Auxin guides germ-cell specification in Arabidopsis anthers

Significance

Germ cells (GCs) (i.e., the cells that are committed to meiosis and gametogenesis) are key carriers for eukaryotes to complete their life cycle, transmitting their genetic information from one generation to the next while generating variations to integrate environmental changes. Compared to what has been known in animals, very little is known about how the GCs in plants are segregated from somatic cells. This work demonstrates that auxin is a key factor guiding GC specification in Arabidopsis anthers. Local auxin biosynthesis interacts with the transcription of SPOROCYTELESS/NOZZLE and a progressive GC specification itself to form a dynamic feedback circuit that ensures the completion of GC specification.

Abstract

Germ cells (GCs) are the key carriers delivering genetic information from one generation to the next. In a majority of animals, GCs segregate from somatic cells during embryogenesis by forming germlines. In land plants, GCs segregate from somatic cells during postembryonic development. In a majority of angiosperms, male GCs (archesporial cells) initiate at the four corners of the anther primordia. Little is known about the mechanism underlying this initiation. Here, we discovered that the dynamic auxin distribution in developing anthers coincided with GC initiation. A centripetal auxin gradient gradually formed toward the four corners where GCs will initiate. Local auxin biosynthesis was necessary for this patterning and for GC specification. The GC determinant protein SPOROCYTELESS/NOZZLE (SPL/NZZ) mediated the effect of auxin on GC specification and modified auxin biosynthesis to maintain a centripetal auxin distribution. Our work reveals that auxin is a key factor guiding GC specification in Arabidopsis anthers. Moreover, we demonstrate that the GC segregation from somatic cells is not a simple switch on/off event but rather a complicated process that involves a dynamic feedback circuit among local auxin biosynthesis, transcription of SPL/NZZ, and a progressive GC specification. This finding sheds light on the mystery of how zygote-derived somatic cells diverge into GCs in plants.

Germ cells (GCs) are initiated early in development in animals (during embryogenesis in Ecdysozoa and Chordata) (1) and later in plants, where they arise from somatic cells in the adult (after flowering in angiosperms) (2, 3). While many achievements have been achieved in understanding the meiosis and postmeiotic germline specification during anther development (4), the mechanism for early GC initiation and specification before meiosis in plants is not well understood (3, 5–7). Early premeiotic anther development occurs in five stages as seen by changes to anther shape and cellular morphology (Fig. 1A) (8, 9). In Arabidopsis and most other angiosperms, male GC precursor cells are derived from L2 cells at the four corners of the anther primordia. The primordial cells at the corners divide further to form outer primary parietal cells (PCs) and inner initial GCs (also termed archesporial cells [ARs]). These initial GCs can be identified by their large cell size and big nuclei. By stage five, a concentric GC–PC pattern is fully established, GCs (also termed pollen mother cells [PMCs] at this stage) are prepared for meiosis, and the anther exhibits a four-lobed butterfly-like shape (Fig. 1A). Many genes influence early anther development and form a complex gene-regulatory network (2, 10–12). In this network, the key upstream gene SPOROCYTELESS/NOZZLE (SPLNZZ) is necessary and sufficient for GC initiation (13–15). However, the information about the positional signals guiding the GC initiation pattern is unclear.

Fig. 1.

Dynamic auxin distribution patterns in wild-type premeiotic anthers in Arabidopsis. (A) Schematic representation of microsporogenesis in early anther lobes. (Left) Cross-section of a four-lobed anther. (Right) The developmental stages of an anther lobe. Each cell lineage is marked by a specific color. L1, the outermost cell layer; L2-d, cells derived from the second layer; PPC, primary parietal cell; Ar, archesporial cell; SPC, secondary parietal cell; PSC, primary sporogenous cell; E, epidermis; En, endothecium; ML, middle layer; T, tapetum. (B) Immunolocalization of IAA in cross-sections of anther lobes from stage 1 to stage 5. IAA gradually forms a centripetal gradient. The GCs are marked with magenta stars. (C) The method used for imaging R2D2 signals in abaxial anther lobes. (D) Auxin signaling in anther lobes from stages 3 to 5 showing higher auxin levels (red) in GCs. The signal intensity calculated from mDII/DII is displayed as a false color scale. The original images are shown in the insets. L1 to L4 indicate the cell layers from outside to inside. The corresponding names of the cell layers are referenced in brackets. (E and F) Quantification of auxin signaling input (mDII/DII) (E) and nucleus size (F) in different cell layers of anther lobes from stages 3 to 5. Note the association between auxin levels and GC specification (characterized by large nucleus size). The number of samples per layer are as follows: L1 = 27, L2 = 30, L3 = 27, and L4 = 24 from four different stage-3 locules; L1 = 49, L2 = 48, L3 = 46, and L4 = 45 from six different stage-4 locules; and L1 = 37, L2 = 41, L3 = 34, L4 = 28, and L5 = 28 from four different stage-5 locules. The individual data points are colored and plotted on the boxplot. The box indicates the interquartile range (IQR), the whiskers show the range of values that are within 1.5 × IQR, and a horizontal line indicates the median. The notches represent the 95% CI for each median. (Scale bars, 10 µm.)

