miR156-Targeted and Nontargeted SBP-Box Transcription Factors Act in Concert to Secure Male Fertility in Arabidopsis

The SBP-box transcription factor SQUAMOSA PROMOTER BINDING PROTEIN-LIKE8 ( SPL8 ) is required for proper development of sporogenic tissues in Arabidopsis thaliana . Here, we show that the semisterile phenotype of SPL8 loss-of-function mutants is due to partial functional redundancy with several other members of the Arabidopsis SPL gene family. In contrast with SPL8 , the transcripts of these latter SPL genes are all targeted by miR156/7. Whereas the introduction of single miR156/7-resistant SPL transgenes could only partially restore spl8 mutant fertility, constitutive overexpression of miR156 in an spl8 mutant background resulted in fully sterile plants. Histological analysis of the anthers of such sterile plants revealed an almost complete absence of sporogenous and anther wall tissue differentiation, a phenotype similar to that reported for sporocyteless/nozzle ( spl/nzz ) mutant anthers. Expression studies indicated a functional requirement for miR156/7-targeted SPL genes limited to early anther development. Accordingly, several miR156/7-encoding loci were found expressed in anther tissues at later stages of development. We conclude that fully fertile Arabidopsis ﬂowers require the action of multiple miR156/7-targeted SPL genes in concert with SPL8 . Either together with SPL/NZZ or independently, these SPL genes act to regulate genes mediating cell division, differentiation, and speciﬁcation early in anther development. Furthermore, SPL8 in particular may be required to secure fertility of the very ﬁrst ﬂowers when ﬂoral transition-related miR156/7 levels might not have sufﬁciently declined.


INTRODUCTION
Flowering plants alternate during their life cycle between diploid sporophytic and haploid gametophytic generations. Initial development of the sporophyte arising from the zygote is generally dominated by vegetative growth. Sexual reproduction requires a developmental phase transition resulting in the formation of flowers with highly specialized organs (i.e., anther-bearing stamen and ovule-bearing carpels). Within these organs, cells are recruited to undergo meiotic divisions to form male and female gametophytes. In Arabidopsis thaliana, the canonical homeotic ABC genes APETALA3, PISTILLATA, and AGAMOUS (AG) determine stamen identity of the third whorl floral organs. The latter is a direct activator of SPOROCYTELESS/NOZZLE (SPL/NZZ) more recently found to play a role in determining stamen identity also (Ito et al., 2004;Liu et al., 2009). After stamen identity is fixed, subsequent anther development can be divided into 14 stages according to defined histological characteristics (Sanders et al., 1999). Stages 1 to 7 are predominantly confined to sporogenous cell and anther wall formation in the four corners of the anther. The postmeiotic stages 8 to 14 mainly consist of pollen maturation and anther structural adaptations for their release at anthesis.
Much research on early anther development has focused on how sporogenous and parietal cells are derived and how they interact in a complex process involving cell proliferation, specification, and differentiation. However, whereas comparative global expression analysis of the wild type and mutants uncovered many genes acting specifically in early anther development (Zik and Irish, 2003;Hennig et al., 2004;Wellmer et al., 2004;Lu et al., 2006;Alves-Ferreira et al., 2007;Ma et al., 2007Ma et al., , 2008Wijeratne et al., 2007), only a few are functionally characterized. Among the earliest acting genes are those that relate to more general aspects of organ growth like regulating polarity and cell proliferation, such as the related transcription factors JAGGED and NUBBIN (Dinneny et al., 2006;reviewed in Husbands et al., 2009). Their activity is followed by the determination of cell fate within anthers, but little is known about genes that participate in the specification of the archesporial cells. However, obviously required for early microsporangium development are SPL/NZZ and ROXY1 and -2, as their respective mutants do not produce microsporocytes and anther walls (Schiefthaler et al., 1999;Yang et al., 1999;Xing and Zachgo, 2008). Subsequent cell-type specification and differentiation is mediated by BARELY ANY MERISTEM1 (BAM1) and BAM2, EXCESS MICROSPOROCYTES1/EXTRA SPOROGENOUS CELLS (EMS1/EXS), SOMATIC EMBRYOGENESIS RECEPTOR-LIKE KINASE1 (SERK1) and SERK2, and TAPETUM DETER-MINANT1 (TPD1) Zhao et al., 2002;Yang et al., 2003;Albrecht et al., 2005;Colcombet et al., 2005;Hord et al., 2006;Mizuno et al., 2007;Jia et al., 2008;Zhao, 2009). Their mutants all produce more micromeiocytes at the expense of tapetum or other parietal cell layers due to ectopic expression of SPL/NZZ. In normal development, SPL/NZZ promotes expression of BAM1 and BAM2 such that in a negative feedback, BAM1 and BAM2 restricts SPL/ NZZ to the center of the locule (Hord et al., 2006). Whereas BAM1, BAM2, EMS1/EXS, SERK1, SERK2, and TPD1 all encode receptor/ligand components, SPL/NZZ remains the only transcription factor studied in detail at these early anther stages (Ma, 2005).
In previous work, we showed that in addition to SPL/NZZ, the unrelated SBP-box transcription factor SPL8 influences anther development in Arabidopsis (Cardon et al., 1999;Unte et al., 2003;Zhang et al., 2007). Histological analysis of spl8 loss-of-function mutants revealed abnormal pollen sac formation, resulting in reduced pollen production per anther and, as a consequence, a reduced fertility. Moreover, severity of the phenotype declined acropetally in the main inflorescence.
As loss-of-function of SPL8 results in a semisterile phenotype that declines in severity in an acropetal fashion, we investigated if a compensatory mechanism could rely on functional redundancy with other Arabidopsis SBP-box genes. Here, we present data that miR156/7 targeted SPL genes together with the nontargeted SPL8 are absolutely required for male fertility in Arabidopsis. Either together with or independently of SPL/NZZ, these SBPbox transcription factors promote sporogenous cell and parietal cell formation in early anthers.

