Cell fate determination in plants relies on positional cues. To investigate the position-dependent gene regulation in plants, we focused on shoot epidermal cell specification, which occurs only in the outermost cells. ATML1, which encodes an HD-ZIP class IV transcription factor, is a positive regulator of shoot epidermal cell identity. Despite the presence of a weak ATML1 promoter activity in the inner cells, ATML1 protein was detected mostly in the outermost cells, which suggests that ATML1 accumulation is inhibited in the inner cells. ATML1 nuclear localization was reduced in the epidermis and there was a positive, albeit weak, correlation between the amount of ATML1 in the nuclei and the expression of a direct target of ATML1. Nuclear accumulation of ATML1 was more strongly inhibited in the inner cells than in the outermost cells. Domain deletion analyses revealed that the ZLZ-coding sequence was necessary and partially sufficient for the post-transcriptional repression of ATML1. Our results suggest that post-transcriptional repressions contribute to the restriction of master transcriptional regulator activity in specific cells to enable position-dependent cell differentiation.
For proper pattern formation of multicellular organisms, cell differentiation should be regulated spatiotemporally. In some animals, cell lineage plays an important role in their development because cell fate is already determined during early embryogenesis (Sulston et al., 1983). Unlike animals, plants have a flexibility in their cell-fate decision. Arabidopsis thaliana embryos show a regular well-organized cell division pattern. However, mutant embryos with disrupted division patterns can still acquire the correct body organization (Torres-Ruiz and Jürgens, 1994). Lineage tracing analyses have shown that the apical cell fate was not strictly determined in two-cell-stage embryos (Saulsberry et al., 2002). Moreover, cell ablation experiments in the root meristem have shown that cortex and endodermis initial cells are regenerated from the neighboring pericycle cells (van den Berg et al., 1995). These reports suggest that cell specification in plants largely depends on the cell position, rather than cell lineage. Namely, plant cells can recognize where they are located and gene activities are changed in response to their positions. However, the mechanisms that underlie position-dependent cell-fate decisions are still unclear. To gain knowledge of these mechanisms, we focused on epidermal cell fate, which is acquired and maintained only in the outermost cells (Stewart and Dermen, 1975).
In A. thaliana, ARABIDOPSIS THALIANA MERISTEM L1 LAYER (ATML1) and its closest homologue PROTODERMAL FACTOR2 (PDF2) are necessary for epidermal cell identity. ATML1 and PDF2 encode HD-ZIP class IV transcription factors, which are composed of four domains: a homeodomain (HD), a zipper-loop-zipper (ZLZ) motif, a StAR-related lipid-transfer (START) domain, and a START-associated domain (SAD) (Mukherjee and Bürglin, 2006; Ariel et al., 2007). Weak and strong loss-of-function alleles of ATML1 have been reported; atml1-1;pdf2 can germinate but cannot form an epidermis on their leaves, whereas atml1-3;pdf2 arrests development around the globular stage (Abe et al., 2003; San-Bento et al., 2014; Ogawa et al., 2015). In addition, constitutive expression of ATML1 was shown to induce ectopic epidermal cell differentiation in the inner tissues of cotyledons and leaves (Takada et al., 2013). Taken together, these reports suggest that ATML1 is a positive regulator of epidermal cell identity.
ATML1 mRNA and promoter activity are detected preferentially in the outermost cells during embryogenesis, which suggests that ATML1 transcription is promoted in the outermost cell layer (Lu et al., 1996; Sessions et al., 1999; Takada and Jürgens, 2007). In addition to transcriptional activation, some reports imply that ATML1 activity is also under post-transcriptional repression. For example, ATML1 mRNA was not detected in suspensor cells although the ATML1 promoter was active in these cells (Takada and Jürgens, 2007; Nodine and Bartel, 2010). MicroRNA biogenesis appears to be necessary for ATML1 mRNA degradation in the suspensor cells (Nodine and Bartel, 2010). In another report, GbML1, an ATML1 homologue in cotton, was not clearly localized to the nuclei in the onion epidermis (Zhang et al., 2010). These reports imply that ATML1 mRNA accumulation and the nuclear localization of ATML1 may be decreased by post-transcriptional mechanisms. However, it is still not clear whether ATML1 activity is indeed reduced post-transcriptionally and whether this reduction contributes to the restriction of ATML1 activity to the outermost cells. In this paper, we show that ATML1 protein was rarely detected in the inner cells of the embryos, although ATML1 was weakly transcribed in these cells. Moreover, we demonstrate that nuclear accumulation of ATML1 was attenuated, especially in the inner cells, during embryogenesis. We also performed domain deletion experiments and found that the ZLZ-coding sequence was necessary and partially sufficient for the attenuation of ATML1 nuclear accumulation and the suppression of ATML1 activity in the inner cells. Our study implies that spatial restriction of ATML1 activity by post-transcriptional mechanisms may facilitate the formation of the single epidermal layer, a common feature in many seed plants.
Reliable new reporters for ATML1 transcription in the embryos
In our previous reports, ATML1 transcription was visualized using only its promoter sequence (Takada and Jürgens, 2007). Therefore, the reporter expression pattern might not reflect the real transcriptional activity of ATML1. To visualize the native transcriptional dynamics of ATML1, we generated transgenic plants that expressed the triple GFP reporter gene fused with the Simian Virus 40 (SV40) T-antigen nuclear localization signal (NLS) under the control of the ATML1 regulatory sequence, which included a 3.4 kb promoter sequence, exons, introns and a 3′ downstream sequence (gATML1-nls-3xGFP; Fig. 1A). gATML1-nls-3xGFP signals were detected from as early as the one- or two-cell stage (12 of 12 lines) and were detected preferentially in the outermost cells from the early globular stage (12 of 12 lines; Fig. 1D-H). These expression patterns were comparable with the 3.4 kb ATML1 promoter activity shown by Takada and Jürgens (2007). These results suggest that the 3.4 kb ATML1 promoter contains sufficient regulatory regions to drive the same expression pattern as the 9.2 kb genomic sequence used in gATML1-nls-3xGFP.
