Accession-specific modifiers act with ZWILLE/ARGONAUTE10 to maintain shoot meristem stem cells during embryogenesis in Arabidopsis

ZLL is required for meristem maintenance in an accession specific manner

The frequency of homozygous seedlings showing shoot stem cell termination in Ler zll alleles varies from 10 to 90%, depending on the mutation[8, 13, 20]. By contrast, the putative null T-DNA insertion mutants zll ^ago10-1 and zll ^ago10-3 in the Col accession have no or minimal effects on stem cell maintenance and meristem development (~0.2% in zll ^ago10-1 [11, 18]). To test if this difference is related to the nature of the respective mutant alleles, zll ^ago10-1 was backcrossed three times to Ler-0 wild-type. In homozygous Ler zll ^ago10-1 lines, 29% (n = 194) of the seedlings showed stem cell termination. In a converse experiment, the strong zll-1 EMS mutant allele was introduced into the Col background by crossing Ler zll-1 to Col-0 wild-type. Only a small fraction (0.5%, n = 5736) of the expected 25% zll-1 homozygous F₂ seedlings showed defects in stem cell maintenance (Figure 1B, C). This equates to an approximate phenotype of 2% in the homozygous state, compared to 61% (n = 315) in a cross between Ler zll-1 and Ler-0. This indicates that the different expressivity of the zll-1 mutation between Ler-0 and Col alleles is not due to the nature of the mutant alleles, but must be caused by genetic modifiers. To determine if the different requirement for ZLL between Ler and Col might be due to the erecta mutation in Ler, the Col zll ^ago10-1 allele was crossed to the strong er-102 and intermediate er-103 Col alleles (Figure 1B)[21]. Double mutants showed the characteristic erecta phenotype, but did not show an increased frequency of stem cell defects compared to the zll ^ago10-1 single mutant.

Differences in embryonic meristem size between Ler and Col were also investigated as a possible explanation for the different expressivity of zll mutations. Because the number of L1 cells in the mature embryonic meristem is indicative of the size of the meristem, wild-type embryos from the Ler-0 and Col-0 accessions were stained with propidium iodide and examined by confocal laser microscopy (Figure 1D). Grown under the same conditions, Ler-0 embryos (n = 17) contained ~11.4 (SD +/-1.4) cells in the L1 layer of the meristem at maturity, while Col-0 embryos (n = 25) contained ~10.1 (SD +/-1.5) cells. This indicates that the L1 layer of the Ler-0 embryonic meristem, and hence the meristem itself, contains more cells than the Col-0 meristem (Students t-test P < 0.01). This opposes the hypothesis that a smaller meristem in Ler-0 might be the cause of increased sensitivity to zll mutations, but suggests that fundamental differences in embryonic meristem regulation may contribute to variable zll expressivity in these accessions.

ZLL is required for meristem maintenance in multiple Arabidopsis accessions

To determine if the differences in ZLL-dependency for stem cell maintenance are restricted to Ler-0 and Col-0, the Ler zll-1 allele was crossed to 28 different wild-type Arabidopsis accessions originating from diverse countries (Figure 2). These accessions and others were analysed previously with 149 single nucleotide polymorphisms (SNPs) to address Arabidopsis population structure[22]. Analysis of 59 SNPs that produced clear genotypes in the 28 accessions used here confirmed that apart from the pairs of Berkley and Col-0, and Ct-1 and En-1, the accessions were different (Figure 3).

The frequency of seedling meristem termination, indicative of stem cell defects during embryogenesis, was assessed in the F₂ generation of each accession cross to determine the approximate phenotypic frequency in the homozygous zll-1 state (Figure 2). Of the 28 accessions, five showed stem cell termination in more than 20% of the homozygous seedlings, eight showed a phenotype in 10 to 20%, and fifteen showed a phenotype in less than 10% of seedlings. Curiously, the five accessions that showed the highest zll-1 expressivity, Ler-0, Freiburg-1 (Fe-1), Neuweilnau-0 (Nw-0), Bayreuth-0 (Bay-0) and Enkheim-1 (En-1), were all derived from locations in the southern half of Germany. Despite this geographical association, there is no obvious clustering of these accessions in a phylogenetic tree based on 59 SNPs to suggest they were more related to each other than accessions showing weak stem cell termination phenotypes (Figure 3).

Multiple quantitative trait loci (QTL) influence ZLL-dependent stem cell maintenance

To locate candidate modifiers in the Arabidopsis genome, we utilized a population of Ler/Col Recombinant Inbred Lines (RILs)[23]. Each line contains a different combination of Ler-0 and Col-0 genomic regions and has been genotyped, providing an excellent resource for mapping QTL. The Ler zll-1 allele was crossed to 99 RILs and the two parents, and F₂ progeny were scored for stem cell termination phenotypes (Figure 4A). The expectation from this cross was that RILs containing Ler-0 alleles at the position of putative QTL would show a higher frequency of seedlings with stem cell defects than lines containing Col-0 alleles at the same position.

