Tapetal cell fate, lineage and proliferation in the Arabidopsis anther

Unlike animals, which can initiate germlines as early as embryogenesis, plant reproductive cells form late in development. All male reproductive cells of Arabidopsis develop within concentrically organised microsporangia derived from archesporial cells in the L2 of the stamen primordium (Fig. 1A,E). These archesporial cells divide periclinally to form an inner primary sporogenous cell and an outer primary parietal cell (Fig. 1B,E). Primary sporogenous cells continue division to give rise to a central sporogenous mass, and consecutive divisions of neighbouring cells result in the formation of three concentric parietal layers – the tapetum adjacent to the sporogenous cells, a middle layer, and the endothecium subjacent to the epidermis (Fig. 1C-E) (Canales et al., 2002; Feng and Dickinson, 2007; Scott et al., 2004; Sorensen et al., 2002). In common with the development of other plant reproductive organs, where consistency in size and shape is essential for effective function, microsporangial formation does not involve ‘classical’ plant meristems, but rather finite numbers of divisions within apparent cell lineages – reminiscent of development in some animal tissues.

Patterning of the concentric microsporangia involves the interplay of many genes (Feng and Dickinson, 2007; Ma, 2005; Wilson and Yang, 2004). The transcription factor SPOROCYTELESS (SPL; also known as NOZZLE, NZZ) (Schiefthaler et al., 1999; Yang et al., 1999) acts directly downstream of the floral ‘C’ gene AGAMOUS (AG) (Ito et al., 2004) to promote sporogenous cell identity (Schiefthaler et al., 1999; Yang et al., 1999) and is reported to interact with the leucine-rich repeat receptor-like kinases (LRR-RLKs) BARELY ANY MERISTEM 1 and 2 (BAM1 and BAM2) (Hord et al., 2006). BAM1 and BAM2 are held to promote somatic cell fates, first in the outward-facing products of the archesporial cell division, and subsequently during microsporangial expansion by restricting the expression domain of SPL to the locular centre (Hord et al., 2006). In turn, SPL appears to promote BAM1 expression, possibly creating a regulatory loop similar to that involving CLAVATA 1 (CLV1) and WUSCHEL (WUS) in shoot apical and floral meristems (Hord et al., 2006).

Also playing an important part in microsporangial patterning, the LRR-RLK EXCESS MICROSPOROCYTES 1 (EMS1; also known as EXTRA SPOROGENOUS CELLS, EXS) (Canales et al., 2002; Zhao et al., 2002) is reported to form a receptor complex with two further LRR-RLKs – SOMATIC EMBRYOGENESIS RECEPTOR-LIKE KINASE1 and 2 (SERK1 and SERK2) – in the plasma membrane of tapetal cells (Albrecht et al., 2005; Colcombet et al., 2005). This complex is held to bind TAPETUM DETERMINANT 1 (TPD1) (Yang et al., 2005; Yang et al., 2003), a putative ligand synthesised by the microsporocyte mass, and is reported to specify tapetal cell fate (Albrecht et al., 2005; Feng and Dickinson, 2007; Jia et al., 2008; Ma, 2005; Yang et al., 2005). ems1 microsporangia have no tapetal layer but excess microsporocytes (Canales et al., 2002; Zhao et al., 2002), a phenotype shared with serk1 serk2 double mutant lines (Albrecht et al., 2005; Colcombet et al., 2005) and tpd1 plants (Yang et al., 2003). How this EMS1 complex mediates the formation of a functional tapetal layer is unknown; models range from a role in regulating the number of functional archesporial cells (Canales et al., 2002) to the specification of tapetal fate in the inner division products of the inner secondary parietal cells, cells which, in ems1 mutant lines, are held to revert to a default microsporocyte fate – implying a switch in fate within a cell lineage (Zhao et al., 2002).

In an attempt to define further the role of the EMS1 complex in the specification of the Arabidopsis tapetum and the mechanism by which it proliferates as a monolayer, we expressed EMS1 ectopically in ems1 plants after its normal period of expression. These experiments show that a small population of founder cells is committed to a tapetal fate very early in development, but that the EMS1 complex is not involved in this early specification. However, the EMS1 complex is required for the subsequent proliferation of these cells to form the tapetal monolayer, but not divisions in any other cell layer. We also show competition for space to exist between the products of the microsporocyte and tapetal cell lineages. This early independence of the tapetal lineage and its ability to interact with neighboring cell layers provide the first clues as to how these small but complex plant organs develop – apparently in the absence of functional meristems.

