The wild-type (WT) and mutant lines atino80-5 (Zhang et al., 2015) and m56-1 (Zhu et al., 2006) are all derived from the Columbia (Col) ecotype background. The reporter lines WOX5:GFP (Blilou et al., 2005), pPIN1:PIN1-GFP (Benková et al., 2003) and DR5rev:GFP (Friml et al., 2003) in Col-background have been described in previous studies. Seedlings were grown vertically on agar-solidified MS medium M0255 (Duchefa) supplemented with 0.9% sucrose at 21°C under 16 h light/8 h dark conditions. For the inhibition of polar auxin transport, N-naphthylphthalamic acid (NPA, 33371, Sigma-Aldrich) was added to the medium at the indicated concentrations.

The images of inflorescence were acquired by using a TM-3000 scanning electron microscope according to the manufacturer instructions (HITACHI). Differential interference contrast (DIC) images were taken with an Imager A2 microscope (Zeiss). For Lugol staining, roots were immersed in Lugol iodine solution containing 5% iodine for 2 min. After washing, roots were cleared with chloral hydrate solution (chloral hydrate: water: glycerol, 8:3:1, w/v/v). Confocal images were acquired by using a LSM710 microscope (Zeiss) with the following excitation/emission wavelengths: 561 nm/591–635 nm for Propidium Iodide (PI), 488 nm/505–530 nm for GFP. The antibody against γ-H2A.X was generated in our previous study (Zhou et al., 2016). The whole-mount root immunostaining was performed as previously described (Ma et al., 2018).

Plant organs were dissected by using a sharp blade and quickly frozen in liquid nitrogen. We used TRIzol kit to extract RNA according to standard procedures (Invitrogen). RT was performed using Improm-II reverse transcriptase (Promega). Quantitative RT-PCR was performed in three biological replicates. ACTIN2 (ACT2) was used as a reference gene to normalize the data. The gene-specific primers are listed in the Supplementary Table 1.

Chromatin immuno-precipitation was performed as described in our previous study (Zhang et al., 2015). All analysis was performed in three biological replicates. Antibodies used in this study were anti-GFP (A-11122, Invitrogen), anti-H2A.Z (Zhang et al., 2015), and anti-H3 (ab1791, Abcam). The gene-specific primers are listed in the Supplementary Table 1.

Our previous study has showed that the aerial part of m56-1 double mutant seedling resembles that of WT, while atino80-5 m56-1 triple mutant seedling resembles the single mutant atino80-5 (Zhou et al., 2016). Here, we confirmed the maintenance of such epistatic effect on aerial growth throughout the whole vegetative stage. Except the decrease in leaf size observed for atino80-5 and atino80-5 m56-1, no significant change of leaf phyllotaxy has been found in all the mutants (Supplementary Figure 1). After flowering, the WT flowers and siliques successively appeared along the branch axes in a Fibonacci spiral pattern. Such spiral pattern was not lost in atino80-5 and m56-1 inflorescences, albeit the spacing of atino80-5 siliques was shortened (Figure 1A). Intriguingly, we found an obviously disordered positioning pattern of siliques along floral branches of the atino80-5 m56-1 triple mutant. In many cases, several siliques appeared adjacent to each other without a spiral pattern. In addition, the development of most siliques and their fertility were greatly impaired in the triple mutant (Figure 1A). Notably, although the differentiation of flower organs was not generally affected in all the mutants, the organ size was reduced in atino80-5 and more severely in atino80-5 m56-1 (Supplementary Figure 2).

We observed and compared the IM by using electron microscopy (EM) (Figure 1B). There was no significant difference in IMs between WT, atino80-5 and m56-1, in which several floral primordia locate in a spiral pattern around the periphery zone (PZ). However, in the atino80-5 m56-1 triple mutant IM, the CZ significantly expanded but still with isotropy, and at the same time, extraordinary number of flower primordia at various growth stages emerged concurrently around the PZ, in line with the observed disordered inflorescence phyllotaxy. Our EM observation indicated that AtINO80 and NRP1/2 play a synergistic role in the maintenance of normal IM size as well as proper pattern of lateral organ initiation in IM.

Both AtINO80 and NRP1/2 participate in local chromatin remodeling for transcription modulation (Zhang et al., 2015; Zhu et al., 2017). Given the vital role of auxin in determining floral primordia formation and in controlling IM size, we wonder whether auxin pathway is interrupted in the atino80-5 m56-1 triple mutant. Therefore, we examined the transcription levels of several auxin-related genes in inflorescences. These include PIN1, ARF5, and the more downstream genes FIL, TMO3, ARR7 and ARR15. Notably, the transcription levels of most examined genes are synergistically mis-regulated (fold change > 1.5) in the triple mutant (Figure 2A), in line with its growth abnormality in inflorescence phyllotaxy and SAM size.