ATML1 protein was not detected in the inner cells of the embryos from the 32-cell stage
Detailed observation of gATML1-nls-3xGFP plants revealed that weak GFP signals were detected in the inner cells of the embryos from the 32-cell stage (Fig. 1E,F and Fig. S1A-C). Weak GFP signals were still observed in the inner cells of the heart-stage embryos, especially in the L2 layer (12 of 12 lines), which suggests that ATML1 is weakly transcribed in the inner cells (Fig. 1G,H and Fig. S1D-F). Our previous study has shown that ectopic ATML1 expression induces epidermal cell identity in the inner tissues of the cotyledons and leaves (Takada et al., 2013). Therefore, ATML1 activity should be repressed in the inner cells to avoid the formation of multiple epidermal layers. To test whether ATML1 activity is suppressed after its transcription, we observed ATML1 protein localization by generating translational fusion lines, in which a single or triple GFP reporter gene fused in-frame with the ATML1-coding sequence was expressed under the control of the native regulatory sequence of ATML1 (gATML1-1xGFP and gATML1-3xGFP; Fig. 1B,C). These constructs rescued the embryonic lethal phenotype of atml1-3;pdf2-1, which indicates that these fusion proteins were functional (Table S1). GFP signals were detected from the one- or two-cell stage in gATML1-1xGFP and gATML1-3xGFP (9 of 9 and 10 of 11 lines, respectively; Fig. 1I,N), as observed in gATML1-nls-3xGFP. Unlike in gATML1-nls-3xGFP, however, gATML1-1xGFP and gATML1-3xGFP signals were decreased or not detected in the inner cells of some 16-cell-stage embryos (11 of 26 and 28 of 69 embryos, respectively; Fig. 2A-C). From the 32-cell stage, GFP signals were detectable mostly in the outermost cells (10 of 10 lines in gATML1-1xGFP and 10 of 10 lines in gATML1-3xGFP; Fig. 1J-M,O-R and Fig. S1G-L). These results suggest that ATML1 protein accumulation is post-transcriptionally restricted to the outermost cells from the 32-cell stage. Treatment of the gATML1-3xGFP embryos with a proteasome inhibitor MG132 did not affect the 3xGFP-ATML1 protein accumulation pattern, which suggests that 26S proteasome-mediated protein degradation is not involved in the inhibition of ATML1 accumulation in the inner cells (Fig. S2).
Our results also suggested that ATML1 protein does not move intercellularly, at least between the outermost cells and the inner cells, because 3xGFP-ATML1 fusion protein, the cell-cell movement of which should be limited due to the large triple GFP reporter, showed the same localization as 1xGFP-ATML1 protein (Fig. 1J-M,O-R; Kim et al., 2005).
Subcellular localization of ATML1 changed during embryogenesis
We found that ATML1 protein showed a change in subcellular localization during embryogenesis. gATML1-1xGFP and gATML1-3xGFP signals were detected only in the nuclei until the eight-cell stage. Whereas some 16-cell-stage embryos of gATML1-1xGFP and gATML1-3xGFP showed GFP signals only in the nuclei, other embryos showed GFP signals not only in the nuclei but also in the cytoplasm of the embryo-proper cells (Fig. 2A-C). After the 32-cell stage, gATML1-1xGFP and gATML1-3xGFP signals were detected in both the nuclei and cytoplasm in the outermost cells of the embryo proper (10 of 10 and 12 of 12 lines, respectively; Fig. 1J,K,O,P). The cytoplasm-to-nuclei signal ratios in gATML1-1xGFP and gATML1-3xGFP were significantly higher than those in gATML1-nls-3xGFP (Fig. S3A). Weak nuclear localization of GFP-ATML1 was evident, particularly in the protoderm of the shoot apical meristem (SAM) region and the adaxial side of cotyledons at the heart stage (9 of 9 lines in gATML1-1xGFP and 11 of 11 lines in gATML1-3xGFP; Fig. 1L,M,Q,R). By contrast, the suspensor cells did not show a cytoplasmic GFP-ATML1 signal in all gATML1-1xGFP lines and most gATML1-3xGFP lines examined (10 of 10 and 9 of 10 lines, respectively; Fig. 1I-K,N-P).
It is possible that these cytoplasmic gATML1-1xGFP and gATML1-3xGFP signals were caused by spontaneous cleavage of GFP from the fusion proteins. To test this possibility, we fused an NLS sequence to the N- or C-terminus of 3xGFP-ATML1 (NLS-3xGFP-gATML1 and 3xGFP-gATML1-NLS, respectively). If the cytoplasmic GFP signals indeed represented cleavage products of 3xGFP-ATML1, these cytoplasmic GFP signals should be decreased in NLS-3xGFP-gATML1, but not in 3xGFP-gATML1-NLS. If the cytoplasmic signals were not derived from cleavage products, GFP signals in NLS-3xGFP-gATML1 should show the similar localization as those in 3xGFP-gATML1-NLS. Nuclear-to-cytoplasmic GFP signal intensity ratios were increased in NLS-3xGFP-gATML1 compared with gATML1-3xGFP embryos, although GFP signals were still detected in the cytoplasm (4 of 4 lines; Fig. 2D and Fig. S3B). This increased nuclear accumulation of GFP was also detected in 3xGFP-gATML1-NLS embryos (11 of 11 lines), which suggests that addition of a strong NLS sequence at either the N- or C-terminal end similarly enhances the nuclear import of 3xGFP-ATML1, and at least a large part of cytoplasmic gATML1-3xGFP signals are not caused by the cleavage of 3xGFP-ATML1 (Fig. 2E and Fig. S3B). We also performed western blot analyses using antibodies against GFP and confirmed that free GFP was not detected in protein extracts from gATML1-3xGFP (Fig. S4). These results suggest that nuclear localization of ATML1 is repressed in the outermost cells of the embryo proper from the 16-cell stage. We also found that gATML1-1xGFP and gATML1-3xGFP signals were detected in both the nuclei and cytoplasm in the epidermis of the post-embryonic root tip, which suggests that the nuclear localization of ATML1 tends to be inhibited in the young epidermis, independently of organ identity (Fig. S5). To gain additional mechanistic insights into ATML1 protein localization, gATML1-3xGFP plants were treated with a nuclear export inhibitor, leptomycin B (LMB; Fig. S6). We found that cytoplasmic gATML1-3xGFP signals were reduced in LMB-treated embryos and roots, which suggests that shuttling between the nucleus and cytosol is involved in determining ATML1 nuclear localization (Fig. S6).