Loss-of-Function of miR156/7-Targeted SBP-Box Genes Enhances the spl8 Semisterile Phenotype
Based on similarities in both SBP domain sequence and genomic structure, the 17 Arabidopsis SPL genes are easily divided into two groups. The first group is represented by the more complexly organized genes SPL1, -7, -12, -14, and -16, which consist of 10 or more exons and encode proteins of over 800 residues. The second group represents the remaining 12 SPL genes, with two to four exons and coding for proteins of less than half the size of those within the first group ( Figure 1A; Riese et al., 2008). Moreover, with the exception of SPL8, all members of the second group carry a miR156/7 response element and are thus subjected to posttranscriptional regulation by miR156/7 ( Figure  1A). It should be noted, however, that SPL8 and other members of this group share very limited sequence similarity outside the SBP domain. In fact, additional similarity could be detected only weakly in the N-terminal region of SPL2, -10, -11, -9, and -15 and possibly also of SPL13. This area upstream of the SBP domain encompasses a motif with unknown function previously described to be evolutionary conserved for orthologous SPL8-like proteins (Unte et al., 2003;Riese et al., 2007).
To test the possibility that other SBP-box genes contribute redundantly to sustain fertility in spl8, we first identified the triple mutants spl8 spl9 spl15, spl8 spl2 spl9, and spl8 spl2 spl15 and the quadruple mutant spl8 spl2 spl9 spl15 from an F2 population generated after a cross between spl8-1 and the triple mutant spl2-1 spl9-1 spl15-1 (Schwarz et al., 2008), a loss-of-function mutant for the SPL2, SPL9, and SPL15 genes, which are posttranscriptional regulated by miR156 and probably by miR157 as well ( Figure 1A). When compared with the spl8-1 single mutant, the identified multiple mutants were found to form, as expected, fewer pollen sacs and less pollen in the anthers of later arising flowers (Figures 1B and 1C). Hence, seed production of these plants dramatically decreased ( Figure 1D). Pollen production and seed set in the triple mutant spl2-1 spl9-1 spl15-1, however, was only slightly reduced ( Figures 1C and 1D). These data provided evidence that the miR156/7-targeted SPL genes do affect fertility. That they regulate anther development, most probably redundantly with SPL8, was therefore further investigated as described below.