Wild-type (wt) Arabidopsis thaliana (Linneaus) Landsberg erecta (Ler) plants were used unless otherwise specified. exs1, exs2, exs3 seeds were available in the laboratory (Canales et al., 2002). ems1 seeds were kindly provided by Hong Ma (Pennsylvania State University, USA) (Zhao et al., 2002); tpd1 seeds by De Ye (China Agricultural University, China) (Yang et al., 2003); and serk1 serk2 seeds by Sacco de Vries (Wageningen University, Holland) (Albrecht et al., 2005). Rod Scott (Bath University, UK) provided pA9::Barnase lines in both Ler and Columbia-0 backgrounds (Dilkes et al., 2008). The ems1 line was genotyped using XF301, XF504 and XF3R primers, and tpd1 (Yang et al., 2003) and serk1 serk2 (Albrecht et al., 2005) mutants were genotyped as previously described. Plants were grown at 21°C on a 16 hour light/8 hour dark cycle.

Biological replicates comprised floral buds at flower stages (Smyth et al., 1990) 1-14, collected from three plants, and three independent replicates were used from each line. Quantitative RT-PCR (qRT-PCR) was performed with an Applied Biosystems 7300 Real-Time PCR system, using XG102F/XG102R primers for AtA9 and At/AhEF1a primers (Becher et al., 2004) as a control. Thirty-five cycles of RT-PCR were carried out (PyrobestDNA polymerase, TaKaRa) using XG1206F/XG1206R primers for EMS1 and GAPDH primers (Hay et al., 2002) as a control.

4005 bp of EMS1 genomic DNA and 979 bp of the A9 promoter (pA9) were cloned using primers XF201/XF202, and XF203/XF204, respectively. The pA9::EMS1 fragment was amplified through interlaced PCR using primers XF201/XF204. This fragment was cloned into pMLBART (Gleave, 1992) using NotI digestion. The same fragment was also cloned into Gateway pDONR207 vector (Invitrogen) after amplification using primers XL101/XL102, and finally recombined into Gateway vector pGWB16 (Invitrogen). Both constructs were then introduced into ems1/+ heterozygous plants using floral dip transformation (Zhang et al., 2006). The pMLBART-pA9::EMS1 and pGWB16-pA9::EMS1-6xcMYC T1 plants resulting were screened for BASTA, and for Kanamycin and Hygromycin resistance, respectively, and were genotyped for the transgene using XH201/XH203 primers and for the genomic EMS1 allele using XF301/XF504/XF3R primers. T4 plants from each independent transgenic line were genotyped and examined for phenotype.

Interpretation of our experiments required precise knowledge of the timing of pA9 expression in our plant lines; previous reports had only described pA9 driven reporter constructs as being tapetal specific (Sorensen et al., 2002). We thus investigated tapetal ablation in pA9::Barnase expressing lines of Arabidopsis (Dilkes et al., 2008; Paul et al., 1992). These plants form a full complement of microsporocytes and tapetal cells, but the latter become highly vacuolated during the proliferation phase (S5; see Fig. S1 in the supplementary material), expand rapidly and degenerate after S5, confirming pA9 to be tissue specific and active during the proliferation of the tapetum.

Material was prepared as previously described (Sorensen et al., 2002), observed under an Olympus BX50 microscope and imaged using Image-Pro 6.2 software. The longitudinal mid-point of abaxial microsporangia from the medial anthers of anther stage 6 was chosen for cell number measurement. Mean value and s.d. were calculated from 150 independent locules of each line.

Material was prepared as previously described (Sorensen et al., 2002).