Auxin is a morphogen-like compound that plays a pivotal role in pattern formation in plant morphogenesis (16). As expected, auxin is also involved in anther development (17–20), including anther initiation (21), pollen maturation, anther dehiscence, and other aspects of postmeiotic anther development (18–20, 22). However, its function in microsporogenesis (i.e., GC initiation and specification during premeiotic anther development) is surprisingly unknown. Therefore, we investigated the role of auxin in guiding GC specification.

Results and Discussion

Auxin Distribution Parallels with GC Initiation in Premeiotic Anthers.

We first studied auxin distribution during early anther development using immunostaining and auxin biosensor. Using antibodies specific to the major form of auxin, indole-3-acetic acid (IAA), we labeled native auxin in early anthers from stage 1 to stage 5 (Fig. 1 A and B and SI Appendix, Fig. S1). In stage-1 anthers, auxin was homogeneously distributed in all cells. In stage 2, auxin spread into lateral domains. Starting from stage 3, auxin gradually congregated in central GC domains, accompanying the differentiation of the cell layers; stage-5 anthers displayed a clear difference in auxin levels between GCs and PCs (Fig. 1B and SI Appendix, Fig. S1).

R2D2 is a ratiometric auxin biosensor that is used to indicate the relative in vivo auxin concentration at the cellular level (23, 24). We mapped R2D2 signals in abaxial anther lobes (Fig. 1 C–F). The longitudinal view of anther lobes yielded a fine map of auxin distribution patterns from stage 3 to stage 5: in stage 3, the auxin gradient was steep, with a minimum in the outer cells and a maximum in the inner cells; from stage 4 to stage 5, auxin was concentrated in the cells of the GC domains (Fig. 1 D and E), which are inherently characterized by their large nuclei (Fig. 1F).

TAA1 and TAR2 Are Involved in Microsporogenesis.

The auxin signals detected through immunolocalization and biosensors are the combined output of auxin transport (25), metabolism [including biosynthesis (26), modification (27), and degradation (28)], and complex signaling pathways that stimulate downstream events (29). To pinpoint the particular components responsible for the centripetal auxin distribution and its potential effects on GC specification, we conducted RNA sequencing (RNA-seq) and genetic manipulation experiments. RNA-seq analysis revealed a variable expression pattern for auxin-related genes in stage 3 and stage 4 anthers, but no clear hints pinpoint whether auxin polar transport, metabolism, or signaling pathways potentially play a key role in early anther development (SI Appendix, Fig. S2). We then created transgenic plants that interfered with a range of auxin-related functions in early anther development (SI Appendix, Fig. S3) and used Alexander staining to evaluate GC formation (6, 30). Phenotyping of these plants showed that altered auxin synthesis caused severe pollen defects (SI Appendix, Fig. S3). These genetic data suggested that the regulations in the auxin biosynthesis pathway may be important for microsporogenesis. Therefore, we examined the role of auxin biosynthesis in GC formation.

The L-Trp–dependent indole-3-pyruvate biosynthesis pathway is the main route for IAA production (26). This two-step pathway includes enzymes encoded by genes in the TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS 1 (TAA1) and YUCCA (YUC) families. YUC2 and YUC6 are well-studied genes important for anther development (18, 31). In yuc2 yuc6, the anthers are sterile, but the GC initiation and specification are normal (SI Appendix, Fig. S4). This may indicate the redundant function of YUC family genes in premeiotic anther development. Because the TAA1 family works upstream of the YUC family (32), and TAA1 and TRYPTOPHAN AMINOTRANSFERASE RELATED 2 (TAR2) are expressed in early anthers in our RNA-seq data (SI Appendix, Fig. S2), we then focused on TAA1 and its close homolog TAR2 in our investigation of the role of auxin biosynthesis in GC formation. TAA1 and TAR2 were initially expressed throughout anther primordia but became concentrated in GC domains by stage 5 (SI Appendix, Fig. S5). As reported, the flower development of taa1 tar2-1 [a strong allele of taa1 tar2 (33)] was severely affected, and the anther numbers decreased dramatically (SI Appendix, Fig. S6). In taa1 tar2-2 [a weak allele of the taa1 tar2 mutant (33)], where the anther initiation was less affected (SI Appendix, Fig. S6), around 40% of anthers formed pollen in two locules, ∼20% harbored pollen in only one locule, and the remaining 40% were completely sterile (SI Appendix, Fig. S7A). By contrast, the anthers of taa1 tar2-1 were completely devoid of pollen (SI Appendix, Fig. S7A). Thus, TAA1 and TAR2 affect microspore formation.

