART27 binary vector and were introduced into Agrobacterium tumefaciens strain GV3001 by electroporation. Transgenic lines were generated by Agrobacterium-mediated transformation, and transformants were selected on soil on the basis of resistance to the BASTA or kanamycin. Primers used to clone the various cDNAs and promoters are described in S2 Table.

Histology and microscopy

Scanning electron microscopy was performed according to Alvarez and Smyth [71]. For light microscopy, cleared samples were prepared. Leaf samples were fixed overnight in 9:1 (v:v) ethanol:acetic acid at room temperature. After rehydration in a graded ethanol series, samples were rinsed with water and were cleared with chloral hydrate solution [1:8:2 (v:w:v) glycerol:chloral hydrate:water]. For histochemical analysis of GUS activity, samples were infiltrated with GUS staining solution [0.2% (w/v) Triton X-100, 2 mM potassium ferricyanide, 2 mM potassium ferrocyanide, and 1.9 mM 5-bromo-4-chloro-3-indolyl-β-glucuronide in 50 mM sodium phosphate buffer, pH 7.0] and incubated at 37°C.

Phylogenetic analyses

Publically available KNOX and BELL coding nucleotide sequences representing taxa across land plants were manually aligned as amino acid translations using Se-Al v2.0a11 (http://tree.bio.ed.ac.uk/software/seal/). We excluded ambiguously aligned sequence to produce alignments for subsequent Bayesian analysis. Bayesian phylogenetic analysis was performed using Mr. Bayes 3.2.1 [72,73]. Three separate analyses were performed. The first included Chlorophyte algal and land plant KNOX sequences (S1 Fig.); the second included only land plant KNOX2 sequences (S2 Fig.); and the third included land plant BELL sequences (S17 Fig.). The fixed rate model option JTT + I was used based on analysis of the alignments with ProTest 2.4 [74]. Sequence alignments and command files used to run the Bayesian phylogenetic analyses are provided upon request.