A family of pathogen-induced cysteine-rich transmembrane proteins is involved in plant disease resistance

Main conclusion Overexpression of pathogen-induced cysteine-rich transmembrane proteins (PCMs) in Arabidopsis thaliana enhances resistance against biotrophic pathogens and stimulates hypocotyl growth, suggesting a potential role for PCMs in connecting both biological processes. Abstract Plants possess a sophisticated immune system to protect themselves against pathogen attack. The defense hormone salicylic acid (SA) is an important player in the plant immune gene regulatory network. Using RNA-seq time series data of Arabidopsis thaliana leaves treated with SA, we identified a largely uncharacterized SA-responsive gene family of eight members that are all activated in response to various pathogens or their immune elicitors and encode small proteins with cysteine-rich transmembrane domains. Based on their nucleotide similarity and chromosomal position, the designated Pathogen-induced Cysteine-rich transMembrane protein (PCM) genes were subdivided into three subgroups consisting of PCM1-3 (subgroup I), PCM4-6 (subgroup II), and PCM7-8 (subgroup III). Of the PCM genes, only PCM4 (also known as PCC1) has previously been implicated in plant immunity. Transient expression assays in Nicotiana benthamiana indicated that most PCM proteins localize to the plasma membrane. Ectopic overexpression of the PCMs in Arabidopsis thaliana resulted in all eight cases in enhanced resistance against the biotrophic oomycete pathogen Hyaloperonospora arabidopsidis Noco2. Additionally, overexpression of PCM subgroup I genes conferred enhanced resistance to the hemi-biotrophic bacterial pathogen Pseudomonas syringae pv. tomato DC3000. The PCM-overexpression lines were found to be also affected in the expression of genes related to light signaling and development, and accordingly, PCM-overexpressing seedlings displayed elongated hypocotyl growth. These results point to a function of PCMs in both disease resistance and photomorphogenesis, connecting both biological processes, possibly via effects on membrane structure or activity of interacting proteins at the plasma membrane. Supplementary Information The online version contains supplementary material available at 10.1007/s00425-021-03606-3.