Overexpression of MIR156 in an spl8 Mutant Background Leads to Full Male Sterility
The quadruple mutant spl8 spl2 spl9 sp15 occasionally produced a few seeds per silique ( Figure 1D). To examine whether this leakiness is due to other still functioning miR156/7-mediated SBP-box genes in the spl8 mutant, we crossed spl8-1 into a homozygous 35S:MIR156b transgenic line in which the targeted SPL genes were confirmed to be all downregulated (Schwab et al., 2005; see Supplemental Figure 1 online). In their reproductive phase, these miR156-overexpressing plants were bushy with short siliques (Figure 2A). In addition, pollen production per anther and seed set per silique of these plants were also reduced to ;15 and 50%, respectively, of that of the wild type (30 anthers and 22 siliques examined, respectively). However, whereas the parental spl8-1 and 35S:MIR156b transgenic plants still produced some seeds, 29 out of 160 F2 plants showed a fully sterile phenotype under comparable growth conditions ( Figure 2A). All of these sterile plants were genotyped as homozygous for spl8-1 with one or two copies of the 35S:MIR156b transgene. Further geno-and phenotyping also revealed that a loss of SPL8 function enhanced the short and bushy phenotype of both the hemi-and homozygous 35S:MIR156b transgenics in a dosagedependent manner. In addition, the flowers of the fully sterile double mutant plants produced trichomes on the adaxial instead of on the abaxial sides of their sepals ( Figure 2D). In their overall size, the flowers did resemble those of the 35S:MIR156b transgenic line in size (i.e., displaying a squashed appearance but anthers remained small and did not release pollen) (Figures 2B to 2D;Schwab et al., 2005). To examine the fertility of the female organs of double mutant plants harboring one or two copies of the 35S:MIR156b transgene, we manually pollinated the clubshaped and short-stalked pistils with wild-type pollen ( Figure 2C, inset). Three weeks after pollination, siliques of such plants homozygous for spl8-1 and hemizygous for the 35S:MIR156b transgene had slightly elongated and contained four to eight seeds per silique (20 siliques examined). Siliques of plants homozygous for both spl8-1 and the 35S:MIR156b transgene carried only zero to three seeds per silique (21 siliques examined). Similar results were obtained with 35S:MIR156a spl8-2 and 35S:MIR156 h spl8-1 mutant plants (20 and 24 siliques examined, respectively). These results show that downregulation of miR156-targeted SPL genes in spl8 mutants resulted in nearly fully sterile plants. The observed residual female fertility may reflect leakiness in the repression of the miR156-targeted SPL genes by the 35S:MIR156 transgene. (A) Unrooted phylogram of all SPL genes obtained using the neighbor-joining algorithm and based on the conserved SBP domain. Nodes representing bootstrap values >50% from 10,000 replicates are encircled. Both subfamilies I and II are indicated by dashed lines. miR156/7-targeted SPL genes are depicted in green, whereas nontargeted large SPL genes and SPL8 are in blue and red, respectively. The alignment used to generate this tree is available as Supplemental Data Set 1 online. (B) and (C) Mean number of pollen sacs (B) and mean number of pollen (C) formed per anther of the wild type and mutants for SPL8, SPL2, SPL9, and SPL15 and combinations thereof, as labeled in the image. A total of 30 anthers from the 5th and 15th flowers of six primary inflorescences were analyzed. Small pollenless locules are referred to as underdeveloped. (D) Mean number of seeds set per silique of the wild type and multiple SPL gene mutants as labeled in the image. Siliques from the 5th and 15th flowers of a total of 20 primary inflorescences were examined. Error bars indicate SD.
miR156 Depletion by Sequestration Partially Rescues the spl8 Semisterile Phenotype So far, our results clearly showed that downregulation of miR156/7-targeted SPL genes strongly enhanced the spl8 fertility problem. Usually, the first few flowers formed in the primary inflorescence of spl8-1 mutants do not set seed at all. To investigate whether upregulation of miR156/7-targeted SPL genes could restore fertility in such flowers, we crossed a homozygous spl8-1 mutant to a homozygous transgenic miR156 target mimicry line. These MIM156 transgenic plants display a lengthened plastochron and flower after fewer leaves due to the upregulation of miR156-targeted SPL genes (Franco-Zorrilla et al., 2007). Furthermore, fertility of MIM156 transgenic plants, as deduced from the seed set per silique, seemed only slightly reduced compared with the wild type ( Figure 3C). In a resulting F2 population of 300 plants, we found 22 plants homozygous for spl8-1 that carried at least one copy of the MIM156 transgene. In comparison to the homozygous spl8-1 mutant alone, these plants showed a significantly (t test, P < 0.05) increased seed set, in particular with respect to the first flowers, although wild-type levels were not reached ( Figures 3A to 3C). This partial rescue through upregulation of miR156-targeted SPL genes (see Supplemental Figure 2 online) further support the hypothesis that at least some of the latter genes act redundantly with SPL8 to maintain fertility.
miR156-Resistant Forms of Several miR156/7-Targeted SPL Genes Attenuate the spl8 Semisterile Phenotype Eleven of the 17 SPL genes are targeted by miR156/7 ( Figure 1A). To get a better impression of which of these are able to restore the spl8 semisterile phenotype, we selected SPL3, -2, -11, -9, -15, -6, and -13 and constructed miR156/7-resistant forms of (A) Morphology of 6-week-old 35S:MIR156b and 35S:MIR156b spl8-1 plants under long-day growth conditions. Homo-or hemizygosity for the 35S:MIR156b transgene is indicated by (+/+) or (+/À), respectively, and homo-or heterozygosity for the spl8-1 mutant allele respectively by (À/À) or (+/À).  these genes (see Methods). Driven by a 2.1-kb upstream promoter region of the SPL8 gene, these miR156/7-resistant versions, hereafter referred to as rSPL3, rSPL2, etc., were subsequently introduced into an spl8-1 mutant background. As a control for full complementation, we transformed a similarly driven SPL8 transgene into the spl8-1 mutant. In the case of this control, 98% of all T1 plants displayed a wild-type-like fertility (i.e., full restoration of seed set; Table 1). By contrast, none of the T1 plants (n = 300) with rSPL3 or rSPL6 exhibited an improved fertility compared with spl8-1. For rSPL2, -9, -15, -11, and -13, only a low percentage (10 to 25%; n = 80 to 100) of the T1 plants displayed better fertility than spl8-1. To get a clue as to what might have caused only a minority of T1 plants to restore the spl8 mutant fertility partially, we examined the expression of the rSPL13 transgene in three partially rescued and three nonrescued transgenic plants selected randomly from a pSPL8: rSPL13 T1 population. The results showed that all the partially rescued T1 plants expressed the transgene to higher levels than did the nonrescued plants (see Supplemental Figure 3 online).
To quantify the degree of fertility rescue, progeny of five partially rescued plants from each of the respective T1 populations were grown and their seed set per silique determined. The partially rescued spl8 phenotype could still be observed in these T2 populations, although some segregation occurred, probably due to different transgene copy numbers. The seed set per silique is shown in Table 1.
Early Anther Development Is Affected in 35S:MIR156b spl8-1 Plants As described before, 35S:MIR156b spl8-1 plants form small and fully sterile anthers. To investigate what aspects of anther histogenesis are affected in these plants, we prepared transverse semithin sections through anthers at different developmental stages for comparison to the wild type and the respective single mutants. In the wild type (Figures 4A to 4E), as reported previously, subepidermal archesporial cells became detectable in stage 2 anthers ( Figure 4A; staging according to Sanders et al., 1999). At stage 3 ( Figure 4B), their periclinal division gave rise to the primary parietal and sporogenous cells. Subsequent mitotic divisions of these cells and their progeny then resulted in the distinct anther wall layers surrounding the pollen mother cells (PMCs) at stage 5 ( Figure 4D). Meiotic divisions of the latter cells followed by microspore release from the tetrads, together with further differentiation of somatic anther tissues, eventually resulted in anther locules filled with mature pollen as seen in stage 12 ( Figure 4E). In comparison, anther development in 35S: MIR156b transgenics seemed to follow the wild-type pattern, although less pollen was seen per locule ( Figures 4F to 4J). By contrast, anther development in 35S:MIR156b spl8-1 flowers deviated clearly ( Figures 4P to 4T). Archesporial-like cells appeared and underwent a few mitotic divisions (Figures 4P and 4Q), similar to stages 2 and 3 in the wild type. However, there appeared to be fewer cells already in stage 3-like anther sections, probably due to reduced cell proliferation. In subsequent developmental stages, no typical sporogenous cells differentiated. One or two subepidermal cell layers could still be observed, but from these no distinct anther wall layers developed ( Figure  4R). Instead, most of the anther cells became vacuolated (i.e., similar to the connective tissue in the wild type). A vascular strand developed, but formation of any pollen sacs failed (Figures 4S and 4T). This phenotype is much more severe than that of spl8-1 single and of sp8-1 spl2-1 spl9-1 spl15-1 quadruple mutants. In the anthers of these mutants, probably due to abnormal cell proliferation, the number and size of the final pollen sacs were reduced, but at least some pollen matured ( To verify further that cell proliferation was affected in the 35S: MIR156b spl8-1 anthers, we performed in situ hybridization using a Histone H4 (H4) probe as a cycling cell marker. Comparison of wild-type and 35S:MIR156b spl8-1 expression patterns at different anther developmental stages revealed that in anther stage 2, H4 expression was already slightly reduced in 35S:MIR156b spl8-1 ( Figures 5A and 5E). This difference became more obvious at later stages. Whereas developing wild-type anthers kept highly active dividing cells, particular at their four corners, at least up until stage 5 (Figures 5B to 5D), expression of H4 rapidly decreased in 35S:MIR156b spl8-1 anthers (Figures 5F to 5H), indicative of fewer cell divisions. Taken together, these data demonstrate that SPL8 together with miR156/7-targeted SPL genes are essential for proper cell proliferation in early anther development.

