All A. alpina mutants were isolated from EMS-mutagenized A. alpina Pajares and pep1-1 populations (Wang et al., 2009; Chopra et al., 2014). The A. thaliana scn1-3 mutant was described previously (Carol et al., 2005).

For root hair analysis, seeds were surface sterilized with chlorine gas for 3h. Sterilized seeds were sown on full Murashige-Skoog plates (Murashige and Skoog, 1962) w/o sucrose and stratified at 4°C for 5days. Plants were grown vertically for 7days under long-day conditions (16-h light, 8-h darkness) at 21°C.

For whole-genome sequencing, plants were grown on plates as described before and the root phenotypes were confirmed in the M3 generation. Next, the seedlings were transferred to soil for three more weeks. Fresh leaf samples of 1-month-old plants were used for DNA extraction.

Arabis alpina gene sequences were obtained from the Genomic resources for A. alpina website¹ and analyzed with CLC DNA Workbench 5.6.1.

For sequence analysis by Sanger sequencing, primers were designed outside the CDS of a given A. alpina gene to sequence it in the mutants. Library preparation and whole-genome sequencing were carried out by the Cologne Center for Genomics, and the raw sequence reads were deposited into National Center for Biotechnology Information² under accession number PRJNA745061. Short reads of each mutant sample were aligned to the A. alpina reference genome version 5.1 (Jiao et al., 2017) using Bowtie2 version 2.2.8, which were further processed with SAMtools version 1.4 (Li et al., 2009) to get BAM files. Next, each BAM was provided to SHORE version 0.8 (Ossowski et al., 2008) to call consensus and variants for each mutant sample. SNPs were stringently selected for each mutant with a minimum mutant allele frequency of 0.85 and a coverage of at least 3. Effects of selected mutations on gene integrity were annotated with SHOREmap version 3.0 (Schneeberger et al., 2009; Sun and Schneeberger, 2015).

The EMBL-EBI database (Pfam 32.0³; Finn et al., 2016) was used to identify conserved domains.

GBrowse from TAIR 10,⁴ along with the assembled A. alpina genome, was used to confirm the synteny of the selected genes regarding conserved order and appearance of the neighboring genes.

AtSCN1 and AaSCN1 were amplified by PCR, using A. thaliana Col-0 or A. alpina Pajares cDNA as templates, and subsequently cloned into pDONOR201 and/or pDONOR207 using the Gateway system (Thermo Fisher Scientific). For construction of 35S::AtSCN1-YFP and 35S::AaSCN1-YFP, the Gateway destination vector pEXSG-YFP (Feys et al., 2005) was used. For construction of the Aascn1-1 mutant CDS, site-directed mutagenesis was performed on pDONOR207-AaSCN1 and Aascn1-1-YFP was subsequently cloned into pEXSG-YFP (Feys et al., 2005). Plants were transformed using floral dip (Clough and Bent, 1998).

The number of root hairs was determined in a 1mm long section as described before (Grierson and Schiefelbein, 2002). The position of H-files was determined on 7-day-old seedlings with respect to the position of underlying cortex cells.

For epidermal cell length analysis, roots were stained with 100mg/ml propidium iodide for 1min and subsequently washed with water. Nitroblue tetrazolium chloride (NBT, Sigma) was used to identify sites of superoxide production (Fryer et al., 2002). Seedlings were covered with NBT solution (0.5mg/ml NBT in 0.1M potassium phosphate pH 7) for 10min.

Stereo microscopy was carried out with the Leica MZ 16F stereo microscope and the LAS AF software (Leica Microsystems, Heidelberg, Germany). Light and fluorescence microscopy were performed with the Leica DMRA2, DMRB, and DM5000B fluorescence microscope (Leica Microsystems, Heidelberg, Germany). Confocal laser scanning microscopy was carried out with the Leica DM5500 CS Microscope and documented with the TCS-SPE imaging systems (Leica Microsystems, Heidelberg, Germany).

Adobe Photoshop CS4 version 14.0.0 and ImageJ (Fabrice Cordelieres, Institut Curie, Orsay, France) were used for image processing.