A large portion of ATML1 protein was localized to the cytoplasm in the inner cells of the embryos
Although ATML1 protein was detected mainly in the epidermis, gATML1-3xGFP signals were still detected in a small population of subepidermal cells. We found that 11.1% of the embryos showed GFP signals in an inner cell at the globular stage (n=9) and 60.0% at the heart stage (n=15). In these ATML1-positive inner cells, the cytoplasmic-to-nuclei signal ratios of gATML1-3xGFP were higher compared with those in the outermost cells and GFP signals showed less co-localization with a nuclear marker at the heart stage (7 of 9 embryos), which implies that ATML1 is primarily localized to the cytoplasm in the inner cells of the embryos (Fig. 3A-F and Fig. S7). To further investigate the subcellular localization of ATML1 protein in the inner cells, we generated transgenic plants in which GFP-ATML1 expression was induced in the whole embryo upon estradiol treatment (RPS5A>>GFP-ML1). The GFP-ATML1-overexpressing globular-stage embryos did not show clear nuclear localization of GFP signals in most of the inner cells (14 of 15 embryos; Fig. 3M and Fig. S8). GFP-ATML1 signals in some inner cells were strong in the nuclei, as seen in the outermost cells (Fig. 3M).
Next, to examine whether the enhanced cytoplasmic accumulation of ATML1 protein requires the inner cell lineage, we observed GFP-ATML1 localization in the inner daughter cells of the protoderm. We utilized the fass (fs) and hanaba taranu (han) mutant embryos that exhibit aberrant cell division orientation. The FS gene encodes a B″ regulatory subunit of PP2A, which is necessary for the proper orientation of interphase cortical microtubules and the formation of preprophase bands (Camilleri et al., 2002). Mutations in the FS gene cause abnormal cell division orientation during embryonic and postembryonic development (Torres-Ruiz and Jürgens, 1994). Loss-of-function mutations in HAN, which encodes a GATA transcription factor, cause pleiotropic developmental defects including abnormal cell division orientation during embryogenesis (Zhao et al., 2004; Nawy et al., 2010; Kanei et al., 2012). In fs mutant embryos, the nuclear-to-cytoplasmic ATML1 signal ratios were reduced in putative inner daughter cells of the epidermis compared with those in the outer daughter cells (Fig. 3G-I). Because cell arrangements were severely disrupted in fs embryos, it was sometimes difficult to distinguish whether cells were located at the surface or not, under a confocal laser scanning microscope. For this reason, we then observed gATML1-3xGFP signals in han, which showed much milder cell division defects. han mutant embryos often showed periclinal divisions in the epidermal layer, which enabled us to easily observe 3xGFP-ATML1 localization in the inner daughter cells of the epidermis. In han embryos, 3xGFP-ATML1 signals were detected throughout the cytoplasm and nuclei in the inner daughter cells of the epidermis (Fig. 3J-L). Although nuclear shapes were clearly visible in the outermost cells, the outlines of nuclei were obscure in 13 of 16 inner cells when observed under GFP channel. These results suggest that nuclear localization of ATML1 is more strongly inhibited in the inner cells compared with the outermost cells, irrespectively of the lineage of the inner cells.
The amount of ATML1 protein in the nuclei showed a weak positive correlation with the promoter activity of ACR4
Next, we investigated the biological significance of the inhibition of ATML1 nuclear localization. It has been reported that presence of a protein in the cytoplasm is required for the intercellular movement of the protein (Crawford and Zambryski, 2000; Gallagher et al., 2004). However, as described above, gATML1-1xGFP and gATML1-3xGFP showed the same GFP localization pattern, which suggests that ATML1 protein accumulation in the cytoplasm does not facilitate ATML1 movement between the outermost cells and the inner cells. Rather, we speculated that reduction of ATML1 nuclear localization might decrease the ATML1-mediated transcription in the nuclei. To examine this hypothesis, ATML1 protein localization and the promoter activity of ARABIDOPSIS THALIANA HOMOLOGUE OF CRINKLY4 (ACR4), a direct target of ATML1, were visualized simultaneously (see Materials and Methods; Fig. 3A-F). Expression of ACR4, which encodes a leucine-rich repeat receptor-like kinase, is directly activated by ATML1 in the embryos (Tanaka et al., 2002; San-bento et al., 2014; N. Takada, A.Y., H.I. and S.T., unpublished). Therefore, the ACR4 promoter activity is expected to reflect the transcriptional activity of the ATML1 protein. We measured the fluorescence intensities of GFP and TdTomato signals in the nuclei of the outermost cells at the heart stage and found that there was a weak, but statistically significant, positive correlation between these signal intensities (Fig. 3N and Fig. S9H). We also examined the correlation between ATML1 nuclear accumulation and ATML1 promoter activity in the heart-stage embryos using the nls-TdTomato reporter (proATML1-nls-TdTomato) as ATML1 itself is also a direct target of ATML1 (Takada et al., 2013; San-bento et al., 2014). The analysis showed that there was a statistically significant positive correlation between these signal intensities (Fig. S9F,G).
In addition, we also examined whether the nuclear localization of ATML1 is correlated with the activation of downstream genes FIDDLEHEAD (FDH) and ATML1 in the overexpression line (Takada et al., 2013; Takada, 2013). We found that constitutive expression of ATML1 was able to induce ATML1 and FDH promoter activity in the L3 layer of cotyledon primordia but not in the stele of the hypocotyl in the heart-stage embryos (Fig. S9A-D). Consistently with the absence of downstream gene activation, ATML1 protein accumulation was weak in the nuclei of the stele (Fig. S9E).