Expression of miR156/7-Targeted SPL Genes Becomes Detectable in Anthers of Later Arising Flowers
Our data clearly showed the importance of at least some of the miR156/7-targeted SPL genes in early anther development. As SPL2, -11, -13, and -15 were found to be able to rescue the spl8 phenotype partially, we selected these as representatives and tried to determine their temporal and spatial expression pattern by RNA in situ hybridization in anthers. To this purpose, apices of wild-type primary inflorescences were fixed at two time points (i.e., after 21 and 30 d of growing in long day) ( Figure 6A). At the 21-d time point, the inflorescences had just started to bolt and thus exposed the first floral buds formed. We examined all four miR156/7-targeted SPL genes at this time point, and no detectable signal could be observed in young anthers ( Figure 6B; see Supplemental Figure 5 online). However, at the 30-d time point and after the inflorescence produced many more flowers, at least SPL11 could easily be detected in stage 2 anthers ( Figure 6C). SPL11 expression continued to be in the sporogenous tissues at stage 3 to 4 ( Figure 6D) and in PMCs as well as tapetum at stage 5 ( Figure 6E), while the expression signal disappeared after meiosis (see Supplemental Figure 5 online). A rather weak signal could also be detected in stage 2 to 4 anthers for SPL13 ( Figures  6F and 6G). Up until stage 7, SPL13 expression seemed to be mainly restricted to tapetum ( Figure 6H). No expression was found in microspores and mature pollen (see Supplemental Figure 5 online). Sporogenous cells of stage 4 anthers also expressed SPL15 ( Figure 6I). Concurrently, a strong signal could be detected for SPL15 in the inner, placental side of the ovary before ovule initiation. SPL2 was not easily detected in young anthers, but occasionally sporogenous cells of stage 3 and 4 anthers seemed to react positively ( Figures 6J and 6K). As for SPL11 and SPL13, we did not find any expression of SPL15 and SPL2 at later (i.e., postmeiotic) anther developmental stages. It should be noted that expression of SPL8 was found to be similar at both time points and comparable to that shown in Figure   Our in situ hybridization data demonstrate that at least some of the miR156/7-targeted SPL genes are active in early anther development. Interestingly, however, their expression did not seem to become properly established in the first-arising flowers within the primary inflorescence. This would be fully consistent with the observation that the spl8 mutant does not set seeds in its first flowers despite the likely functional redundancy between SPL8 and the miR156/7-targeted SPL genes in anther development.

MIR156/7 Loci Show Highly Dynamic Expression Patterns during Anther Development
Although expression and function of miR156/7-targeted SPL genes seemed to correlate well with early anther development, the role of miR156/7 remained unclear. To gain insight into the temporal and spatial expression of MIR156/7 loci, we constructed promoter b-glucuronidase (GUS) reporter lines for all 12 MIR156/7 loci known in the Arabidopsis genome (MIR156a to -h and MIR157a to -d). After having screened 20 to 30 primary transformants (T1) per construct, we ended up with reproducible GUS activity in different developmental stages and various tissues for 10 out of the 12 loci (for some examples of flower expression patterns, see Supplemental Figure 6 online). For pMIR156e:GUS and pMIR156g:GUS, no or no consistent GUS staining could be detected. Among the 10 loci showing the GUS signal, pMIR156a-, pMIR156 h-, pMIR157c-, and pMIR157d:GUS were found strongly expressed in anthers; pMIR156 h:GUS was also expressed in ovules. pMIR156a:GUS activity was detected mainly in late anther development and in pollen (see Supplemental Figure 6 online). We selected pMIR156 h-, pMIR157c-, and pMIR157d:GUS lines for further analysis and focused on their expression patterns in cross sections through young an-thers. Interestingly, a strong GUS signal for all three loci could be detected around meiosis stage. pMIR156h and pMIR157c:GUS were not detected before anther stage 3 and 4, respectively ( Figures 7A, 7B, 7E, and 7F). From stage 4 anthers on, staining in entire anthers increased for pMIR156 h:GUS until stages 6 and 7 ( Figures 7A, 7C, and 7D). pMIR157c:GUS became detectable somewhat later at anther stage 6 ( Figures 7E to 7H). Whereas pMIR157d:GUS was detected in inflorescence apical meristems and young flower buds containing stage 2 anthers, in stage 3 and 4 anthers, its activity had disappeared ( Figures 7I and 7J). A strong GUS signal reappeared in the stage 5 anthers and was maintained through stages 6 and 7 ( Figures 7I, 7K, and 7L). For all three loci, the GUS signal was still detectable in anthers immediately after meiosis but vanished within two or three further developmental stages in most cases (see Supplemental Figure 6 online).
To quantify changes in GUS transcript levels in these transgenic plants over time, we harvested the very tip of their primary inflorescences at three successive time points: 25, 40, and 52 d after sowing and cultivation in long days. The transcript levels of pMIR157c:GUS and pMIR157d:GUS steadily decreased when the plants grew old, whereas the transcript levels first increased and then declined in the pMIR156h:GUS lines ( Figure 7M). We also compared the temporal GUS reporter transcript patterns to those of the corresponding endogenous pri-miRNA precursors. With the exception of the pri-miR156h that remained at a relative high level at 52 d, GUS reporter transcript and pri-miR levels were largely consistent ( Figure 7N).
Taken together, the GUS reporter lines unveiled a dynamic expression pattern for MIR156h, MIR157c, and MIR157d during early anther development and likely reflected in local miR156/7 levels. Before anther stage 4, low levels of miR156/7 would allow transcribed and targeted SPL genes to be fully active at these MIR156b spl8-1 anthers at this stage are nearly devoid of detectable H4 activity. Note that H4 activity also strongly declined in the gynoecium, whereas it is still easy detectable in developing petals. an, anther; g, gynoecium; p, petal. Bars = 50 mm.
stages. The subsequent rise in miR156/7 levels could be required to repress these SPL genes below a certain threshold before and during meiosis. As the inflorescence develops, early anthers from late-arising flowers may generally express lower levels of miR156/7 than those arising earlier.
A Relationship between SBP-Box Genes and the Early Anther Gene SPL/NZZ SPL8 together with miR156/7-targeted SPL genes promoted sporogenous cell formation, and their complete loss-of-function resulted in loculeless anthers similar to those of spl/nzz mutants.
To determine a possible interdependency, we examined the expression pattern of SPL8 and the miR156/7-target SPL11 in spl-1 mutant anthers with the help of RNA in situ hybridization. In addition, the expression of SPL/NZZ was determined in spl8-1 35S:MIR156b plants. In the wild type, the domain of SPL8 expression covered the four corners of stage 2 anthers where archesporial cells became initiated. Its expression was maintained in PMCs and in the anther walls except the epidermis up until stage 5 (Figures 8A to 8C). In the spl-1 mutant, the expression domain of SPL8 seemed to be expanded in stage 2 anthers ( Figure  8D), and at later stages, SPL8 expression could still be detected in the anther lobes ( Figures 8E and 8F). In comparison to SPL8, SPL11 seemed to be more broadly expressed in stage 2 and 4 wild-type anthers ( Figures 8G and 8H), and later in PMCs and surrounding cell layers at stage 5 ( Figure 8I). Strikingly, SPL11 expression was found to be slightly elevated in young spl-1 mutant anthers, and the signal appeared to be spread over the entire anther ( Figures 8J to 8L). In spl8-1 35S:MIR156b transgenic anthers, however, the normal expression of SPL/NZZ seemed not to be affected at very early stages of development ( Figures 8M  and 8P). At later stages, transcript levels of SPL/NZZ decreased in the transgenic anthers ( Figures 8N and 8Q, and 8O and 8R). Together, these data suggest that SPL/NZZ represses the SPL genes during early anther development and that these SPL genes may not be required for the initiation of SPL/NZZ expression.