Auxin Distribution and GC Specification Are Disturbed in taa1tar2

To investigate the role of auxin biosynthesis in centripetal auxin patterning and GC differentiation, we crossed R2D2 with taa1/+ tar2. In the F₃ generation, as expected, auxin production was significantly reduced in taa1 tar2-2 and taa1 tar2-1 anthers (Fig. 2 A and B), and auxin-responsive genes were broadly down-regulated (SI Appendix, Fig. S7B). Even though a weak centripetal auxin gradient persisted in taa1 tar2-2 abaxial anther lobes, this gradient was lost in taa1 tar2-1 (Fig. 2 A and B). Consistent with the perturbation of the auxin distribution pattern, the nucleus size also changed in taa1 tar2 mutants. In wild-type stage 3/4 anthers, the nuclei of GCs located in L4 were bigger than those of somatic cells (Fig. 1F), similar to those in taa1/+ tar2-2 (Fig. 2C and SI Appendix, Fig. S7A). However, in taa1 tar2-2, the L4 nuclei were significantly smaller (Fig. 2C), and taa1 tar2-1 anthers had smaller nuclei in L4 and L3 cells (Fig. 2C).

Fig. 2.

TAA1/TAR2 is indispensable for GC specification. (A) Ratiometric image of R2D2 shows auxin signaling input in taa1 tar2 anther lobes at stage 3/4. The signal intensity calculated from mDII/DII is displayed as a false color scale. The original images are shown in the insets. Compared with taa1/+ tar2-2, the auxin level is lower in taa1 tar2-2 and extremely low in taa1 tar2-1. (B and C) Quantification of auxin signaling input (B) and nucleus size (C) in stage-3/4 taa1 tar2 anther abaxial lobes. As the auxin level decreases in the taa1 tar2 mutants, the nucleus becomes smaller in the inner cells (L3 and L4), implying the loss of germinal cell fate. The number of samples per layer was as follows: L1 = 36, L2 = 32, L3 = 30, and L4 = 28 from four control (taa1/+ tar2-2) anther lobes; L1 = 46, L2 = 47, L3 = 37, and L4 = 49 from five taa1 tar2-2 anther lobes; and L1 = 47, L2 = 35, L3 = 30, and L4 = 28 from four taa1 tar2-1 anther lobes. n.s., no significant difference; *P < 0.05; **P < 0.001 by Student’s t test. (D–G) Semithin section images showing the transverse sections of wild-type (Col) (D), taa1 tar2-2 (E and F), and taa1 tar2-1 (G) anthers at stage 5. The ad- and abaxial sides of the anthers were marked as Ad and Ab, respectively. (H–K) Electron microscope images showing the detailed phenotype of wild-type (Col), adaxial (Ad) (H), taa1 tar2-2 abaxial (Ab) (I and J), and taa1 tar2-2 Ad locules (K). In H, the normal PMC is marked with a black arrow, and different somatic cell layers are indicated along the dashed line. The impaired PMC (hollow arrow) and somatic cell layer (dashed lines) are marked in I–K. Note that most cells in the taa1 tar2-2 adaxial locules are vacuolated (K). (L) Quantification of PMC formation in cross-sections of taa1 tar2 anthers. (Scale bars, 10 µm.)