This result suggests that inhibition of nuclear localization potentially represses the transcriptional activity of ATML1 protein. Considering the weak nuclear localization of ATML1 in the inner cells, this inhibition might contribute to restrict ATML1 activity to the outermost cells.
ZLZ domain was required for the inhibition of ATML1 nuclear localization during embryogenesis
To further gain the mechanical insights of the post-transcriptional repression of ATML1, we performed domain deletion analyses. Among the four domains of ATML1, we focused on ZLZ and START domains because of their potential to interact with other regulatory molecules: the ZLZ domain is known as a dimerization motif and the START domain is predicted as a lipid/sterol-binding domain (Schrick et al., 2014). To assess the role of each domain in ATML1 localization, we generated transgenic lines that expressed 3xGFP-ATML1 fusion genes, with or without deletion of ZLZ or START, under the control of the ATML1 promoter (proML1-3xGFP-ML1, proML1-3xGFP-ML1ΔZLZ and proML1-3xGFP-ML1ΔSTART). GFP signals were detected in both the nuclei and cytoplasm at the heart stage in proML1-3xGFP-ML1 as observed in gATML1-3xGFP plants (8 of 8 lines; Fig. 4A,D,E and Fig. S10A,E). 3xGFP-ATML1ΔSTART signals, although slightly decreased in one line, showed cytoplasmic GFP signals in the epidermis of the heart-stage embryos, which suggests that the START domain is not essential for the inhibition of nuclear localization at this stage (10 of 10 lines; Fig. 4C,H,I and Figs S3C and S10C,E). In contrast, 3xGFP-ATML1ΔZLZ signals were not detected in the cytoplasm of the epidermis at the heart stage (10 of 10 lines; Fig. 4B,F,G and Figs S3C and S10B,E). The wild-type plants did not show fluorescence signals above background (Fig. S10D-F). This observation suggests that there is an active mechanism to attenuate nuclear localization of ATML1 and that the ZLZ domain is necessary for the inhibition of ATML1 nuclear localization in the epidermis.
Next, to examine the roles of ZLZ and START in subcellular localization of ATML1 in the inner cells, we generated estradiol-inducible lines that expressed the GFP-ATML1ΔZLZ or GFP-ATML1ΔSTART fusion gene under the control of the RPS5A promoter (RPS5A>>GFP-ML1ΔZLZ and RPS5A>>GFP-ML1ΔSTART). Two days after estradiol induction, GFP-ATML1ΔZLZ and GFP-ATML1ΔSTART signals were primarily found in the nuclei of the inner cells at the globular stage (19 of 20 and 12 of 12 embryos, respectively), whereas no preferential nuclear accumulation of ATML1 was observed in the inner cells of RPS5A>>GFP-ML1 embryos at the same stage (Figs 3M and 4J,K). In addition, we noticed that RPS5A>>GFP-ML1ΔSTART showed weak GFP signals, which suggests that the START-coding sequence may be required to increase the stability or amount of ATML1 mRNA or ATML1 protein (Fig. 4K). To clearly observe ATML1ΔSTART localization, 3xGFP was fused to ATML1ΔSTART. Constitutive expression of 3xGFP-ATML1ΔSTART and 3xGFP-ATML1ΔZLZ showed stronger nuclear localization compared with that of 3xGFP-ATML1 even in the inner cells of the embryos (Fig. S8). These results suggest that both ZLZ and START domains are necessary for the inhibition of nuclear localization in the inner cells of the embryos.
Deletion of ZLZ caused accumulation of ATML1 in the inner cells of the 32-cell-stage embryos
As GFP signals in proML1-3xGFP-ML1, proML1-3xGFP-ML1ΔZLZ and proML1-3xGFP-ML1ΔSTART were weak at the early stages compared with those in gATML1-3xGFP, it was difficult to observe ATML1ΔZLZ and ATML1ΔSTART protein localization in the early embryos (data not shown). Weak GFP signals in these transgenic lines at the early stages imply that there are enhancer sequences in the introns and/or the 3′ downstream region of ATML1. Because the ZLZ domain but not the START domain is encoded within a single exon, we were able to delete the ZLZ-coding sequence from the gATML1-3xGFP construct without affecting the exon-intron boundary structure to express 3xGFP-ATML1ΔZLZ under the control of the native ATML1 regulatory sequence (gATML1ΔZLZ-3xGFP). gATML1ΔZLZ-3xGFP signals were detected only in the nuclei even after the 32-cell stage (8 of 8 lines), which is consistent with the observation in proML1-3xGFP-ML1ΔZLZ (Fig. 5A-D). Whereas gATML1-3xGFP signals were restricted to the outermost cells from the 32-cell stage, weak gATML1ΔZLZ-3xGFP signals were detected also in the nuclei of the inner cells at the 32-cell stage (6 of 6 lines; Figs 1O,P and 5A,B). However, at the heart stage, gATML1ΔZLZ-3xGFP signals were not detected in the inner cells (8 of 8 lines; Fig. 5C,D). These observations suggest that the ZLZ-coding sequence inhibits ATML1 protein accumulation in the early embryos. Moreover, absence of detectable gATML1ΔZLZ-3xGFP signals in the inner cells of the heart-stage embryos implies the existence of ZLZ-coding sequence-independent mechanisms to prevent ATML1 protein accumulation.