The miR156/7 Levels Are Reduced in spl-1 Mutant Inflorescences
As miR156/7-targeted SPL11 transcript levels seemed to be elevated in spl-1 mutant anthers, we performed quantitative RT-PCR (qRT-PCR) on inflorescence tips to test if other miR156/7targeted SPL genes were also upregulated in the spl-1 mutant. In addition to SPL11, the related genes SPL10 and SPL2 were found to be upregulated more than 1.5 fold. SPL9 and SP15 transcript levels were also slightly increased ( Figure 9A). We then investigated the possibility that the levels of these different miR156/7-targeted SPL transcripts were elevated due to a decrease of miR156/7 in the spl-1 mutant. We followed the approach described by Yang et al. (2009) and detected mature miR156 and miR157 in the same inflorescence tips (see Methods and the Supplemental Protocol online). Compared with the wild type, the miR156 level in the spl-1 mutant was decreased 4-fold and the miR157 level 10-fold ( Figure 9B). These data support the idea that SPL/NZZ represses miR156/7-targeted SPL genes through promoting miR156/7, but a more direct repression cannot be ruled out.

DISCUSSION
Using genetic approaches and histologic analyses, we demonstrated a clear role for SBP-box genes in early anther development. The finding that loss-and gain-of-function of selected miR156/7targeted SPL genes enhances or attenuates, respectively, the semisterile phenotype of the spl8 mutant reveals a degree of functional redundancy among these SPL genes.

miR156/7-Targeted SPL Genes Are Required for Sporogenous Cell Formation in Anthers
After periclinal division of a subepidermal archesporial cell, the inner daughter cell (i.e., the primary sporogenous cell [PSC]) divides further (anther stage 2 to 4) to give rise to the PMCs seen at stage 5 of anther development. Progeny of the outer daughter cell (i.e., the primary parietal cell [PPC]) become part of the three cell layers (i.e., tapetum, middle layer, and endothecium) surrounding the PMCs (Sanders et al., 1999). In the spl8 mutant, PMC and parietal cell layer formation failed for some locules but succeeded in others thereby following the wild-type pattern and leading to viable pollen (Unte et al., 2003; this article). A disruption of the earliest stages of pollen sac formation is particularly well known from spl/nzz mutants. In spl/nzz mutant anthers, archesporial-like cells seem to be specified. However, these cells either do not divide periclinally to form PPCs and PSCs at all (Schiefthaler et al., 1999), or else their daughter cells do not give rise to microsporocytes and the different anther wall layers (Yang et al., 1999). This phenotype is very similar to what we observed after downregulation of miR156/7-targeted SPL genes by overexpressing MIR156b in the spl8-1 mutant. In this case as well, archesporial-like cells seem to appear and to divide but the PPC-and PSC-like daughter cells cannot form the typical anther wall layers and microsporocytes. As a result, small anthers develop, fully occupied with vacuolated cells (Figures 4P  to 4T).
These phenotypic similarities between spl/nzz and spl8 35S: MIR156b plants imply that these genes are involved in the same pathway for regulating pollen sac formation in early anther development. RNA in situ hybridization results suggest that SPL8 and SPL/NZZ become independently activated in stage 2 anthers. SPL/NZZ then seems to keep SPL8 expression restricted to the four corners of the anther. In spl-1 mutant anthers comparable to stage 4 to 5 in the wild type, tissues do not differentiate properly and SPL8 is found to be slightly downregulated. Similarly, SPL/NZZ is also slightly downregulated in anthers of the comparable stage in spl8-1 35S:MIR156b flowers. Furthermore, we observed that a SPL-GUS reporter transgene remains strongly expressed in the whole anther during its entire development in spl8-1 35S:MIR156b flowers (see Supplemental Figure 7 online). Together, these data show that sporogenous tissues do not form in the absence of SPL8 and other miR156/7targeted SPL genes and that these genes are required for SPL/ NZZ to act in promoting sporogenous cell formation.
Both male and female organs express SPL/NZZ during early and late stages (Schiefthaler et al., 1999;Yang et al., 1999). Likewise, the expression of SPL8 and other miR156/7-targeted SPL genes can also be detected in both male and female organs, consistent with an important role in reproductive organ development. However, in anthers, their expression remains restricted mainly to early stages (i.e., before meiosis), whereas SPL/NZZ activity can still be observed thereafter during pollen development (Dinneny et al., 2006).
As a direct target of the floral organ homeotic gene AG, and based on its dramatic mutant phenotype, SPL/NZZ is considered one of earliest genes required for initiation of micro-and  megasporogenesis. Despite its fundamental role in Arabidopsis fertility, both maize (Zea mays) and rice (Oryza sativa) seem to lack true SPL/NZZ homologs GenBank, July 2010). This strongly suggests that sporogenous cell formation in these, and probably other monocots, requires another regulator (s). A SPL/NZZ-like function in maize anthers may be fulfilled by MSCA1. The msca1 mutant also lacks sporogenous cell formation, although archesporial cells seem to become specified in its anthers (Chaubal et al., 2003).
Interesting in this context is that both miR156/7-targeted and nontargeted SBP-box genes, which according to our data regulate anther development similarly and as early as SPL/NZZ, are well conserved between monocots and dicots (Xie et al., 2006;Riese et al., 2007;Hultquist and Dorweiler, 2008). It would thus be interesting to determine if SBP-box genes play an evolutionarily more conserved role in sporogenesis. A difficulty may be that due to a high degree of functional redundancy, loss of function of one or two genes within this group may cause only minor fertility defects, as seen in Arabidopsis.