A 1157-bp fragment for EMS1 detection was amplified using primers XH402F/XH402R (Zhao et al., 2002) using wild-type cDNA as template; similarly, a 305-bp A9 fragment was amplified using primers XH401F/XH401R, and a fragment for A7 detection was made using DNA from vector pMC1577 kindly provided by Hong Ma (Zhao et al., 2002; Rubinelli et al., 1998). Fragments were cloned into the pGEM T-Easy vector (Promega) and checked by sequencing. In situ hybridisation was carried out on 8-μm paraffin wax-embedded sections using antisense and sense probes for each gene as previously described (Lincoln et al., 1994).

See Table S1 in the supplementary material for a list of primers used.

To investigate the role of the EMS1 complex in tapetal development, we first examined events immediately following the division of the archesporial cell in wild-type Arabidopsis (Fig. 1A-I). At this stage, and only in the region under the radius of the anther lobe (distant from the connective), the primary parietal cell and its adjacent two or three L2 cells (in transverse section) surrounding the primary sporogenous cell divide to form outer, and inner, secondary parietal cells (Fig. 1C). The primary sporogenous cells commence division at this point and, simultaneously, the inner secondary partietal cells adjacent to them divide periclinally to give rise to tapetal and middle cell layer initials, which we term primary tapetal cells (PTCs) and primary middle layer cells, respectively. This results in the sporogenous cells becoming surrounded by a small number of PTCs (Fig. 1F). Dramatic growth of the microsporangium then follows, achieved principally by cell expansion and anticlinal division of the investing layers, and one or two further, apparently unpolarised, divisions of the sporogenous cells. The PTCs participate actively in this proliferation through successive anticlinal divisions, maintaining a monolayer of cells between the microsporocyte mass and the dividing middle layer (Fig. 1G). The microsporocytes then enter meiosis and produce tetrads (Fig. 1H) from which individual microspores are released (Fig. 1I) to develop, via pollen mitosis 1 and 2, into pollen grains.

ems1 microsporangia develop similarly to wild type prior to anther stage 5 [S5, anther stages refer to Sanders et al. (Sanders et al., 1999) (Fig. 1J)]; however, the PTCs in these lines fail to proliferate, and as both sporogenous cells (on the inside), and the outer layers of the microsporangium (on the outside) continue to divide and expand, individual PTCs either remain grouped (Fig. 1K,L), or become separated from each other, isolated at the locular periphery (Fig. 1M,N). As these cells do not develop into functional tapeta in ems1 locules, we have termed them locular peripheral cells (LPCs) from this stage onwards. On average, each ems1 microsporangium contains 3.15±0.34 LPCs in transverse section at the mid-point of the abaxial microsporangium of medial anthers at anther stage 6. The locular content of ems1 plants degenerates shortly following the second meiotic nuclear division (Canales et al., 2002; Zhao et al., 2002), with LPCs showing dramatic hypertrophy (Fig. 1L,N).

To determine what ‘tapetal’ features are possessed by the LPCs in the ems1 mutants, we examined these cells using transmission electron microscopy (TEM) and investigated the expression of two tapetal markers, At5g07230 (A9) expressed early in development (Paul et al., 1992) and At4g28395 (A7) expressed post meiosis (Rubinelli et al., 1998). TEM analysis showed the LPCs to feature a high density of ribosomes, and active populations of Golgi bodies and mitochondria (see Fig. S2 in the supplementary material) – all features of tapetal cells. Quantitative RT-PCR (qRT-PCR) shows A9 to be expressed in wild-type and ems1 plants – but at a lower level than in the mutant (see Fig. S3 in the supplementary material). In situ hybridisation confirms the presence of A9 transcript in the developing tapetal cells of wild type (Fig. 2A-D), as well as solely in the LPCs of ems1 anthers (Fig. 2E-H). A7 is highly expressed in the active, late tapetal cells of wild-type anthers (see Fig. S4A in the supplementary material) and, although the loculi of ems1 plants degenerate after meiosis 2 (Canales et al., 2002), the A7 transcript can nevertheless be detected in the hypertrophic LPCs of these anthers using in situ hybridisation (see Fig. S4C,D in the supplementary material). A combination of position and marker gene expression thus clearly identifies the LPCs of ems1 locules as partially developed tapetal cells.