In IM, PIN1 determines the polar distribution of auxin and triggers the consequent transcriptional cascade and organogenesis (Reinhardt et al., 2000). Hence, the roles of AtINO80 and NRP1/2 in PIN1 transcriptional regulation were particularly examined in the following ChIP analysis by using inflorescences expressing EYFP-AtINO80 (Zhang et al., 2015) or EYFP-NRP1 (Zhu et al., 2017). Our ChIP results showed that EYFP-AtINO80 displayed enrichment at both 5′- and 3′-ends of the PIN1 gene, while a single peak of EYFP-NRP1 was found after the transcription start site of PIN1 (Figure 2B,C). These results indicated that PIN1 is the target gene of chromatin-remodeling factor AtINO80 and histone chaperones NRP1/2, and at the same time suggested that the observed higher mRNA level of PIN1 is not just the indirect result of enlarged IM size in the triple mutant inflorescence.

We also examined the occupancy of core histone H3 in WT and mutants in ChIP analysis. Relative H3 occupancy was slightly decreased in atino80-5 and m56-1, but was clearly decreased (fold change > 1.5 and P-value < 0.05) in most examined regions of PIN1 in the atino80-5 m56-1 triple mutant when compared to WT (Figure 2D), which is consistent with the observed PIN1 transcriptional change.

AtINO80 can regulate the local enrichment peak of histone variant H2A.Z within chromatin region of FLC, a key flowering suppressor gene (Zhang et al., 2015). Next, we analyzed the enrichment of H2A.Z relative to H3 (H2A.Z/H3) in PIN1. The H2A.Z/H3 peak was found near the 5′-end of PIN1 in WT, which is largely maintained also in all the mutants (Supplementary Figure 3). The atino80-5 mutant showed a reduction of H2A.Z/H3 but this reduction is compromised in atino80-5 m56-1, suggests that the H2A.Z dynamics is not associated with the synergistic effect of atino80-5 and m56-1 on the transcriptional up-regulation of PIN1 in the atino80-5 m56-1 triple mutant.

Compared with atino80-5 and m56-1, the triple mutant atino80-5 m56-1 also displayed an additive short-root phenotype (Figure 3A). We measured the primary root elongation of vertically grown seedlings. The root length of m56-1 became significantly shorter than that of WT from 8 day-after-germination (DAG), and that of atino80-5 mutant became significantly shorter than WT from 10 DAG, which are consistent with our previous studies (Zhu et al., 2006; Zhang et al., 2015). Remarkably, as early as from 4 DAG, the triple mutant has already shown an obvious inhibition of root elongation and the synergistic effect between atino80-5 and m56-1 became evident along the time course of root growth analysis (Figure 3B).

We observed and compared the root tips through DIC microscopy (Figure 3C). At 6 DAG, although the root length of atino80-5 and m56-1 were comparable to WT, their meristem size was smaller than that of WT, and again, we observed a much smaller meristem size in the triple mutant roots. At 10 DAG, only the meristem in WT sustained the original size, whereas the corresponding size in all mutants gradually decreased when compared with their younger state. Among them, the change in the triple mutant was most severe.

Dark treatment (skotomorphogenesis) can cause a decrease in both PIN1 transcription level and the shoot-to-root polar auxin transport in hypocotyl, resulting in auxin depletion in the RAM as well as the consequent reduced meristem size (Sassi et al., 2012). The skotomorphogenesis-associated mechanism seems to be compatible with the observed phenotype in triple mutant, thus prompting us to examine the mutants in dark treatment.

Under dark growth conditions, the hypocotyls of all the mutants elongated as those of WT (Supplementary Figure 4A). Moreover, dark treatment caused similar thinner roots and much smaller RAM in all the examined roots (Supplementary Figure 4B). These findings indicate that skotomorphogenesis is epistatic to AtINO80 and/or NRP1/2 depletion. Moreover, transcriptional analysis by using RNA extracted from hypocotyls revealed that PIN1 transcription level remained at a basal level in all the hypocotyls grown in dark. After light exposure, PIN1 was potently transcriptionally activated in hypocotyls in the triple mutant atino80-5 m56-1 (Supplementary Figure 4C), consistent with the synergistic role of AtINO80 and NRP1/2 in PIN1 transcriptional repression.

We further analyzed the transcription levels of several well-studied genes involved in RAM organization. They include: WUSCHEL-RELATED HOMEOBOX 5 (WOX5), SCR, SHR, PIN1, PIN2, PLT1 and PLT2. WOX5 is a homeobox gene specifically expressed in quiescent center (QC) in RAM identity (Kong et al., 2015). PIN2 encodes another PIN-family member which plays a root-specific role of auxin transport (Luschnig et al., 1998). Notably, the transcriptional levels of WOX5, PIN1 and PLT1/2 genes were synergistically and significantly up-regulated in the atino80-5 m56-1 triple mutant (fold change > 1.5) (Figure 4), suggesting that auxin pathway also undergoes a transcriptional mis-regulation in the triple mutant roots.