In examination of the differentiation status of GC cells using histological analysis in stage-5 taa1 tar2 anthers (Fig. 2 D–K), we found that in some of the abaxial locules of taa1 tar2-2 anthers, GC-like cells were visible, but the cell number and size were reduced (Fig. 2 E and I). In other abaxial locules, cells were mostly vacuolated, which complicated the observation of GCs or well-differentiated PCs (Fig. 2 F and J). All cells in the adaxial locules were vacuolated, so we were unable to identify GCs or well-organized PC layers (Fig. 2 E, F, and K). We quantified these phenotypes: 48.8% of taa1 tar2-2 anthers failed to form GCs or well-differentiated PCs, and 51.2% of the anthers only showed GCs at abaxial locules (Fig. 2L and SI Appendix, Fig. S7C). The strong allele taa1 tar2-1 contained only locules with vacuolated cells, which caused sterility (Fig. 2 G and L and SI Appendix, Fig. S7C). Consistent with our histological analysis, the expression of PMCs and tapetum-specific genes was dramatically repressed in taa1 tar2-2 (SI Appendix, Fig. S8). Thus, we concluded that TAA1 and TAR2 are required for dynamic auxin distribution during early anther development and are indispensable for GC specification.

TAA1 and TAR2 Trigger GC Specification by Activating SPL/NZZ.

How is auxin involved in GC specification? In Arabidopsis, SPL/NZZ is the determinant for early GC differentiation (13–15), and the SPL/NZZ expression pattern overlaps with the distribution of auxin (SI Appendix, Fig. S9A) (6). In spl/nzz anthers, ARs and their adjacent cells stop differentiating at stage 3 (14, 15). In stage 4, cells become vacuolated, hindering the formation of GCs and well-organized PCs. To test if auxin affects GC specification by regulating SPL/NZZ expression, we probed SPLNZZ for transcripts in taa1 tar2-2 anthers by in situ hybridization. A decrease of SPL/NZZ transcripts was found in early anthers from stage 1 onwards (SI Appendix, Fig. S9A). In 50% of anthers, SPLNZZ transcripts were diminished or absent in the adaxial lobes and completely absent in the rest at stage 5 (Fig. 3A). We quantified SPLNZZ transcripts using quantitative polymerase chain reaction (qPCR) in taa1 tar2 flowers. taa1 tar2-2 contained 70% fewer SPLNZZ transcripts, and taa1 tar2-1 contained almost no SPLNZZ transcripts when compared with wild-type (Fig. 3B). To test whether auxin mediates SPL/NZZ expression at the transcriptional or at the posttranscriptional level, we examined SPL/NZZ promoter activity in taa1 tar2 by crossing pSPL:GUS with taa1/+ tar2. In the F₃ progeny, β-glucuronidase (GUS) signals became weaker in taa1 tar2-2 and were absent in taa1 tar2-1 anthers (Fig. 3C and SI Appendix, Fig. S9B). Thus, SPL/NZZ transcriptional activity is reduced in taa1 tar2 anthers.

Fig. 3.

TAA1 and TAR2 activate SPL/NZZ transcription. (A) SPL/NZZ transcripts in cross-sections of stage-4 Col and taa1 tar2-2 anthers, revealed by in situ hybridization. The SPLNZZ signals are indicated by arrowheads. (B) Quantification of SPL/NZZ transcripts in Col and taa1 tar2 mutants by qPCR. Bars indicate the mean, and the error bars indicate SE (n = 3 technical repeats). Three independent experiments yielded similar results. (C) The SPL promoter activity shown by pSPL:GUS signals in Col and taa1 tar2. pSPL activity is reduced in taa1 tar2-2 and absent in taa1 tar2-1 anthers (arrowheads). (D) The fertility of taa1 tar2-2 is partially restored by pREM22:SPL. Arrowheads (magenta) indicate fertile long siliques. (E) Quantification of anther fertility in Col, taa1 tar2-2, and taa1 tar2-2 pREM22:SPL by Alexander staining assay. (F) Quantification of PMC formation in cross-sections of taa1 tar2-2 pREM22:SPL anthers. (Scale bars: 20 µm in A, 1 mm in C, and 1 cm in D.)

To test if auxin affects GC specification by regulating SPL/NZZ transcription, we ectopically expressed SPL/NZZ under another AR-active promoter, pREM22 (34). If this transgene complemented the taa1 tar2 phenotype, that would indicate SPL/NZZ is downstream of auxin in mediating GC specification. Considering the anther initiation was severely inhibited in taa1 tar2-1 (SI Appendix, Fig. S6), which may increase the complexity for interpretation, we chose to ectopically express SPL/NZZ driven by pREM22 in taa1 tar2-2. As expected, in at least four independent taa1 tar2-2 pREM22:SPL-Myc T2 lines, the fertility of taa1 tar2-2 was partially restored (Fig. 3 D and E) (34). In a taa1 tar2-2 pREM22:SPL-Myc plant, GCs appeared in adaxial locules of 18.2% of the anthers, which were fully sterile in taa1 tar2-2 (Figs. 3F and 2L). The proportion of abaxial locules with GCs increased from 51.2% in taa1 tar2-2 (Fig. 2L) to 81.8% in taa1 tar2-2 pREM22:SPL-Myc (Fig. 3F). These data demonstrated that SPL/NZZ functioned downstream of auxin to mediate GC specification. Similar to the previous finding that overexpression of SPL/NZZ could suppress YUC genes in different organs (35), we found an elevation of TAA1 and TAR2 transcripts in spl flowers (SI Appendix, Fig. S10). Together with the abolishment of a centripetal auxin gradient in spl mutant anthers (Fig. 4 A–C and SI Appendix, Fig. S11), we conclude that SPL/NZZ also has a feedback-regulatory effect on the auxin biosynthesis pathway to maintain auxin homeostasis and distribution in premeiotic anthers.