The ZLZ-coding sequence alone was not fully sufficient for the post-transcriptional repressions of ATML1
To further ascertain whether the ZLZ-coding sequence alone is sufficient for the post-transcriptional repression of ATML1, NLS-3xGFP fused with the ZLZ-coding sequence was expressed under the control of the native ATML1 regulatory sequence (gATML1-nls-3xGFP-ZLZ). gATML1-nls-3xGFP-ZLZ signals were detected only in the nuclei even after the 32-cell stage, which suggests that the ZLZ domain is not sufficient for the inhibition of nuclear localization (10 of 10 lines; Fig. 5E-H and Fig. S3B). We also examined whether GFP signals were detected in the inner cells of the gATML1-nls-3xGFP-ZLZ embryos. Whereas gATML1-nls-3xGFP-ZLZ signals were weakly detected in the inner cells of the embryos around the 32-cell stage (9 of 10 lines), GFP signals were not observed in the inner cells of the heart-stage embryos (10 of 10 lines; Fig. 5E-H). These results suggest that ZLZ is sufficient for the inhibition of protein accumulation only in the late-stage embryos. Taken together, we conclude that the ZLZ-coding sequence is necessary, but not entirely sufficient, for the post-transcriptional repressions of ATML1.
Position-dependent and position-independent repressions of ATML1 activity
Despite weak transcriptional activity of ATML1 in the inner cells, ATML1 protein was accumulated only in the outermost cell layer, which implies that ATML1 protein synthesis is inhibited or degradation of ATML1 mRNA or ATML1 protein is enhanced in an inner cell-specific manner. In addition to position- or cell-type-dependent inhibition, it is also possible that ATML1 protein accumulation is reduced equally in both the outermost cells and inner cells. Because the expression of ATML1 mRNA was weak in the inner cells, general attenuation of ATML1 protein accumulation in the whole embryo would result in the outermost cell-specific localization of ATML1 protein. Although we cannot exclude either of these two possibilities, the attenuation of protein accumulation in all cells appears to be more simple and feasible than the inner cell-specific inhibition. Equal accumulation of ATML1 protein in the inner and outermost cells of ATML1-overexpressing embryos supports the latter possibility. Therefore, we assume that the reduction of ATML1 protein accumulation might not be position-dependent.
By contrast, nuclear localization of ATML1 in the inner cells appeared to be strongly inhibited compared with that in the outermost cells. Firstly, gATML1-3xGFP signals were occasionally detected in subepidermal cells and these cells showed weaker nuclear localization of ATML1 than those of the epidermis. Secondly, in the GFP-ATML1 overexpression experiments, ATML1 protein was present equally both in the nuclei and in the cytoplasm in most of the inner cells of the embryos. Lastly, in fs and han mutant embryos, ATML1 was not clearly localized to the nuclei in the epidermis-derived inner cells. These results suggest that the subcellular localization of ATML1 is affected not by cell types or cell lineage, but by the position of the cells.
ZLZ might, in combination with other domains, facilitate the post-transcriptional or post-translational repression of ATML1
Our domain deletion analyses showed that there are ZLZ-dependent mechanisms for post-transcriptional repression of ATML1. The ZLZ-coding sequence was necessary for the repression of ATML1 protein accumulation in the inner cells of the 32-cell-stage embryos. Also, the ZLZ-coding sequence alone was sufficient to inhibit protein accumulation in the inner cells of the heart-stage embryos. However, ATML1ΔZLZ protein accumulation was still restricted to the outermost cells in the heart-stage embryos, which suggests that a ZLZ-independent mechanism also exists to attenuate ATML1 protein accumulation in the inner cells. This idea is consistent with the observation that the ZLZ-coding sequence alone was not sufficient for the inhibition of the nls-3xGFP accumulation in the inner cells of 32-cell-stage embryos.
We showed that ATML1ΔZLZ protein was clearly localized to the nuclei of the outermost cells and the inner cells. However, ZLZ did not decrease the nuclear localization of nls-3xGFP, which suggests that ZLZ is necessary, but not sufficient, for the repression of nuclear localization. This result implies that nuclear localization is reduced through the combinatory effects of ZLZ and other domains. The START domain might be a candidate domain that affects nuclear localization because ATML1ΔSTART was mainly localized to the nuclei in the inner cells of the ATML1ΔSTART-overexpressing embryos. These results suggest that the outermost cells and the inner cells use different mechanisms to reduce nuclear localization of ATML1, and that ATML1 nuclear localization in the inner cells was inhibited through ZLZ and START domains.
ZLZ and START domains were shown to interact with other molecules or act as target sites for post-translational modification (Tron et al., 2002; Schrick et al., 2014). It has been reported that the START domain of ATML1 stimulated transcription factor activity in yeast in response to increased sterol biosynthesis (Schrick et al., 2014). Therefore, HD-ZIP class IV START domains may be involved in post-translational activation of transcription factors. Finding interactors of ZLZ and START domains and analyzing their functions should be promising approaches for future research to elucidate the position-dependent changes of ATML1 activity.
Post-transcriptional repressions may facilitate the position-dependent cell fate decision
We propose that both transcriptional activation and post-transcriptional repression are important to restrict ATML1 activity to the outermost cells. Our results suggest that outermost cell-specific localization of ATML1 might be explained by the combinatory effects of strong ATML1 transcription in the outermost cells and the reduction of ATML1 protein accumulation in the whole embryos. Even if a small amount of ATML1 protein were mistakenly accumulated in the inner cells, it would not induce ectopic epidermal cell identity because the inner cell-specific strong attenuation of ATML1 nuclear localization inhibits ATML1 activity in the inner cells. Transcriptional activation appears to play a major role in determining ATML1 activity because overexpression of ATML1 induced ectopic epidermal cell differentiation even in the inner cells of the leaves and cotyledons (Takada et al., 2013). We assume that post-transcriptional repressions of ATML1 counteract an excess of ATML1 activity.
As the ATML1 promoter contains an L1-box sequence, an ATML1 binding site, it is expected that ATML1 directly enhances its own expression (Abe et al., 2001). This positive feedback suggests that once ATML1 is expressed in a cell, ATML1 expression is maintained in its daughter cells. However, when epidermal cells accidentally undergo periclinal cell divisions, inner daughter cells do not maintain epidermal cell identity (Stewart and Dermen, 1975; Takada and Iida, 2014). Thus, we speculate that the position-dependent and position-independent post-transcriptional repressions of ATML1 activity are likely to suppress the positive feedback loop of ATML1 and, therefore, ectopic epidermal cell specification in the inner cells, to enable single epidermal layer formation.