Data were analyzed with Microsoft Excel 2007. Statistical analysis was carried out with OriginPro 8.5 0G SR0 and Microsoft Excel 2010. Significance was tested using a two-sided t test for normally distributed data and one-way ANOVA for data with non-normal distribution.

Root hair development in A. alpina is similar to A. thaliana. Root hairs are initiated at the basal end of root epidermal cells, followed by extensive tip growth (Figures 1A,D; Grierson and Schiefelbein, 2002). However, all cell files in A. alpina can produce root hairs [(Chopra et al., 2014), Figure 1A]. Nevertheless, trichoblasts and atrichoblasts differ in cell length (Chopra et al., 2014), so it is conceivable that A. alpina also produces morphologically distinct H- and N-files in the root epidermis. To substantiate this, root hairs were stained with propidium iodide to identify the H-file position by their position over the cleft of the underlying cortex cells (Figures 1B,C). We measured the length of four consecutive cells in N- and H-files starting immediately below the first cell displaying the formation of a root hair bulge. Cells located close to the root tip show no length difference. More distant cells were about 30% shorter in H-positions as compared to N-positions (Figure 1E), indicating that a morphological difference between these cell types is conserved between the two species.

We used a forward genetic approach to define different functional steps during root hair development in A. alpina. Two EMS screens were performed in independent populations: one in Pajares (Wang et al., 2009), representing 4,205M1 plants, and the other in the pep1-1 background (Wang et al., 2009; Albani et al., 2012; Bergonzi et al., 2013; Zhou et al., 2021), representing 6,800M1 individuals. Seeds of five M1 plants were pooled, and 50 M2 seedlings from each pool were screened for root hair phenotypes. Forty-five root hair pattern and root hair morphology mutant phenotypes were confirmed in the M3 generation (Table 1). We found two types of patterning mutants: Seven mutants showed more root hairs (Figure 2B), and six had fewer or no root hairs (Figure 2C) compared to the wild type (Figures 2A and 3A). The morphology mutants were divided into six classes. Three lines produced only small bulges (Figure 3B), suggesting that cell elongation is disturbed. In six mutants, hairs initiated from a swelling of the epidermal cell (Figure 3C). Three lines showed a bursting phenotype (Figure 3D). Typically, these root hairs ruptured at the tip soon after initiation. One line produced multiple sites of growth from each epidermal cell (Figure 3E). Sixteen mutants exhibited short hairs (Figure 3F), and three mutants produced branched (Figure 3G) and wavy root hairs (Figure 3H). A similar range of phenotypes has been previously described in A. thaliana (Grierson and Schiefelbein, 2002).

A. alpina and A. thaliana are both members of the Brassicaceae and therefore relatively closely related (Koch et al., 2006; Beilstein et al., 2010). Hence, we reasoned that most mutant phenotypes found in the A. alpina root hair screen are probably caused by mutations in orthologs of genes shown to affect root hair development in A. thaliana. In a first step, we selected genes known to be involved in root hair development of A. thaliana (Supplementary Table S1). We also included genes involved in trichome development, because of genetic overlap between the two processes, and pollen tube genes, since pollen tubes show tip growth similar to root hairs. In total, we considered 136 A. thaliana genes. In a next step, we identified 283 corresponding paralogs or orthologs based on the annotated A. alpina genome (Willing et al., 2015; Supplementary Table S1). This way, we constructed a list of candidate A. alpina genes that might be impaired in function in our set of root hair mutants.

We then sequenced candidate genes in selected root hair mutants, as previously described for trichome genes in A. alpina (Chopra et al., 2014, 2019). However, the high number of candidate genes rendered gene by gene sequencing inefficient. Therefore, we performed whole-genome sequencing of the genomic DNA from different mutants, aiming for an average of 15X fold coverage. Each SNP identified in the genes of interest is listed in Supplementary Table S2 along with the position in the chromosome, the sequence of the reference allele and the mutant allele, the location of exchange regarding the gene model, mutant allele coverage, and mutant allele frequency. Annotated SNPs, stringently selected for each mutant, with a minimum mutant allele frequency of 0.85 and minimum mutant allele coverage of 3, were chosen for further analysis. Next, we correlated SNPs to the phenotype of the respective mutants. As summarized in Supplementary Table S3, 26 mutant-specific alleles were identified. We considered stop codons as mutant-specific alleles, while other mutations were only considered when leading to a non-synonymous amino acid exchange.