miR156/7-Targeted SPL Genes Promote Cell Proliferation
Cell proliferation is an important determinant of organ patterning and growth, which relies on many coordinated regulation factors, including cell cycle genes and other developmental factors (Rodriguez et al., 2010;Sozzani et al., 2010). Several lines of evidence indicate that SPL genes promote cell proliferation. The msc-1D mutant, representing a miR156-resistant spl15 mutant allele gene, shows increased cell number and decreased cell size in leaves. Overexpression of SPL3 also increases cell number and decreases cell size in leaves. By contrast, overexpression of miR156 results in the opposite (Usami et al., 2009). The spl8-1 mutant, displaying larger cotyledons and young leaves with reduced cell numbers and increased cell sizes, provides another example. Overexpression of SPL8 leads again to the opposite phenotype (Zhang, 2005). Here, we used the cell proliferation marker H4 to show that SPL8 and miR156/7-targeted SPL genes are required for maintaining cell proliferation in early anthers ( Figure 5). Taken together, these data clearly point to a role for SPL genes in cell proliferation, but the molecular genetic mechanisms involved need to be clarified.

miR156/7 Levels Follow Complex Spatial and Temporal Changes in Anthers
Recent studies demonstrated that variation in the level of miR156 and its targets is sufficient to promote or delay the juvenile-toadult development in Arabidopsis (Wu and Poethig, 2006;Schwarz et al., 2008;Wang et al., 2009;Wu et al., 2009). Starting high in young seedlings, the overall miR156 level declines as plants grow Wu et al., 2009). By contrast, the targeted SPL genes exhibit an opposite expression pattern. During the reproductive phase, flowers form with more complex tissues and organs and determining miR156/7 expression patterns in these developing floral organs is not an easy task, although in situ hybridization can be applied (Vá ló czi et al., 2006;Wang et al., 2008). Moreover, several different loci are able to contribute to miR156/7 levels. To get an impression of miR156/7 distribution during plant development, we generated transgenic Relative expression levels as determined by qRT-PCR on RNA from inflorescence apices of 40-d-old plants grown in long days, representing floral buds up until stage 12 (Smyth et al., 1990).  At the onset of the anther developmental program, the floral C function gene AG promotes the expression of SPL/NZZ. SPL/NZZ is required for proper expression of genes involved in cell division, differentiation, and specification and leads to PMC and tapetum formation. Equally critical at this stage appear to be the SPL8 and miR156/7-targeted SPL gene functions, either required to act together with SPL/NZZ or independently to regulate genes mediating cell division, differentiation, and specification. Lines ending with arrowheads indicate positive and those with rhombs negative genetic interactions. A question mark indicates a possible but yet unclear interaction. GUS reporter lines for the different MIR156/7 loci. Different organs at different stages of development showed highly dynamic expression patterns (see Supplemental Figure 6 online). Also, different loci showed overlap in expression in both time and space, suggesting some functional redundancy. For example, expression of MIR157a and -b, predicted to encode identical miRNAs, as well as the slightly different miRNA-encoding MIR156a, could be detected in late stages of pollen development (see Supplemental Figure 6 online). Overall, the GUS staining patterns obtained imply that the miR156/7 levels in different tissues and organs vary; accordingly, transcript levels of the targeted SPL genes are affected. In anther development, we found that expression of MIR156h and MIR157c and -d coincided with that of their targets around meiosis. This could mean that instead of fully repressing their targets, miR156/7 action at these stages involves fine tuning target transcript levels, a mode of action shown for some other conserved plant miRNAs (Cartolano et al., 2007). A study of MIR156/7 mutants is required to reveal the degree of redundancy and their mode of action.