In the present work, it was not possible to test whether enhanced nuclear localization and stabilization of ATML1 affected single epidermal layer formation. Although ATML1ΔZLZ showed nuclear accumulation in the inner cells of the globular-stage embryos, deletion of the dimerization domain eliminates the function of ATML1. In addition, NLS-3xGFP-gATML1 and 3xGFP-gATML1-NLS, which were able to rescue the embryonic lethal phenotype of atml1-3;pdf2-1, did not cause ectopic epidermal cell differentiation in the inner tissues (Fig. S11 and Table S1). Because reduction of ATML1 protein accumulation can also counteract ectopic activation of ATML1 in the inner tissues, multiple layers of ATML1 repression mechanisms might mask the defects of the cytoplasmic retention in NLS-3xGFP-gATML1 and 3xGFP-gATML1-NLS.
MATERIALS AND METHODS
Plant materials and growth conditions
Mutants used in this experiment, atml1-3;pdf2-1, fs-1 and han-30, have been described previously (Torres-Ruiz and Jürgens, 1994; Kanei et al., 2012; Ogawa et al., 2015). Plants were germinated and grown on Murashige and Skoog (MS) plates that contained 0.4% phytagel and 1% sucrose at 22°C. After 2-3 weeks, plants were transferred to soil and grown at 18°C or 22°C.
Ovules were cultured in liquid Nitsch medium (Duchefa Biochemie) that contained MES-KOH, vitamins and trehalose dihydrate, as described previously (Gooh et al., 2015). For the estradiol induction, β-estradiol, dissolved in dimethyl sulfoxide (DMSO), was added to the liquid medium to a final concentration of 10 µM. The same volume of DMSO was added to the control medium.
In inhibitor treatment experiments, seedlings or ovules were grown for 24 h in liquid media (MS for seedlings and Nitsch for ovules) that contained 10 µM MG132 (Sigma-Aldrich) or 2 µM LMB (Santa Cruz Biotechnology). The same volume of DMSO (for MG132) or ethanol (for LMB) was added to the control medium.
Plasmid construction and transgenic plants
gATML1-nls-3xGFP, gATML1-1xGFP and gATML1-3xGFP in pBar vector
A 4.9 kb region including the ATML1 promoter sequence and the 5′ untranslated region (from −3055 to +1926) was amplified by polymerase chain reaction (PCR) using primers 11218 and 11109 (primer sequences can be found in Table S2) and cloned into the KpnI and SmaI sites of the pBluescript KS (pBKS) vector (Stratagene) (KpnI-AscI-gATML1-half-SmaI/pBKS). The ATML1-coding sequence and 3′ downstream sequence (from +1927 to +6190) was amplified by PCR using primers 11110 and 11111 (Table S2) and cloned into the SmaI and SacI sites of the KpnI-AscI-gATML1-half-SmaI/pBKS (KpnI-AscI-gATML1-full-SacI/pBKS). The nls-3xGFP-nost sequence was excised from the nls-3xGFP-nost/pBKS vector (Takada and Jürgens, 2007) and cloned into the SmaI site of the KpnI-AscI-gATML1-full-SacI/pBKS vector (KpnI-AscI-gATML1-nls-3xGFP-nost-SacI/pBKS). The gATML1-nls-3xGFP-nost fragment was excised from KpnI-AscI-gATML1-nls-3xGFP-nost-SacI/pBKS with AscI and cloned into the pBar-linker vector, in which an AscI-XhoI linker sequence was added to the multiple cloning site of the pBar-A vector (GenBank: AJ251013), to generate gATML1-nls-3xGFP/pBar. The gATML1-nls-3xGFP/pBar was used to transform the wild-type Columbia.
Fragments of single and triple GFP sequences followed by three glycine codons were generated by PCR and cloned into the SmaI site of the KpnI-AscI-gATML1-full-SacI/pBKS vector (KpnI-AscI-gATML1-1xGFP-SacI/pBKS and KpnI-AscI-gATML1-3xGFP-SacI/pBKS). The gATML1-1xGFP and gATML1-3xGFP fragments from the KpnI-AscI-gATML1-1xGFP-SacI/pBKS and KpnI-AscI-gATML1-3xGFP-SacI/pBKS vectors were inserted into the AscI site of the pBar-linker vector (gATML1-1xGFP/pBar and gATML1-3xGFP/pBar). gATML1-1xGFP/pBar and gATML1-3xGFP/pBar were used to transform atml1-3;pdf2-1/+.
NLS-3xGFP-gATML1 and 3xGFP-gATML1-NLS in pBar vector
An SV40 NLS sequence was fused to the 5′ end of the 3xGFP reporter gene that lacked the stop codon, and cloned into the pGEM-5zf vector (Promega) (nls-3xGFP-nonstop/5zf). The nls-3xGFP-nonstop sequence was cloned into the BamHI site of the pBKS-Sma-Gly vector, in which SmaI sites were located both upstream and downstream of the BamHI site in pBKS and three glycine codons were added to the 5′ end of the downstream SmaI site (SmaI-nls-3xGFP-nonstop-Gly-SmaI/pBKS). The triple GFP sequence in the gATML1-3xGFP/pBar vector was replaced with the nls-3xGFP-Gly sequence that was excised from the SmaI-nls-3xGFP-nonstop-SmaI-Gly/pBKS vector using SmaI (NLS-3xGFP-gATML1/pBar). NLS-3xGFP-gATML1/pBar was used to transform atml1-3;pdf2-1/+.