Near Isogenic zll-1 Lines (NILs) showing shoot stem cell defects preferentially retain genomic regions linked to the Ler FHE loci

To further assess the contribution of FHE^Ler loci to stem cell maintenance, the zll-1 mutation was introgressed from Ler to Col. Twenty independent F₂ plants showing stem cell defects in the first backcross of Ler zll-1 to Col-0 (termed BC₁F₂) were further backcrossed to Col-0 wild-type three to six times. In each backcross, F₂ plants that showed stem cell defects at the seedling stage were selected for further backcrossing. At the same time, F₃ seeds from self-fertilised siliques were harvested to assess the frequency of stem cell defects in the progeny. Each backcross was expected to remove non-essential Ler DNA, while retaining Ler genomic regions that enhance zll stem cell termination phenotypes. Although some lines showed relatively stable levels of stem cell defects over several backcrosses (Figure 5), the majority showed a decrease with each subsequent backcross (the average phenotypic frequency of all lines decreased from 31 ± 16% after BC₁ to 13 ± 8% after BC₃), consistent with the gradual accumulation of Col-0 modifiers suppressing stem cell termination. This gradual decrease strongly indicates that multiple loci suppress stem cell termination in the Col-0 accession in a quantitative manner.

Two lines (NIL22 and NIL28) that maintained relatively high frequencies of stem cell termination over several backcrosses were selected for mapping and phenotypic analysis after BC₅. Genomic DNA from phenotypic NIL22 BC₅F₂ plants (n = 48) and NIL28 BC₅F₂ plants (n = 48) was used for bulk mapping with 19 PCR-based markers (Additional file2) that detect Ler/Col polymorphisms in the vicinity of the predicted FHE QTL and at unlinked sites throughout the genome. In both NILs, PCR results identified a strong bias towards Ler DNA at markers MT435 (~32 cM) and nga168 (~73 cM) on chromosome 2 and ciw9 (~90 cM; zll-1) on chromosome 5, while the rest of the genome was biased towards Col.

Changes in gene expression between ZLL-independent and dependent accessions identify candidates for the FHE loci and downstream pathways

Despite further backcrosses, most zll-1 Col NILs preferentially retained a large fragment of Ler genomic DNA around FHE2^Ler which complicated fine mapping. To further delineate putative FHE factors and pathways that influence stem cell maintenance in a ZLL-dependent manner, microarray profiles were generated from specific ZLL-dependent and independent Arabidopsis accessions using Affymetrix ATH1 chips.

Torpedo-stage embryos were harvested separately from four ZLL-dependent accessions (Ler-0, Fe-1, Nw-0, and Bay-0), and three ZLL-independent accessions (Col-0, Van-0 and Ts-1). Consistent with previous SNP genotyping results[22], none of the accessions showed an identical pattern when genotyped with 14 INDEL markers, confirming that they are genetically unique. Multiple comparisons were made between expression profiles derived from the two groups of accessions by maintaining a minimum of three arrays per group. A total of 439 genes were identified as being differentially expressed in at least one of the comparisons, based on a p-value of <0.05 and a 3-fold expression change (Additional file3). In the most stringent comparison, all of the ZLL-dependent accessions were grouped as replicates and compared to the ZLL-independent accessions. Two genes were identified as being up-regulated and ten were identified as being down-regulated (Additional file4). None of the genes appeared to be tightly linked to the predicted FHE loci from the RIL analysis, suggesting that they may either represent factors that act downstream of the FHE modifiers, they are unrelated to ZLL function, and/or they represent modifiers that escaped detection in the Ler/Col FHE QTL mapping.

FHE2 maps close to the Arabidopsis Cyclophilin-40 homologue SQUINT, which is differentially expressed between Col-0, Fe-1 and Ler-0 accessions

Changes in SQN expression influence stem cell maintenance in the Ler-0 and Col-0 accessions

Different genetic backgrounds influence the role of ZLL in stem cell maintenance

Previous studies in Arabidopsis have utilised variation between natural accessions to identify the genetic basis for developmental differences[27, 28], including small RNA-mediated regulation of flowering time[29], self-incompatibility[30] and root growth[31]. In the current study, natural genetic modifiers that influence stem cell maintenance in the absence of ZLL function were investigated. The aim was to identify novel components that support ZLL in promoting stem cell maintenance through the regulation of WUS activity[10] and/or modification of miRNA function[12]. QTL mapping in a population of Ler/Col RILs and zll-1 Col-0 NILs suggested that five FHE loci can explain 49% of the variance in stem cell maintenance in Ler-0 and Col-0. In all cases, the presence of the Ler FHE alleles with zll-1 resulted in an increased frequency of stem cell termination, consistent with Col zll mutants showing limited degrees of stem cell termination. The Ler/Col RILs were previously analysed for variation in shoot regeneration from tissue culture, which depends on three QTL on chromosomes 1, 4 and 5[32, 33]. The position of these loci is distinct from the FHE QTL, suggesting that the FHE loci are unlikely to be involved in shoot formation per se, and are more likely to be involved in embryonic meristem function.