In A. thaliana, the R2R3MYB transcription factor WER (Lee and Schiefelbein, 1999) and the WD40 protein TTG1 (Galway et al., 1994) form a complex with the bHLH proteins GL3/EGL3 (Payne et al., 2000; Zhang et al., 2003) that activates the expression of GL2, which in turn suppresses root hair development in non-root hair cells (Masucci et al., 1996).

As expected, various mutants with ectopic root hairs showed relevant mutations in AaTTG1 and AaGL3. Similar as had been shown for trichomes (Chopra et al., 2019), Aagl3 mutants showed a strong phenotype and none of the mutants was defective in AaEGL3. We found no mutations in AaWER. This is particularly interesting as we have previously found no gl1 mutants in glabrous trichome mutants (Chopra et al., 2019). While this may be due to the mutant screen not having been fully saturated, it is also possible that the MYB genes act redundantly in A. alpina, such that single gene mutations do not lead to a phenotype.

Surprisingly, all six mutants displaying a reduced number of root hairs had no mutations in any of the known A. thaliana genes in which mutations lead to fewer root hairs.

The analysis of the root hair morphology mutants derived from our screen revealed that various genes, including ARK1, SPI, COW, PFN1, CAP1, and KJK, have similar functions as described in A. thaliana. In addition, we found one short root hair mutant with a relevant amino acid substitution in another AaKJK family member, the AaCSLD6 gene. Although the corresponding mutant was not yet described in A. thaliana, our finding suggests that CSLD6 is involved in root hair morphogenesis.

We also found possible differences in the gene-phenotype relationships. While in A. thaliana only lrx1 and the lrx1 lrx2 double mutants show a root hair phenotype (Baumberger et al., 2003), we found that a relevant amino acid exchange in the AaLRX2 gene correlates with a short root phenotype. This suggests that the relative importance of the AaLRX1 and AaLRX2 genes is different in A. alpina and A. thaliana. Also, unlike in A. thaliana (Grierson and Schiefelbein, 2002), we observed short but not swollen root hairs in the Aakeu mutant. This suggests that mutations in the same gene lead to slightly different phenotypes in the two species. Similarly, the1 and anx1 mutants in A. thaliana exhibit collapsed, burst, and short root hairs (Duan et al., 2010), while the corresponding A. alpina mutants showed short but intact root hairs.

Our attempt to analyze the function of the AaSCN1 gene was challenging because we failed to generate transgenic lines in A. alpina to prove that the mutant phenotype is indeed caused by the mutation in the gene. However, interspecies experiments using A. thaliana provided evidence that the mutation in the AaSCN1 gene renders it non-functional, suggesting that the scn1 mutant phenotypes are very similar in the two species. This includes not only the morphological phenotype but also the ROS distribution in the mutants of both species (Carol et al., 2005). Moreover, we showed that YFP fused SCN1 from A. thaliana and A. alpina localized to the tip of growing root hairs.

In this work, we discovered interesting differences in the gene-to-phenotype relationship between A. alpina and A. thaliana and their further analysis may help to understand the evolution of the underlying gene regulatory networks. Moreover, we found root hair mutants without relevant SNPs in any of the known A. thaliana root hair genes. In principle, it is possible that in these cases, SNPs in the promoter regions cause transcriptional changes of the candidate genes. However, as this was rarely observed in EMS screens before, it is conceivable that the mutant phenotypes can be explained by genetic redundancies or mechanistic differences in root hair formation in A. alpina. In either case, the identification of the corresponding genes will likely reveal new players important for root hair development.