A part of the ATML1-coding sequence from +3840 to +5590 relative to the transcriptional start site was amplified by PCR using primers 11949 and 11950 (Table S2). The amplified fragment was cloned into the SacI and SpeI sites of the pBKS vector (SacI-AflII-gATML1-SpeI/pBKS). An SV40 NLS sequence was fused in-frame to the 3′ end, i.e. immediately before the stop codon, of the partial ATML1-coding sequence in SacI-AflII-gATML1-SpeI/pBKS by PCR-mediated insertion using primers 12004 and 12003 (SacI-AflII-gATML1-NLS-SpeI/pBKS; Table S2). The AflII-gATML1-NLS-SpeI fragment that was excised from SacI-AflII-gATML1-NLS-SpeI/pBKS was cloned into the AflII and SpeI sites of the gATML1-3xGFP/pBar (3xGFP-gATML1-NLS/pBar). 3xGFP-gATML1-NLS/pBar was used to transform atml1-3;pdf2-1/+.
proML1-3xGFP-ML1, proML1-3xGFP-ML1ΔZLZ and proML1-3xGFP-ML1ΔSTART in the pBar vector
A BsrGI fragment was excised from the 2xGFP/5zf vector, in which two GFP-coding sequences were translationally fused and cloned into the pGEM-5zf vector, and inserted into the BsrGI site of the GFP-ATML1/pBKS, at which the GFP-ATML1 fusion gene was cloned in pBKS. Insertion of two BsrGI-GFP fragments resulted in 3xGFP-ML1/pBKS. 3xGFP-ML1 was excised from 3xGFP-ML1/pBKS with XhoI and SpeI and inserted into the XhoI and SpeI sites of the proATML1-35St/pBar vector, in which the 3.4 kb ATML1 promoter sequence and the SpeI-linker sequence followed by the cauliflower mosaic virus 35S terminator sequence (35St) were cloned upstream and downstream of the XhoI site of the pBar-linker, respectively (proML1-3xGFP-ML1/pBar; N. Takada, A.Y., H.I. and S.T., unpublished). proML-3xGFP-ML1/pBar was used to transform the wild-type Columbia.
GFP-ATML1 without the ZLZ- or START-coding sequence was generated by PCR amplifying GFP-ATML1/pBKS using primers 6422 and 6423, or 6420 and 6421, respectively (Table S2; GFP-ATML1ΔZLZ/pBKS and GFP-ATML1ΔSTART/pBKS). 3xGFP-ML1ΔZLZ and 3xGFP-ML1ΔSTART were generated with the same procedure as 3xGFP-ML1/pBKS by adding two BsrGI-GFP fragments, and these coding regions were inserted into the XhoI and SpeI sites of the proATML1-35St/pBar vector (proML1-3xGFP-ML1ΔZLZ/pBar and proML1-3xGFP-ML1ΔSTART/pBar). proML-3xGFP-ML1ΔZLZ/pBar and proML-3xGFP-ML1ΔSTART/pBar were used to transform the wild-type Columbia.
RPS5A>>GFP-ML1/ER8, RPS5A>>GFP-ML1ΔZLZ/ER8, RPS5A>>GFP-ML1ΔSTART/ER8, RPS5A>>3xGFP-ML1/ER8, RPS5A>>3xGFP-ML1ΔZLZ/ER8 and RPS5A>>3xGFP-ML1ΔSTART/ER8
GFP-ATML1, GFP-ATML1ΔZLZ, GFP-ATML1ΔSTART, 3xGFP-ATML1, 3xGFP-ATML1ΔZLZ and 3xGFP-ATML1ΔSTART regions from the GFP-ATML1/pBKS, GFP-ATML1ΔZLZ/pBKS, GFP-ATML1ΔSTART/pBKS, 3xGFP-ATML1/pBKS, 3xGFP-ATML1ΔZLZ/pBKS and 3xGFP-ATML1ΔSTART/pBKS vectors were inserted into the XhoI and SpeI sites of the RPS5A/ER8, which was described previously (RPS5A>>GFP-ML1/ER8, RPS5A>>GFP-ML1ΔZLZ/ER8, RPS5A>>GFP-ML1ΔSTART/ER8, RPS5A>>3xGFP-ML1/ER8, RPS5A>>3xGFP-ML1ΔZLZ/ER8 and RPS5A>>3xGFP-ML1ΔSTART/ER8; Takada et al., 2013). The wild-type Columbia was transformed with the RPS5A>>GFP-ML1/ER8, RPS5A>>GFP-ML1ΔZLZ/ER8, RPS5A>>GFP-ML1ΔSTART/ER8, RPS5A>>3xGFP-ML1/ER8, RPS5A>>3xGFP-ML1ΔZLZ/ER8 and RPS5A>>3xGFP-ML1ΔSTART/ER8 vectors.
gATML1ΔZLZ-3xGFP and gATML1-nls-3xGFP-ZLZ in the pBar vector
A part of the ATML1 genomic sequence that included the ZLZ-coding sequence was amplified by PCR using primers 13238 and 13239 (Table S2) and inserted into the BamHI and KpnI sites of the pBKS vector (BamHI-NruI-gATML1-AflII-KpnI/pBKS). The ZLZ-coding sequence was removed from BamHI-NruI-gATML1-AflII-KpnI/pBKS by PCR-based deletion using primers 6422 and 13427 (Table S2; BamHI-NruI-gATML1ΔZLZ-AflII-KpnI/pBKS). The gATML1ΔZLZ fragment from the BamHI-NruI-gATML1ΔZLZ-AflII-KpnI/pBKS vector was inserted into the NruI and AflII sites of the gATML1-3xGFP/pBKS vector using the SLiCE method (gATML1ΔZLZ-3xGFP/pBKS; Motohashi, 2015). The gATML1ΔZLZ-3xGFP fragment was inserted into the AscI site of the pBar-linker vector (gATML1ΔZLZ-3xGFP/pBar). gATML1ΔZLZ-3xGFP/pBar was used to transform the wild-type Columbia.