We investigated the possibility that the LPCs in ems1 plants result from low levels of normal or aberrant EMS1 expression by searching for functional EMS1 transcripts in ems1 plants using RT-PCR (see Fig. S5 in the supplementary material). No transcripts were detected, and further evidence that the LPCs do not result from incomplete penetrance of the ems1 mutation is provided by the presence of identical numbers of morphologically similar LPCs in exs-1, and exs-2 and exs-3 mutants in the C24 ecotype (data not shown).

To discover whether the few LPCs present in ems1 microsporangia can be ‘reactivated’ to generate functional tapetal cells (mimicking wild-type development), we expressed EMS1 in ems1 plants under the A9 promoter (pA9), which drives tapetal-specific expression during the proliferation stage of the tapetum (see Materials and methods) and which, in ems1 locules, is expressed only in the LPCs. In all 42 ems1 pA9::EMS1 and pA9::EMS1-cMYC lines generated, LPCs both proliferated and differentiated into apparently normal tapetal cells (see Fig. 3) – to which the EMS1 protein was restricted (data not shown). All lines produced fertile pollen, despite the comparatively late expression of EMS1 when driven by pA9 [EMS1 expression precedes that of A9 in the wild-type anther (Canales et al., 2002; Paul et al., 1992; Zhao et al., 2002)] (Fig. 2A-D,I-L), but each transgenic line possessed a different proportion of sterile locules. The tapetum was also restored in different patterns in different lines, varying from a normal monolayer (e.g. in ems1 pA9::EMS1 line #10; Fig. 3A-C) to clumps of multilayered tapetum (e.g. in ems1 pA9::EMS1 line #5; Fig. 3D-F).

pA9::EMS1 expression in ems1 microsporangia restores tapetal function sufficiently to produce viable pollen and, because only the LPCs express A9 in these locules, it follows that the functional tapeta of the ‘restored’ plants must be derived from this cell type. Furthermore, as the tapetum of some of the transgenic lines (e.g. ems1 pA9::EMS1 line #10) so closely resembles that of wild type, and because the LPCs themselves resemble tapetal cells both in position and in expression of the A9 (early) and A7 (late) tapetal markers, we conclude that the wild-type tapetal layer must also be derived from the few PTCs formed around the early sporogenous cells. The extra tapetal cells that are frequently generated in ems1 pA9::EMS1 lines are likely to be the result of the higher activity of pA9 compared with pEMS1 (Fig. 2I-P).

Although pA9 is activated after EMS1 transcription in wild-type plants (Canales et al., 2002; Paul et al., 1992; Zhao et al., 2002) (Fig. 2A-D,I-L), it must direct expression sufficiently early for the transgenic tapeta to follow closely the ‘wild-type’ developmental pathway. pA9::EMS1 rescues the ems1 mutant phenotype, but it fails to do so in either ems1 tpd1 double or ems1 serk1 serk2 triple mutant backgrounds, although EMS1 is strongly expressed in the LPCs of the mutant anthers (see Fig. S6 in the supplementary material). However, these transgenic plants have phenotypes very similar to those of ems1, tpd1 and serk1 serk2 mutants (Fig. 4), providing further evidence that these four genes are members of the same signalling pathway (Albrecht et al., 2005; Colcombet et al., 2005).

As the reduction in tapetal cell number in ems1 lines is accompanied by an increase in the microsporocyte population, we investigated the relationship between microsporocyte number and the level of tapetal proliferation in ems1 pA9::EMS1 transgenic plants. In wild-type plants, mid-point transverse sections of medial anthers with microsporangia at stage 6 transect 7.17±0.17 microsporocytes, whereas similar ems1 locules contain 14.38±0.32 microsporocytes. The number of microsporocytes in ems1 pA9::EMS1 microsporangia varies markedly between lines, but they are always in excess of those in wild-type plants, and less than those found in ems1 anthers. For example, ems1pA9::EMS1 line #5 sporangia contain 10.38±0.18 microsporocytes in transverse section (see Fig. S7 in the supplementary material). Similar comparisons cannot be carried out for tapetal cells as the number in ems1 pA9::EMS1 microsporangia cannot be measured accurately owing to their multiple division planes and irregularity in size and shape. However, the fact that ems1 anthers (which contain excess microsporocytes) contain few LPCs (about three per section), transgenic ems1 pA9::EMS1 lines (which contain fewer microsporocytes than ems1 plants) have significant numbers of tapetal cells, and wild-type anthers (which possess fewer microsporocytes than ems1 pA9::EMS1 lines) feature a complete tapetal monolayer, points to a clear inverse relationship between tapetal cell and microsporocyte numbers.