Our findings also indicate that differences in ZLL-dependency are not limited to the Col-0 and Ler-0 accessions and considerable variation exists between different accessions from North America, Europe and Asia. The two accessions showing the highest frequency of stem cell defects in the presence of zll-1, in 55% and 61% of zll-1 seedlings respectively, were Fe-1 and Ler-0. Limited information is available for the Fe-1 accession, but previous studies of natural genetic variation show that it diverges from Ler and Col in its response to pathogen susceptibility[34, 35]. No obvious phenotypic differences in growth habit, flower development or embryo morphology were detected between Fe-1 and Col-0 to suggest such a prominent difference in response to loss of ZLL function.

Although our analyses are far from saturating, most ZLL-dependent accessions analysed here were collected from regions within middle and southern Germany. Recent advances in SNP detection and the availability of large genomic sequence datasets from diverse accessions allows trait variation to be dissected by genome wide association studies (GWAS), which offers a much higher mapping resolution compared to the RILs[28, 36–38]. The number of accessions analysed in this experiment was insufficient for robust GWAS, but such an approach might be useful in future studies to fine map the FHE loci and to identify additional loci that contribute to ZLL function in stem cell maintenance. Identification of the genes underlying the FHE QTL will show whether the ZLL-dependent accessions share a common recent ancestor containing a set of genetic modifications influencing stem cell maintenance, or if geographical conditions have independently influenced selection of polymorphisms in the modifier loci.

Conserved differences in embryonic gene expression are detected between different Arabidopsis accessions

The first defects in stem cell maintenance in zll-1 mutants are observed at the torpedo stage of embryogenesis[10], suggesting that FHE modifiers of ZLL function should be active at this stage. Microarray analysis identified multiple genes showing natural variation in embryonic gene expression at the torpedo-stage. These expression profiles suggest it is unlikely that any causative polymorphisms influencing ZLL-dependency in the seven analysed accessions lead to common changes in mRNA expression of the genes underlying the Ler/Col FHE loci. This is not surprising, since the FHE QTL may differ between diverged accessions and the specific polymorphism(s) leading to ZLL-dependency may not lead to a change in mRNA expression, but rather have effects on protein function or accumulation.

Of the 12 genes differentially expressed in the combined ZLL-dependent versus ZLL-independent accessions, none were tightly linked to the Ler/Col FHE loci or had documented functions in meristem development or RNAi. In addition, only two of the genes showed any expression correlation across a developmental series (0.84; At1g78820 vs At5g28770; Genevestigator[39]), suggesting that the group are unlikely to be associated closely in the same pathway. Despite this, it is possible that variable expression of these genes in the different accessions is at least partly dependent on activity of the FHE loci. This is also possible for the remaining 427 genes that showed accession-specific expression during embryogenesis.

A hypomorphic SQUINT allele may support stem cell maintenance in Columbia zll mutants

Restricting the embryonic expression profile comparisons to the three most relevant accessions (Col-0, Ler-0 and Fe-1), in combination with double mutant analysis, identified SQN as a candidate modifier underlying the FHE2 QTL. In Col-0, embryonic SQN expression is 5-fold lower than Ler-0 and Fe-1. Consistent with a role in modifying stem cell development, decreased SQN activity in Ler via sqn-4 and amiRSQN partially rescued stem cell maintenance in the zll-1 background, while increased SQN expression in Col zll ^ago10-1 induced stem cell termination. Although the effects were greater than the predicted quantitative contribution of the FHE2 QTL, this may be due to the nature of the polymorphisms between SQN ^Ler and SQN ^Col . Only synonymous SNPs are present in the SQN coding sequence between Ler-0 and Col-0, indicating that differences in enzyme amino acid sequence cannot explain differences in function. In contrast, significant variations including insertions and deletions are present in the 5’ sequence upstream of the SQN gene ([40]; Additional file5). Notably, a 6.6 kb MULE-related transposon sequence, annotated as At2g15800/At2g15810, is inserted close to the transcriptional start site of SQN. This insertion is located 500 bp upstream of the predicted SQN start codon in Col-0 but is absent from Ler-0 (Additional file5). The presence of this insertion varies between Arabidopsis accessions[41], and may contribute to natural variation in SQN expression as detected for other genes tightly linked to transposon sequences[42].