To generate gATML1-nls-3xGFP-ZLZ/pBar, the ZLZ-coding sequence was amplified using primers 13907 and 13908 (Table S2) and inserted into the BglII site of nls-3xGFP-nonstop/5zf (nls-3xGFP-ZLZ/5zf). The nopaline synthase terminator (nost) sequence was amplified using primers 13909 and 13910 (Table S2) and inserted after the ZLZ-coding sequence of nls-3xGFP-ZLZ/5zf to generate nls-3xGFP-ZLZ-nost/5zf. The nls-3xGFP-ZLZ-nost fragment was excised from nls-3xGFP-ZLZ-nost/5zf and inserted into the BamHI and NotI sites of the pBKS vector (XmaI-BamHI-nls-3xGFP-ZLZ-nost-XmaI-NotI/pBKS). The nls-3xGFP-ZLZ-nost fragment of the resulting construct was cloned into the XmaI sites of the gATML1-nls-3xGFP/pBar vector; so that the nls-3xGFP-nost region of gATML1-nls-3xGFP/pBar was replaced with nls-3xGFP-ZLZ-nost (gATML1-nls-3xGFP-ZLZ/pBar). The wild-type Columbia was transformed with the gATML1-nls-3xGFP-ZLZ/pBar vector.
proACR4-nls-TdTomato, proFDH-nls-TdTomato and pATML1-nls-TdTomato
The ACR4 promoter sequence (from −1222 to +616; gene structure AT3G59420.1 from Araport 11) was inserted into the XhoI and SmaI sites of the pGTV-KAN-derived pHM3 vector (Becker et al., 1992). The TdTomato reporter gene with an SV40 NLS sequence followed by the 35S terminator was cloned downstream of the ACR4 promoter sequence (proACR4-nls-TdTomato-35St/pHM3). The proACR4-nls-TdTomato-35St sequence was excised from the pHM3 vector and inserted into the XbaI and EcoRI sites of the BinHygTOp-derived pKH3 vector (proACR4-nls-TdTomato/pKH3; Höfgen et al., 1994). The proACR4-nls-TdTomato/pKH3 was used to transform the gATML1-3xGFP/pBar line T2#5-4.
The FDH promoter sequence (from −938 to +204; gene structure AT2G26250.1 from Araport 11) was inserted into the HindIII and SmaI sites of the pHM3 vector. The TdTomato reporter gene with an SV40 NLS sequence followed by the 35S terminator was cloned downstream of the FDH promoter sequence (proFDH-nls-TdTomato-35St/pHM3). The proFDH-nls-TdTomato/pHM3 was used to transform the RPS5A>>ATML1 line #7-1.
The nls-TdTomato reporter gene was inserted into XhoI and SpeI sites of the proATML1-35St/pBar vector to generate proATML1-nls-TdTomato/pBar. The proATML1-nls-TdTomato/pBar was used to transform the wild-type Columbia. A T3 line #5-2, homozygous for the proATML1-nls-TdTomato/pBar transgene, was crossed with the gATML1-3xGFP/pBar line T2#5-4.
Confocal laser scanning microscope and imaging analysis
For GFP, TdTomato, SR2200, 4′,6-diamidino-2-phenylindole (DAPI) and propidium iodide (PI) fluorescence observation, an LSM710 (Carl Zeiss) or FV1000 (Olympus) confocal laser scanning microscope was used. Embryos were fixed in 4% paraformaldehyde and 5% glycerol solution in phosphate-buffered saline (PBS) (4% PFA) before observation. For SR2200 staining, SR2200 was added to 4% PFA (2 µl/ml). For PI staining, roots were mounted in 10 µg/ml PI solution.
Intensity of GFP or TdTomato signals was measured using the imaging software Fiji (fiji.sc). To measure transcriptional activity of ATML1 protein, a homozygous gATML1-3xGFP line was transformed with the proACR4-nls-TdTomato/pKH3 vector, and T2 generation plants carrying proACR4-nls-TdTomato were used for embryo isolation and observation. To normalize the TdTomato signal intensity, the average of TdTomato signal intensity in each embryo was set to one.
The cytoplasmic-to-nuclear signal ratios shown in Fig. S3 were measured using Fiji. Nuclear regions were visualized by staining with DAPI. An optical section with the largest nuclear area from a confocal z-stack was used for quantification in each cell.
Western blot analysis
Proteins were extracted from ∼100 3 day-old seedlings in lysis buffer [125 mM Tris-HCl (pH 8.8), 1% sodium dodecyl sulfate (SDS), 10% glycerol, 50 mM sodium metabisulfite] and 5% of the total samples were subjected to SDS-polyacrylamide electrophoresis using a 10% gel. The separated protein was blotted onto Hybond-P PVDF membrane (Amersham) using an electroblotter Mini-PROTEAN II Cell (Bio-Rad). The blotted membranes were blocked for 1 h in 4% skimmed milk in PBS with 0.1% Tween-20 (PBST) (pH 7.0) at room temperature and incubated with a mouse monoclonal antibody against GFP (1:1000; Roche, 11814460001) for 1 h at room temperature. Membranes were washed in PBST and incubated with an ECL peroxidase-labeled anti-mouse secondary antibody (1:10,000; Amersham, NA931-100UL) for 1 h. After washing, the signals were detected on ChemiDoc XRS (Bio-Rad) using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).
We thank Prof. Nam-Hai Chua (Rockefeller University, USA), Prof. Taku Takahashi (Okayama University, Japan), Dr Ulrike Mayer (University of Tübingen, Germany) and Dr Gorou Horiguchi (Rikkyo University, Japan) for the ER8 vector, atml1-3;pdf2-1/+ seeds, fs-1 seeds and han-30 seeds, respectively. We also thank Dr Shunsuke Miyashima (Nara Institute of Science and Technology, Japan) and Hirofumi Ohmori (Osaka University, Japan) for sharing a method for cell wall staining and DNA sequencing work, respectively. We greatly acknowledge Prof. Tatsuo Kakimoto (Osaka University, Japan) for supporting our projects in his laboratory. We also thank all other members of the plant growth and development laboratory for their helpful comments and discussions.
Conceptualization: H.I., A.Y., S.T.; Investigation: H.I., A.Y., S.T.; Writing - original draft: H.I.; Writing - review & editing: H.I., S.T.; Supervision: S.T.; Project administration: S.T.; Funding acquisition: H.I., S.T.
This work was supported by grants from the Japan Society for the Promotion of Science to S.T. (JP20657012, JP22687003, JP23657036, JP26440142 and JP18K06286) and to H.I. (JP16J00702).
The authors declare no competing or financial interests.