SPL/NZZ Represses miR156/7-Targeted SPL Genes in Anthers
Our results demonstrated that SPL8, in concert with miR156/7targeted SPL genes, affects sporogenous cell and parietal cell formation very similarly to SPL/NZZ. However, it remains unknown if these genes are involved in the same pathway in regulating this process. A transcriptome analysis showed that both SPL/NZZ-dependent and -independent transcriptional programs are required for normal anther development (Wijeratne et al., 2007). Mutation of SPL/NZZ leads to a fold change of two or more of ;2000 genes on the level of transcription. Which of these represent direct targets of SPL/NZZ remains unknown. Interestingly, in situ hybridization showed that SPL11 was upregulated in the spl-1 mutant, a finding further supported with the help of qRT-PCR. The latter approach also revealed upregulation of other miR156/7-targeted SPL genes, such as SPL10, -2, -9, and -15 in the spl/nzz mutant. Our findings here are consistent with the published microarray data of Wijeratne et al. (2007), who reported upregulation of SPL11, -10, -5, and -15 in the spl-1 mutant. Taken together, the available data strongly supports the hypothesis that SPL/NZZ represses miR156/7-targeted SPL genes in anthers. As multiple and different miR156/7-targeted SPL genes are involved, it might well be that SPL/NZZ is actually repressing these genes through miR156/7. In fact, this assumption finds support in our qRT-PCR data showing that both miR156 and miR157 levels were reduced in the spl-1 mutant.
Our qRT-PCR data indicated that expression of MIR157c and MIR157d was downregulated but that of MIR156h upregulated in the spl-1 mutant (see Supplemental Figure 8 online). Whether SPL/NZZ is involved in general miRNA biogenesis or directly represses the MIR156/7 genes thus remains to be determined. Furthermore, we cannot rule out the possibility that SPL/NZZ directly represses miR156/7-targeted SPL genes, particularly in the first formed flowers where we did not detect expression of these SPL genes in early anther development ( Figure 6B).

A Model for Function of SBP-Box Genes in Early Anther Development
The SPL/NZZ gene plays a key role in early anther development by promoting other early anther genes required for cell division, specification, and differentiation, resulting in PMC and tapetum formation (Schiefthaler et al., 1999;Yang et al., 1999). The SPL/ NZZ gene is a direct target of the floral C function gene AG (Ito et al., 2004) and, based on its expression profile, clusters with SPL8 during early flower development (i.e., flower stages 5 to 7 [Wellmer et al., 2006] representing anther stages 1 to 3 [Sanders et al., 1999]).
Based on these facts together with the data presented in this study, we propose a model to integrate the function of SPL8 and miR156/7-targeted SPL genes in early anther development (Figure 10). In this model, it is assumed that the latter SBP-box genes, with the possible exception of SPL3, -4, -5, and -6, act redundantly to enable SPL/NZZ to activate its targets involved in cell proliferation and differentiation. They may, as well, act directly on the same targets. At these early stages, SPL/NZZ promotes BAM1 and -2 expression; in turn, the latter restricts SPL/NZZ expression to the center of the developing pollen sac (Hord et al., 2006;Wijeratne et al., 2007). Furthermore, SPL/NZZ is assumed to promote miR156/7 levels at later stages of anther development. In the first several arising flowers, miR156/7targeted SPL gene expression remains insufficient, possibly as a result of a still too high miR156/7 level in anthers as a remnant related to the preceding floral transition. Thus, SPL8 is absolutely required for proper early anther development during early inflorescence development. Mutation of SPL8 at this stage results in a severe sterile flower phenotype. Further decline of miR156/7 or other repressor levels will then allow miR156/7-targeted SPL transcripts start to accumulate in the flowers initiated at somewhat later stages of inflorescence development. At this stage, loss of SPL8 function is thus partially compensated for by these other SBP-box genes and fertility is only moderately affected ( Figures 1C and 1D). Obviously, this model will need elaboration and refinement, in particular when more detailed knowledge on the regulation of miR156/7, their SBP-box gene targets and targets thereof, becomes available.

Plant Materials and Growth Conditions
Arabidopsis thaliana ecotypes Columbia (Col-0) or Landsberg erecta (Ler) were used as wild-type controls as indicated per experiment. All mutants and transgenics are in the Col-0 background with the exception of spl-1, which is in the Ler background (N6586; Nottingham Arabidopsis Stock Centre). spl8 single mutants (Unte et al., 2003), the spl2-1 spl9-1 spl15-1 triple mutant (Schwarz et al., 2008), and the 35S:MIR156a (Wu and Poethig, 2006) and 35S:MIR156b (Schwab et al., 2005) transgenic lines have been described before. Seeds were imbibed and stratified for 4 d in the dark at 48C prior to sowing. Plants were cultivated on prefertilized soil mixture (Type ED73; Werkverband) at 21 to 238C under long-day conditions (16 h light) in the greenhouse. For comparative experiments, plants were grown in parallel under the same conditions, and material for gene expression studies was harvested at the same time points.

Plasmid Construction and Plant Transformation
For complementation of the spl8 mutant, a fragment encompassing a 2.1-kb upstream promoter region, a 1002-bp cDNA (open reading frame [ORF]) encoding full-size SPL8 protein, and a 190-bp 39-untranslated region fragment of SPL8 was inserted into the binary vector pGSA1252 (www.chromdb.org) by replacing the 35S promoter and GUS intron following BglII and SpeI digestion. The resulting plasmid is referred to as pSPL8:SPL8. The miR156/7-resistant rSPL2, rSPL11, rSPL9, rSPL15, rSPL6, and rSPL13 transgenes were generated by introducing mutations into the predicted miR156 response elements (see Supplemental  Table 1 online) using recombinant PCR. For the generation of rSPL3, only its ORF was amplified omitting the miR156 response element in its 39-untranslated region. The respective transgenes were subcloned into the pSPL8:SPL8 plasmid by replacing the SPL8 ORF following digestion with SmaI and either SacI or XbaI, resulting in plasmids referred to as pSPL8: rSPL3,pSPL8:rSPL2,pSPL8:rSPL11,pSPL8:rSPL9,pSPL8: rSPL15,pSPL8:rSPL6,and pSPL8:rSPL13. The GUS reporter constructs representing the various MIR156/7 loci resulted from amplification of ;2.5-kb upstream promoter region of the respective loci followed by cloning into the pGPTV-bar binary vector (Becker et al., 1992) linearized using SmaI and XbaI. To generate the p35S:MIR156h construct, a 450-bp genomic fragment encompassing the miR156h encoding region was amplified and cloned into the pBAR-35S vector (Xing et al., 2005) linearized using SmaI and XbaI. Oligonucleotides for mutagenesis and other PCRs used in generating the plasmids mentioned above are listed in Supplemental Table 2 online. All constructs were validated by sequencing before transforming into spl8-1 or spl8-3 mutants or wild-type plants, respectively, by the floral dipping method (Clough and Bent, 1998).