Fig. 4.

Auxin–SPL feedback loop for GC specification. (A) Ratiometric image of R2D2 showing an increase of auxin in the outer cell layers of spl stage-3/4 anther lobes. The signal intensity calculated from mDII/DII is displayed as a false color scale. The original images are shown in the insets. (B and C) Quantification of auxin signaling input (B) and nucleus size (C) in stage-3/4 spl+(control) and spl anther lobes. The auxin level was significantly increased in L1 to L3 of spl anther lobes (B), while the nucleus size was significantly reduced in spl anthers, corresponding to the loss of GCs (C). The number of samples per layer was as follows: L1 = 38, L2 = 27, L3 = 28, and L4 = 29 from four control (spl+) anther lobes and L1 = 68, L2 = 56, L3 = 42, and L4 = 45 from six spl anther lobes. n.s., no significant difference; *P < 0.05; **P < 0.001 by Student’s t test. (D) Electron microscope images showing the transverse sections of a stage-5 pSPL:iaaM anther. All cells in the locules were vacuolated (arrowhead), and no GCs or somatic layers were distinguishable. (E) Quantification of PMC formation in pSPL:iaaM anthers. (F) qPCR showing SPL transcript levels in flowers from three different T₅ pSPL:iaaM alleles and in controls. Numbers indicate alleles. Bars indicates the mean, and the error bars indicate SE (n = 3 technical repeats). Three independent experiments yielded similar results. (G) IAA immunolocalization in Col and pSPL:iaaM anthers at stages 4 and 5. IAA is distributed in the outer layers in pSPL:iaaM, similar to A and B. (H) Schematic representation of IAA distribution shown in G, frame zones. (I) Ratiometric image of R2D2 in pSPL:iaaM stage-3/4 anther lobes. The original images are shown in the insets. (J and K) Quantification of auxin signaling input (J) and nucleus size (K). The auxin distribution pattern was perturbed in anther lobes and similar to the pattern in spl (J), and the nucleus size was reduced in L4, corresponding to the loss of GCs (K). The number of samples per layer was as follows: L1 = 52, L2 = 34, L3 = 35, and L4 = 30 from seven anther lobes. (Scale bars, 10 µm.)

SPL/NZZ Mediates the Effect of Auxin on GC Specification.

Auxin deficiency results in a defect in GC specification; excess auxin also affects GC formation (SI Appendix, Figs. S3 and S12 A and B). We quantified the abnormality of anther development in four independent T₂ lines of pSPL:iaaM. Around 20 to 60% of anthers in these transgenic plants were completely sterile, and only 3.65% of the anthers were normal. These phenotypes were further confirmed by analyzing at least three independent T₁ pREM22: iaaM lines (SI Appendix, Fig. S12A). At the cellular level, cells that normally differentiate into GCs in wild type did not differentiate but became vacuolated (Fig. 4D). The loss of GC cell fate is further confirmed by the in situ hybridization results showing the disappearance of SPL/NZZ transcript in early anther locules (SI Appendix, Fig. S13). This deficiency occurred in 62.6% of abaxial anther lobes, and the remaining 37.4% of anthers were completely sterile (Fig. 4 D and E and SI Appendix, Fig. S12C). In pSPL:iaaM transgenic lines, overall SPL/NZZ gene expression increased (Fig. 4F), consistent with up-regulation by auxin (Fig. 3 A–C). However, the centripetal auxin distribution pattern was severely disrupted (Fig. 4 G–K). This abnormal auxin distribution pattern is similar to what we observed in the spl mutant (Fig. 4 A–C and SI Appendix, Fig. S11). Given that SPL/NZZ negatively regulates auxin biosynthesis to maintain auxin homeostasis and patterning (Fig. 4 and SI Appendix, Fig. S10), our observations in pSPL:iaaM anthers underline the importance of auxin control in GC specification.