However, even when a functional monolayer of tapetal cells is formed in ems1 pA9::EMS1 lines (e.g. as in ems1 pA9::EMS1 line #10), the number of microsporocytes always exceeds that of wild-type anthers. pA9 is activated later in anther development than pEMS1, and this interval between expression stages may provide an opportunity for the microsporocytes to commence proliferation in the absence of a developing tapetal layer. Significantly, in all the ems1 and pA9::EMS1 ems1 lines screened, the numbers of cells composing the middle, endothecial and epidermal layers were identical to those of wild-type plants, showing that the proliferation of these cell layers is unaffected by the presence or absence of both EMS1 expression and a functional tapetal layer.

Many ems1 pA9::EMS1 microsporangia degenerate after S9 (Fig. 3F), and we found a clear relationship between microsporocyte number, the extent of tapetal disruption and locular fertility. Lines with a large number of microsporocytes were likely to possess more fragmented and multilayered tapetum, and a higher proportion of sterile locules. For example, ems1 pA9::EMS1 line #5, which contains 10.38±0.18 microsporocytes in transverse section (the highest number among all 24 transgenic lines), has the most irregular tapetum and the largest number of sterile locules (see Fig. S7 in the supplementary material).

Our screen of microsporangia from different transgenic lines revealed a wide range of phenotypes, including plants containing clumped or multilayered tapetum (Fig. 3D-F). We believe this, like the higher numbers of microsporocytes (see above), to result from differences in the timing of EMS1 function (Fig. 5). Thus, in different ems1 pA9::EMS1 lines, the promoter might drive EMS1 transcription to attain an effective threshold either before or after the fragmentation of the monolayer of LPCs around the sporogenous cells: if before, exclusively anticlinal division takes place; if afterwards, a mixture of anti- and periclinal division occurs to generate tapetal layering (Fig. 5). The association between the integrity of the tapetal monolayer and division polarity is striking, and points to the polarity of division (anticlinal in this case) being maintained by coordinating signals from neighbouring cells. Such signals might be transmitted through plasmodesmata, or generated by mechanical stress set up in the tapetal monolayer itself. Alternatively, division polarity could be regulated by the directionality of the TPD1 signal from the microsporocytes. In wild-type locules, this signal would be received on the inward facing surfaces of the tapetal cells, whereas in ems1 pA9::EMS1 lines, where LPCs are isolated, the TPD1 signal could be perceived on their radial (and perhaps outward) surfaces. Such a system has been reported for stomatal development, where the LRR-RLKs ERECTA (ER), ERECTA LIKE 1 (ERL1) (Shpak et al., 2005) and TOO MANY MOUTHS (TMM) (Geisler et al., 2000; Nadeau and Sack, 2002) form a heterodimer, which orients the cell division plate depending upon from which direction it receives the ligand EPIDERMAL PATTERNING FACTOR 1 (EPF1) (De Smet et al., 2009; Hara et al., 2007).

The sequence of cell divisions leading to tapetal formation is well understood in Arabidopsis (Canales et al., 2002; Feng and Dickinson, 2007; Scott et al., 2004), but how cells generated by the periclinal division of the inner secondary parietal (ISP) cell layer acquire and maintain their tapetal fate is unclear. The similarity in cell numbers between wild type and ems1 mutants (where no tapetum is formed) has reasonably been interpreted as evidence that these ISP cells revert to a default microsporocyte fate in the absence of the EMS1 receptor kinase (Zhao et al., 2002). However, our findings that a small number of cells with tapetal features (PTCs/LPCs) are formed early in development in wild-type and ems1 plants, and that ectopically expressed EMS1 stimulates proliferation of only these cells in the latter indicate that an alternative interpretation of these events might be required.