Although variable SQN expression levels correlate with differences in stem cell maintenance in Ler-0, Fe-1 and Col-0, and to a lesser extent in the Bay-0 and Nw-0 accessions, this is not the case in all accessions examined. SQN mRNA levels in the ZLL-independent Ts-1 and Van-0 lines were unchanged relative to Fe-1/Ler-0. Therefore, alternative FHE loci may play a more important role in these accessions. It is possible that some of this variation may be due to subtle transcriptional or post-transcriptional changes in the function of other meristem or RNAi-related genes physically linked to the FHE loci reported here (Additional file6). Further analysis of F₂ progeny from Ler zll-1 and Ts-1 or Van-0 crosses will allow the major FHE loci that influence stem cell maintenance in these accessions to be positioned.

Plant material

Seeds were germinated on soil and grown as described previously[43]. The Col zll ^ago10-1 [18] and Ler zll-1[8] mutants have also been described previously. Seeds from the Ler/Col Recombinant Inbred Lines (N4859) and various Arabidopsis accessions were obtained from the Nottingham Arabidopsis stock centre (NASC). Single Nucleotide Polymorphism (SNP) haplotype alignments were created using publically available data (http://www.naturalvariation.org) in Geneious (http://www.geneious.com/). Maximum-likelihood trees were generated in Mega5.2[44]. Defects in stem cell maintenance were scored in seedlings between 11–15 days post germination. Seedlings that contained an empty apex, a single filament, a single leaf or two leaves in place of a viable shoot meristem, were scored as showing stem cell defects, as per previous studies[8, 10]. Although phenotypic zll mutants terminate primary meristem development, secondary adventitious meristems produce viable flowers that can be used for crossing.

Meristem measurements

The number of cells in the embryonic meristem was determined by staining with propidium iodide and confocal laser microscopy as described previously[45]. Confocal laser microscopy was performed at the Life Imaging Center (LIC, Freiburg).

Mapping

All new markers used in this study were PCR based, and designed from the Cereon collection[46] to detect insertions/deletions (INDELs) or single nucleotide polymorphisms (SNPs) by derived cleaved amplified polymorphism (dCAPS) primers. Primer sequences are shown in Additional file8.

QTL analysis

Molecular map information for the Ler/Col RILs was downloaded from the NASC website (http://Arabidopsis.info/RI_data/full_markers.text). QTL analysis was performed with the software package PlabMQTL[47] using composite interval mapping[48, 49] and a multiple regression procedure[50]. Cofactors were selected based on the modified Bayesian Information Criterion[51] and critical Logarithm of the odds (LOD) thresholds were determined empirically with 1,000 random permutations[52]. The proportion of variance explained by the detected QTL (p) was obtained from the adjusted R² value of the QTL model and the proportion of variance explained by individual QTL by normalizing to sum up to the total p.

Microarray profiling

Torpedo-stage embryos were dissected from maturing seeds in 1×PBS and stored on ice for no longer than 1 hour before snap freezing in liquid nitrogen. Approximately 100 embryos were harvested from each accession. RNA was extracted using the RNeasy Plant Minikit (with on-column DNAse treatment; Qiagen) according to manufacturer’s instructions. 10 μg total RNA was hybridised to Affymetrix ATH1 chips at ATLAS Biolabs (Berlin, Germany). Expression analysis and normalisation was performed in R using the RMA package, following a previously established pipeline[53]

Cloning

Three artificial miRNAs targeting SQN were designed using the Web MicroRNA designer program; http://wmd3.weigelworld.org/cgi-bin/webapp.cgi[54] and cloned into the pJet2.1 expression vector. After sequencing, the amiRNA constructs were sub-cloned via BamHI digest into pEG278. Finally, p35S:amiRSQN-term was cloned into the PacI site of the pGreen II destination vector[55]. Plant transformation was carried out by Agrobacterium-mediated floral dipping. After selection of T1 and T2 transformants, T3 homomozygous lines were used for final analysis.

To generate pAtRPS5a:SQN, genomic DNA of SQN was amplified by oFR126-Fr/oFR127-Rev primers harboring LIC cloning fragments and subcloned into the pJet2.1 expression vector. After sequencing, the SQN genomic fragment was cloned using the LIC cloning protocol[56], into a modified pGreenII vector containing the AtRPS5a promoter, LIC cloning site and Nos terminator. Full primer details are available on request.

Availability of supporting data

The microarray data sets supporting the results of this article are available in the Gene Expression Omnibus (GEO) repository, accessible via the GSE47884 identifier.