In most higher animals, the GCs differentiate in early embryogenesis to form a germline carried by the soma. In contrast, in most land plants, GCs that are committed into meiosis are segregated from somatic cells during late development. In many angiosperms, such as Arabidopsis, male GCs are initiated at the four corners of anther primordia, but little was known about the positional information that induces GC initiation. Our data demonstrate that auxin plays a key role; in particular, auxin biosynthesis is indispensable. Through spatiotemporal interactions with the GC-determining SPL/NZZ, a centripetal auxin gradient forms, and GCs are specified in the inner layers of anther lobes. When the auxin gradient is abolished, GC specification halts. This is reminiscent of research showing the importance of auxin biosynthesis in patterning of female gametophytes (36). Moreover, the dynamic interactions between auxin and SPL/NZZ we find here provide a context for the spatiotemporal model of GC initiation and specification. This leads to several intriguing questions such as how the homogenous auxin distribution in anther primordia is gradually concentrated at the four corners during anther development. Polar auxin transport is possibly involved (17). SPL/NZZ is proposed to control auxin distribution by interacting with PIN1 during female gametogenesis (37). It will be interesting to test whether a similar situation exists in early anthers. Other questions include the following: What is the detailed mechanism of auxin–SPL feedback, and how do GCs differentiate and induce centripetal auxin distribution? Answering these questions will shed further light on how GCs are induced and specified from somatic cells during plant development. In addition, this knowledge may improve our understanding of how plants and animals evolved such dramatically different routes for GC induction and differentiation.

Materials and Methods

Plant Material and Growth Conditions.

The Arabidopsis reporter lines [R2D2 (23), pSPL::GUS (6), and pSPL:SPL-myc (6)] and mutants [spl (14), taa1 tar2-1(alias wei8-1 tar2-1) (33), and taa1 tar2-2 (alias wei8-1 tar2-2) (33)] have been described previously. The seeds were sterilized, placed on Murashige and Skoog (MS) medium for germination, and cultured in vitro. After 2 wk, the seedlings were transplanted to soil and grown under long-day conditions (16-h light/8-h dark; light bulb, Philips 28 W 840 neon, 4,000 K, 103 lm/W) at 22 °C.

Plasmid Construction and Plant Transformation.

To overexpress auxin-related genes in anthers, the full-length complementary DNA (cDNA) was cloned and fused into an SPL promoter cassette (6). For generating RNA interference lines, artificial microRNAs were designed following the instructions on the website (wmd3.weigelworld.org/cgi-bin/webapp.cgi); these were cloned into the SPL promoter cassette (6). These constructs were transformed into Col-0 using the floral dip method (38). T₁ plants were screened on MS medium containing glufosinate-ammonium; resistant plants were selected for further analysis.

The REM22 promoter was chosen based on its activity in ARs (34). A 1,022-bp region upstream of the start codon was cloned into the pEGAD and pCambia1305.1 target vectors to get REM22 promoter cassettes; these were named pEGAD-REM22 and pCambia1305.1-REM22, respectively. To obtain the pREM22:iaaM construct, the iaaM coding sequence was cloned into pEGAD-REM22. It was then transformed into Col plants, and transgenics were screened using glufosinate-ammonium in vitro.

The pREM22:SPL-Myc construct was generated by fusing the SPL-Myc coding sequence, which was cloned from the pSPL:SPL-Myc construct (6) into the pCambia1305.1-REM22 vector. pREM22:SPL-Myc was then transformed into taa1/+ tar2-2 plants. pREM22:SPL-Myc transgenics were screened using hygromycin B and identified by PCR in T₂ progeny. The primers used for cloning are listed in SI Appendix, Table S1

IAA Immunofluorescence Localization Assay.