The new data point to the small number of cells formed adjacent to the sporogenous cells immediately becoming committed to a ‘pre-tapetal’ fate, and the EMS1 signalling pathway being responsible for the subsequent proliferation of these cells, allowing the tapetal monolayer to keep pace with the rapid expansion of the other layers of the loculus, which, because they are unaffected by the absence of functional EMS1, must be regulated by different signals. This role for EMS1 in proliferation implies that the EMS1 complex cannot be involved in the initial establishment of tapetal fate in the PTCs; the genes responsible for this are unknown, but other LRR-RLKs, such as BAM1/2, remain possible candidates.

Arguably, the LPCs of ems1 anthers could represent a small number of partially developed tapetal cells resulting from the incomplete penetration of the ems1 mutation, with the remainder of the inner division products of the inner secondary parietal cell being converted into microsporocytes. This is unlikely because: (1) EMS1 transcripts are absent from ems1 mutant plants; (2) reactivation of these few LPCs in ems1 lines can generate a complete and functional tapetum that supports the production of viable pollen; and (3) LPCs are present in all plants carrying ems1 alleles in both Ler and C24 backgrounds (some encoding non-functional receptor kinases), as well as in tpd1 and serk1 serk2 plants.

Although the EMS1 complex is required for the proliferation of PTCs, its role in their final differentiation into functional tapetal cells is unclear. LPCs certainly complete some aspects of tapetal development in ems1 lines as they express A7, a gene encoding a lipid transfer protein that is normally expressed post meiosis (Rubinelli et al., 1998), but they nevertheless fail to develop into functional tapetal cells. The EMS1 complex and its putative ligand TPD1 thus emerge as a component of a larger gene network regulating tapetal development, with other network members being responsible for specifying initial tapetal fate in the PTCs and, perhaps, aspects of subsequent development. Although an inverse relationship exists between microsporocyte and tapetal cell numbers in our transgenic plants, our data do not show whether this is the result of intercellular signalling, or simply the proliferation of reproductive cells to fill space that would otherwise be occupied by the tapetal monolayer.

Our experiments begin to explain how complex tissue arrays can develop within small plant organs in the absence of organised meristems. Like the microsporocytes, tapetal cells are specified very early, and once specified their fate cannot be altered. These tapetal ‘founders’ are then stimulated to proliferate (via the EMS1 complex) resulting in the formation of a monolayer investing the microsporocyte population. This signal for proliferation is unique to the tapetum, and differs from that regulating division of the microsporocytes and the other cell layers of the microsporangium. As TPD1 binds to the EMS1 complex (Yang et al., 2005), and is principally synthesised by the microsporocytes (Yang et al., 2003), its presence must serve as a developmental checkpoint confirming the presence of functional microsporocytes (Fig. 6). Maintenance of the developing monolayer requires its continued integrity, revealing that anticlinal divisions within it are either controlled by internal signals, or by microsporocyte-generated TPD1 acting as a directional cue. Cells of the tapetum and the microsporocytes thus develop as ‘communicating’ neighbours, but as independent lineages occupying defined domains within the expanding microsporangium. Whether the proliferation of microsporocytes into ‘tapetal territory’ in ems1 lines results from the lack of signals from the tapetum, or simply from available space, remains to be determined.

Perhaps the most significant difference between development of these complex microsporangial tissues and development of those derived from the apical and root meristems is their early and continued independence; certainly they can receive both directional and ‘checkpoint’ signals and perhaps alter the space they occupy but, once formed, their founder cells generate independently proliferating lineages committed to single developmental fates. Without this degree of independence and level of ‘internal’ control it is difficult to see how small numbers of cells of different fates can assemble into such a range of intricate patterns within a small, terminally differentiated organ such as the anther.

We thank the following for providing mutant lines and reagents: Hong Ma, De Ye, Sacco De Vries, and Rod Scott for providing the pA9::Barnase lines and information on A9 expression patterns. Carla Galinha and Paolo Piazza gave valuable help with in situ hybridisation and qRT-PCR, respectively, and we acknowledge Qing Zhang, Helen Prescott and Matthew Dicks for providing excellent technical assistance. We are indebted to Miltos Tsiantis and Angela Hay for helpful discussion, and the research was funded by Oxford University through a Clarendon Scholarship to X.F., with additional financial support from Magdalen College (Oxford).