Chromatin Immuno-Precipitation analysis by using roots as material was performed to examine whether the recruitment of AtINO80 and NRP1 in PIN1 gene is consistent in different organs. Since transcription level of NRP1 is lower than that of NRP2 in Arabidopsis root (Supplementary Figure 5), we also introduced a transgenic plant expressing FLAG-NRP2 and included root-specific PIN2 gene in the same ChIP analysis. The recruitments of EYFP-AtINO80 and EYFP-NRP1 in PIN1 chromatin regions in roots was observed (Supplementary Figures 6A,B), with a pattern largely comparable to that previously described in inflorescences (Figure 2C), and the distribution pattern of FLAG-NRP2 was closely similar to that of EYFP-NRP1. In contrast, no obvious peaks of these proteins were found in PIN2 chromatin regions (Supplementary Figures 7A,B). The pattern of relative H3 occupancies in PIN1 was similar in roots with those in inflorescences (Supplementary Figure 6C). Meanwhile, reduction of relative H3 occupancy (fold change > 1.5 and P-value < 0.05) was found in some regions near the 5′-end of PIN2 in atino80-5 m56-1 triple mutant when compared to WT (Supplementary Figure 7C).

We also introgressed several fluorescent reporters including WOX5:GFP (Blilou et al., 2005), pPIN1:PIN1-GFP (Benková et al., 2003) and DR5rev:GFP (Friml et al., 2003) into each mutant background and observed their expression in root tips. A slightly stronger GFP signal of WOX5:GFP and pPIN1:PIN1-GFP were detected in the QC and steles in triple mutant, respectively (Figure 5, upper and middle panels). These findings are consistent with the above transcription analysis, and at the same time, also exclude the possibility that the severe short-root phenotype of triple mutant atino80-5 m56-1 may be caused by the depletion of QC, which is crucial for the maintenance of stem cell niche (van den Berg et al., 1997).

In a good proportion of examined triple mutant roots (3 out of 10 samples), ectopic GFP signal of WOX5:GFP reporter was detected in the presumptive position of columella stem cells (Figure 5, upper panel). To verify the function of the columella cells, we also examined the root tips with Lugol solution and found the WT-like accumulation of starches in the columella cell layers in all the mutants (each n > 10) (Supplementary Figure 8), indicating that the differentiation of columella cells was not significantly impaired in the absence of AtINO80 or NRP1/2.

Fluorescent signal of DR5rev:GFP is located in QC/columella cells and enriched on the acropetal side as a polar gradient in WT. This gradient pattern was little affected in atino80-5 and m56-1 root tips, but was moderately interrupted in the columella cell layers in the atino80-5 m56-1 triple mutant (Figure 5, lower panel), indicative of a disturbed auxin polar distribution.

To get more insight into the auxin transport in the triple mutant root, we transferred 4-day-old vertically grown seedlings to the culture medium containing different concentration of N-naphthylphthalamic acid (NPA), a synthetic inhibitor of auxin transport. The presence of NPA inhibited the root elongation, and this inhibitory effect is NPA-concentration dependent (Supplementary Figure 9). We observed the inhibitory effect at different concentrations of NPA on RAM (Supplementary Figure 10). Under NPA treatment at high dosage (5 μM), the WT RAM was not significantly changed even when the root length has been strongly suppressed. In comparison, the RAM structure of atino80-5 and m56-1 were obviously altered: the root hairs were much closer to the tips, a defect largely similar to that observed in the atino80-5 m56-1 triple mutant under untreated conditions. Notably, although treated with a low concentration of NPA (1 μM), the triple mutant root meristem displayed already an unusual expansion, which is accompanied by a quick differentiation of epidermal cells into root hairs (Supplementary Figure 10). Taken together, our observations indicate that the triple mutant roots are more sensitive to exogenous NPA treatment, providing additional evidences for its defects in maintaining functional auxin distribution in RAM.

It has been reported that root stem cells and their early descendants can be selectively killed by genotoxic treatment causing DSB (Fulcher and Sablowski, 2009). PI staining can enter and mark dead cells because of the interrupted membrane integrity. We noticed that the atino80-5 m56-1 triple mutant roots have accumulated PI-marked dead cells, which were barely found in WT or the atino80-5 and m56-1 mutant root tips (Figure 6A, upper panel). This observation suggests that AtINO80 and NRP1/2 synergistically prevent programmed cell death in root tips. H2A.X phosphorylation (γ-H2A.X) at the DNA break site constitutes one of the earliest events in the DNA repair process (Friesner et al., 2005). Although our previous study showed that the whole protein extracts from the triple mutant plants grown in the normal conditions did not show an obvious γ-H2A.X accumulation in Western blot analysis (Zhou et al., 2016), our immunostaining analysis detected weak but significantly visible γ-H2A.X loci in the root tips of the atino80-5 m56-1 triple mutant (Figure 6A, lower panel).

Double-strand break can also induce the early onset of endoreduplication in cortical cells, which is frequently associated with cell enlargement (Adachi et al., 2011). A plot of cortical cell area against the distance from QC revealed that the cortical cell expansion were more pronounced in the triple mutant than in atino80-5 and m56-1 when compared to WT (Figure 6B). Furthermore, we examined transcription levels of DNA damage-sensory genes PARP1/2 and DNA repair genes RAD51/54. All of these tested genes were synergistically up-regulated in the triple mutant roots (Figure 6C). Collectively, our data indicate that AtINO80 and NRP1/2 coordinate to maintain chromatin stability to prevent DNA damage for genome integrity.