To cross-link the IAA to endogenous structural proteins, inflorescences were excised and immediately soaked in freshly prepared ice-cold 3% (wt/vol) 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDAC, Sigma Chemical) with 0.05% (vol/vol) Triton X-100, vacuum infiltrated for 1 h, and incubated in darkness at 4 °C for 1 h. The inflorescences were rinsed three times in phosphate buffer (0.2 M, pH 7.4) for 10 min each, transferred to fixative (4% (wt/vol) paraformaldehyde in 0.2 M phosphate buffer and 0.1% (vol/vol) Triton X-100, vacuum infiltrated for 1 h, and incubated overnight at 4 °C. Paraffin sectioning was performed as described previously (39) with a minor modification. The 7-μm sections were spread on poly-Lys–coated slides. After dewaxing, the slides were incubated in fixative again and washed twice with washing buffer (0.2 M phosphate buffer, 0.1% [vol/vol] Tween 20) for 10 min each. For blocking, the sections were soaked in 10 mM phosphate buffer containing 3% (wt/vol) bovine serum albumin (BSA) blocking solution for 1 h (at room temperature) or overnight (at 4 °C). The anti-IAA antibodies (Phytodetek, catalog No. PDM 09346/0096; diluted 1:150 in blocking solution) were added and the sections incubated in a humid chamber for 3 to 4 h (at room temperature, 25 °C) or overnight (at 4 °C) in darkness. After incubation, the samples were washed vigorously twice or thrice in 10 mM phosphate buffer containing 2.9% (wt/vol) NaCl, 0.1% (vol/vol) Tween 20, and 0.1% (vol/vol) BSA for 10 min at a time and then washed with 10 mM phosphate-buffered saline (PBS), 0.88% (wt/vol) NaCl, 0.1% (vol/vol) Tween 20, and 0.8% (wt/vol) BSA for 10 min. Anti-mouse Fluor 488 (affinity anti-IAA antibody) secondary antibodies were diluted 1:500 in blocking solution and incubated for 4 h at room temperature. Two washes (15 min each) with 10 mM PBS, 0.88% (wt/vol) NaCl, 0.1% (vol/vol) Tween 20, and 0.8% (wt/vol) BSA were followed by a 1-min rinse in 10 mM PBS. Images were photographed using a Zeiss microscope (Axio Imager D2), processed by ZEN lite 2011 (blue edition; Carl Zeiss), and edited with Photoshop CS6 (Adobe Systems). At least three biological replicates were performed, and similar results were obtained in all experiments.

Confocal Microscopy.

Flowers between stages 7 and 9 were dissected, and the sepals were removed. The anthers (at stages 3 to 5) were excised from the meristem and quickly placed on a slide with double-sided tape (3M) to hold the abaxial locules upward. A drop of water was placed on the anthers, and they were imaged using an LSM 710 confocal microscope (Carl Zeiss) equipped with a water immersion objective (W Plan-Apochromat 20×/1.0 DIC VIS-IR). Fluorescence images were analyzed using ZEN lite 2011 software (black edition, Carl Zeiss). taa1 tar2-2 and spl were sterile and in different genetic backgrounds than R2D2. To compare the R2D2 signal in different genetic backgrounds, we chose taa1/+ tar2-2 and spl/+ (which are as fertile as wild type; see SI Appendix, Supplementary Information) as the controls for taa1 tar2 and spl, respectively.

To quantify the R2D2 signals, contours of nuclei were manually selected in the mDII channel using the elliptical selection tool in Fiji freeware, and regions of interest (ROIs) were added to the ROI manager. The area of the nuclei and the mean gray values of different ROIs in mDII and DII channels were measured. The mDII/DII ratio was calculated in Microsoft Excel. The results were plotted using the PlotsOfData web tool (40).

To generate ratiometric images of R2D2, ratios between signal intensities of each pixel from the mDII and DII channels were calculated using Fiji; signal intensities in both channels below 10 to 60 (based on the average signal intensities between the nuclei in the mDII channel) were set to 0 in ratio images to subtract the background.

RNA Sequencing.

Stage-3 and stage-4 stamens were dissected and pooled separately for each replicate. RNA was extracted using RNeasy Plant Mini Kit (Qiagen, Cat NO. 74904) according to the manufacturer’s instructions. Libraries were constructed using the TruSeq Stranded Total RNA Sample Prep Kit (Illumina). The RNA-seq libraries were sequenced using a HiSeq2000 Pair End 2 × 100 bp at the High-throughput Sequencing Center in the Biomedical Pioneering Innovation Center, Peking University. The original image data generated by the sequencing machine were converted into sequence data via base calling (Illumina pipeline CASAVA version 1.8.2) and then subjected to standard quality control (QC) criteria to remove all of the reads that fit any of the following parameters: 1) reads that aligned to adaptors or primers with no more than two mismatches, 2) reads with >10% unknown bases, and 3) reads with >50% of low-quality bases (quality value #5) in one read. Finally, 1,308.3 Gb (94.4%) of filtered reads were left for further analysis after QC, and reads mapped to ribosomal RNA were discarded. After that, the remaining reads were mapped to the TIR9 reference genome using Bowtie 2 and TopHap.