In Situ Hybridization
RNA in situ hybridization was performed as described previously (Xing et al., 2005). To generate antisense probes and to avoid cross-hybridization among SPL gene transcripts, only cDNA sequences downstream of the SBP-box from SPL2, SPL11, SPL13, and SPL15 were amplified. For SPL8, a fragment upstream of the SBP-box was added to a downstream fragment. Generation of an SPL/NZZ antisense probe was done as described previously (Xing and Zachgo, 2008). All primers used for the amplification of the probes referred to above, as well as for an antisense H4 probe, are listed in Supplemental Table 2 online (with the presence of a T3 or T7 RNA polymerase binding motif indicated in lowercase). T7 RNA polymerase (Roche) was used for in vitro transcription.

Histology and Microscopy
For comparative anther histology, the young floral buds representing different developmental stages were processed and sectioned as described by Sorensen et al. (2002). Transgenic GUS-expressing plant material was stained following the protocol of Mü ller et al. (2001) and analyzed using a Leica MZFLIII fluorescence stereomicroscope. For preparation of semithin sections through GUS-stained floral buds, samples were dehydrated in an increasing ethanol series (70, 96, 100%, v/v). Samples were then stepwise infiltrated with Technovit 7100 (Hereaus Kulzer) via a 3:1, 1:1, and 1:3 ethanol:Technovit series by incubating for 45 min at each step. Finally, the samples were incubated overnight in pure Technovit at room temperature. Technovit preparation and polymerization were done according to the manufacturer's protocol. One-to three-micrometer-thick sections were prepared using a Reichert ultramicrotome and mounted with entalan. All sectioned material was observed under a Zeiss Axiophot microscope. GUS precipitates were visualized using dark-field optics in combination with UV epifluorescence to visualize the tissue. Micrographs were recorded using a Leica DFC490 digital camera.
To determine pollen production, 30 unopened anthers were dissected from different flowers just before dehiscence and collected in Eppendorf tubes. Pollen were released by homogenizing in 20% glycerol and 13 PBS, pH 7, buffer (gly-PBS), pelleted by centrifugation, and resuspended in 100 mL gly-PBS containing 1% lacto-phenol fuchsin stain. Droplets of 5 mL were pipetted on microscopy slides covered with semidry 1% agarose to absorb the liquid, thereby generating flat spots of ;4 mm in diameter. These were photographed entirely and the fluorescing pollen grains counted using the public domain software ImageJ (Wayne Rasband, National Institutes of Health). Seed set was determined by clearing advanced but unopened siliques in Hoyer's solution and counting seeds directly under the stereomicroscope. To check pollen sac number in anthers, 8-mm-thick transverse sections were prepared from whole flower buds after dehydration and embedding in paraffin.

Real-Time qRT-PCR
Total RNA from inflorescence tips, carrying stage 1 to 12 floral buds, was isolated using the RNeasy plant mini kit (Qiagen). Five micrograms of total RNA, treated with DNase I, were used for first-strand cDNA synthesis with Superscript III (Invitrogen). qPCR master mix was prepared using the IQ SYBR Green Supermix (Bio-Rad) with specific primers for SPL2, SPL9, SPL10, SPL11, and SPL15 genes and MIR156h, MIR157c, and MIR157d precursors. PCR reactions were run and analyzed using the Bio-Rad IQ5 sequence detection system. Results were normalized to the expression of PP2A (Czechowski et al., 2005) mRNA. For determination of mature miR156/7 levels, total RNA was extracted using the mirVana miRNA isolation kit (Ambion). After the DNase I treatment, 200 ng of RNA were taken for cDNA synthesis with specific RT primers designed according to Yang et al. (2009). However, detection in the subsequent qRT-PCR relied on SYBR green instead of TaqMan technology as originally described by these authors (see Supplemental Protocol online). All specific primers used are listed in Supplemental Table 2 online. Representative qRT-PCR data of one of two to three independent biological replicates are presented in the figures with error bars representing the SD of three technical replicates.

Phylogenetic Tree Construction
Multiple alignments of protein sequences and phylogenetic reconstructions were performed by the program ClustalW and the neighbor-joining algorithm, both at default settings, within the MacVector 9.5.2 software package. MacVector's Best tree mode output was converted to the radial unrooted phylogram shown in Figure 1A using Dendroscope (Huson et al., 2007). MacVector's Bootstrap mode output with 10,000 replications was used to identify branching points that occurred in >50% of the resampling trees. For alignment and reconstruction, the SBP domain of 74-amino acid residues was used with two additional downstream residues (cf. to sequence logo provided in Birkenbihl et al., 2005). The alignment is available as Supplemental Data Set 1 online.

Remaining Techniques and Image Processing
Standard molecular biology techniques were performed as described by Sambrook et al. (1989). Graphical plots of numerical data were generated with Excel (Microsoft Germany), and statistical tests were performed using the Student's t test within this program. P values lower than 0.05 were considered to be statistically relevant. Digital photographic images were cropped using Adobe Photoshop CS (Adobe Systems). Color and contrast corrections were performed on entire images only. Figures were assembled in PowerPoint (Microsoft Germany).

Supplemental Data
The following materials are available in the online version of this article.   Figure 1A.
Supplemental Protocol. Detection and Quantification of miR156/7 Using qRT-PCR in Arabidopsis