Alexander Staining Assay.

To stain pollen, flowers were dissected from the inflorescence, opened using tweezers, and soaked in Alexander staining solution (30) for at least 10 h. The stamens were dissected from the flowers, placed on slides, and sealed with chloral hydrate solution (4 g chloral hydrate, 1 mL glycerol, and 2 mL deionized water). To compare the different genotypes, flowers were dissected at the same developmental stage.

In Situ Hybridization.

In situ hybridization was performed according to a previously described protocol (41). Images were taken using a Zeiss microscope (Axio Imager D2). Primers used for preparing the probes targeting TAA1, TAR2, and SPL transcripts are listed in SI Appendix, Table S1

Histological Analysis.

Cross-sections of anthers were obtained by paraffin and semithin sectioning. The procedures were conducted as described previously (39, 42). Inflorescences were collected at the same developmental stage in order to compare the phenotypes between different lines. Anthers at stages 4 to 11 were chosen to quantify locule sterility. Quantitative analysis was carried out using Microsoft Excel.

For electron microscopy, flowers from stages 7 to 9 were selected from the inflorescences and soaked in 4% (wt/vol) paraformaldehyde and 2.5% (wt/vol) glutaraldehyde (Sigma, G5882). After vacuum infiltration for 1 h, the samples were incubated at 4 °C overnight. They were postfixed in 2% (wt/vol) OsO₄ (Ted Pella, 18451) in phosphate buffer (0.1 M, pH 7.4) at room temperature for 90 min. The staining buffer was replaced with 2.5% (wt/vol) ferrocyanide (Sigma, 234125) in phosphate buffer (0.1 M, pH 7.4) at room temperature for 90 min. After being washed three times in 0.1 M phosphate buffer, the samples were incubated with filtered thiocarbohydrazide (Sigma, 223220) at 40 °C for 45 min. Then the samples were fixed in unbuffered 2% OsO₄ for 90 min followed by incubation in 1% (wt/vol) uranyl acetate (Zhongjingkeyi) aqueous solution at 4 °C overnight. Then they were incubated in a lead aspartate solution (0.033 g lead nitrate [Sigma, 228621] in 5 mL 0.03 M aspartic acid [Sigma, 11189, pH 5.0]) at 50 °C for 120 min and dehydrated through a graded ethanol series (50%, 70%, 80%, 90%, and 100% ethanol, 10 min each) and pure acetone. The samples were embedded in Epon-812 resin (Structure Probe, Inc., 02660-AB). Ultrathin sections (50 nm) were cut with a diamond knife (Diatome, MC16425) and imaged using a scanning electron microscope (Zeiss Gemini 300) with a resolution of 3 nm/pixel and dwell time of 2 to 5 µs.

Quantitative Real-Time PCR.

Total RNA was extracted from Arabidopsis inflorescences using the E.Z.N.A Plant RNA kit (OMEGA, R6827-01). RNA samples were digested using RQ1 RNase-free DNase (Promega, M6101). For first-strand cDNA synthesis, 1 μg RNA was used with ReverTra Ace qPCR RT Kit (TOYOBO, Code No. FSQ-101). The SYBR Premix Ex Taq (Takara, RR420A) was used to carry out quantitative real-time PCR using the Applied Biosystems 7500 real-time PCR system. GAPDH was used as the internal reference. The sequences of primers used are listed in SI Appendix, Table S1

GUS Staining Assay.

For GUS staining, whole pSPL:GUS inflorescences were soaked in X-Gluc solution (100 mM sodium phosphate buffer, 10 mM EDTA, 0.5 mM K3Fe(CN)₆, 3 mM K₄Fe(CN)₆, 0.1% [vol/vol] Triton X-100) containing 1 mg/mL GUS substrate X-gluc (5-bromo-4-chloro-3-indolylglucuronide) and vacuum infiltrated for 1 h in darkness. The inflorescences were incubated at 37 °C in darkness for 16 h. After staining, the samples were dehydrated in a graded ethanol series (70%, 85%, 95%, and 100% twice [vol/vol]) until the chlorophyll was completely removed. Flowers were then dissected and placed on the slide, sealed with chloral hydrate solution, and photographed using a Zeiss microscope (Axio Imager D2). Images were processed using ZEN lite 2011 (blue edition; Carl Zeiss).

Data Availability

The RNA-seq data have been deposited at Beijing Institute of Genomics Data Center (https://bigd.big.ac.cn/?lang=en) (BioProject: PRJCA003607).