Introduction
In nature and in agriculture, plants are exposed to many different pathogenic microorganisms. To counter these threats, plants have evolved a complex immune system that can perceive pathogens and activate an appropriate response. These induced defense responses aim to fortify physical barriers against pathogen entry, such as callose (Luna et al. 2011). In addition, defensive compounds like secondary metabolites and pathogenesis-related proteins (PRs) accumulate, some of which have been demonstrated to possess in vitro antimicrobial activity and are associated with plant resistance (van Loon et al. 2006;Sels et al. 2008;Gamir et al. 2017). Plants can rely on a rich repertoire of defense compounds to combat different infecting agents. Still, many of the genes induced during pathogen infection have a so far unknown function, even though a role in defense can be expected for many of them.
The plant immune gene regulatory network that is activated in response to pathogen infection instructs which responses are expressed upon recognition of a specific invader. Conserved microbe-associated molecular patterns (MAMPs) and specific pathogen effectors can be perceived by matching receptors in the plant (Dodds and Rathjen 2010), which subsequently activate diverse downstream signaling cascades that involve elevated levels of reactive oxygen species and calcium signaling, the modification of enzymes, and changes in hormone levels (Boller and Felix 2009). The phytohormone salicylic acid (SA) plays a key role as signaling molecule in the regulation of plant immune responses that are primarily effective to fight biotrophic pathogens (Fu and Dong 2013). In SA-activated cells, the transcriptional cofactor NONEXPRESSOR OF PATHOGENESIS-RELATED GENES1 (NPR1) interacts with members of the TGA family of transcription factors, leading to transcriptional activation of different other transcription factors, like members of the WRKY family, and downstream SA-responsive defense genes (Tsuda and Somssich 2015). Microarray analysis of Arabidopsis thaliana (hereafter: Arabidopsis) plants expressing an NPR1-GR (glucocorticoid receptor) fusion protein (Wang et al. 2006) showed that several well-known SA-related genes, like those encoding PR proteins and WRKY transcription factors, were among the differentially expressed genes (DEGs) following SA treatment and dexamethasoneinduced nuclear localization of NPR1. Almost 20% of the 64 direct target genes regulated by NPR1 were described as having an unknown or uncharacterized function.
While the role of SA in regulating responses to pathogen infection is well established, it is also known to have a broader influence, regulating responses to abiotic stresses, such as cold, heat shock, drought, high salinity, UV radiation, and shade avoidance (Hayat et al. 2010;Nozue et al. 2018). SA also impacts plant growth by inhibiting auxin (growth hormone) signaling and contributes to developmental processes, such as flower formation. The latter is delayed in SA-deficient Arabidopsis genotypes (NahG transgenic lines; eds5 and sid2 mutants), suggesting an interplay of SA with photoperiod and autonomous (flowering) pathways (Martinez et al. 2004;Rivas-San Vicente and Plasencia 2011).
Even though the complete Arabidopsis genome has been known for nearly two decades (Arabidopsis Genome Initiative 2000), a large fraction of the protein-coding genes is still lacking a meaningful functional characterization (Niehaus et al. 2015). A common starting point for gene characterization is to reveal the conditions under which a gene is expressed. Transcriptome analysis has been extensively used to pinpoint genes that are active in specific tissue/cell types, at developmental stages or in response to different stimuli. Recently, several research groups utilized time series transcriptome experiments in the model plant Arabidopsis to gain insight into the topology of the gene regulatory network that is engaged under different conditions. These experiments provided a wealth of predictions regarding functional and regulatory roles of sets of genes that are differentially expressed in diverse situations (Krouk et al. 2010;Breeze et al. 2011;Bar-Joseph et al. 2012;Windram et al. 2012;Lewis et al. 2015;Coolen et al. 2016;Hickman et al. 2017). In our recent study, we applied whole transcriptome shotgun sequencing (RNA-seq) time series and found that approximately one-third of the Arabidopsis genome was differentially expressed in leaves upon treatment with SA over a 16-h time course, with changes in gene expression occurring in well-defined process-specific waves of induction or repression (Hickman et al. 2019).
Here, this SA-responsive gene set was analyzed with the comparative genomics tools OrthoMCL and JackHMMER, which identified homologous groups of largely uncharacterized genes that may play a role in SA-associated immunity. This integrated analysis categorized over a hundred groups of SA-responsive genes, including one group of eight genes encoding short proteins that share a predicted cysteine-rich transmembrane domain and are also responsive to various pathogens and immune elicitors. We therefore named them Pathogen-induced Cysteine-rich transMembrane proteins (PCMs). The PCMs are also present in the group of NPR1regulated direct target genes, mentioned above. Cysteinerich repeat proteins have been predicted to be involved in biotic and abiotic stress responses (Venancio and Aravind 2010). For one of the family members, PCC1/PCM4, a role as positive regulator of defense to the biotrophic pathogen Hyaloperonospora arabidopsidis has been demonstrated (Sauerbrunn and Schlaich 2004). For another family member, CYSTM3/PCM8, a role as negative regulator of salt Page 3 of 17 102 stress responses has been reported (Xu et al. 2019). Analysis of Arabidopsis PCM-overexpressing lines revealed that these proteins are positively involved in immunity against pathogens with a (hemi)biotrophic lifestyle. Furthermore, we expanded the potential scope of their function to a role in photomorphogenesis and hypocotyl development.

Plant material and cultivation conditions
Arabidopsis thaliana wild-type accession Col-0, mutant eds1-2 (Bartsch et al. 2006), triple mutant mlo2-5 mlo6-2 mlo12-1 (Consonni et al. 2006), hy5 hyh (Van Gelderen et al. 2018) and PCM-overexpression lines were used in this study. For whole-plant assays with pathogen infection and SA treatment, the seeds were stratified for 48 h in 0.1% agar at 4 °C prior to sowing them on river sand that was saturated with half-strength Hoagland nutrient solution containing 10 μM Sequestreen (Fe-ethylenediamide-di(0-hydroxyphenylacetic acid; Ciba-Geigy, Frankfurt, Germany) to supply iron nutrition. After 2 weeks of growing on sand in a closed container which provided 100% relative humidity (RH), the seedlings were transferred to 60-mL pots containing a soil:river sand mixture (12:5, v/v) that had been autoclaved twice for 1 h at 70% RH. Plants were cultivated under a 10-h day (75 μmol m −2 s −1 ) and 14-h night cycle at 21 °C. Plants were watered every other day and received modified half-strength Hoagland nutrient solution containing 10 μM Sequestreen once a week. To minimize within-chamber variation, all the trays, each containing a mixture of plant genotype or treatments, were randomized throughout the growth chamber once a week. For the hypocotyl elongation assay, seeds were surface-sterilized and sown on MS plates (8 g L −1 agar and 1 g L −1 Murashige and Skoog; Duchefa Biochemie B.V., Haarlem, The Netherlands). The seeds were stratified in the dark at 4 °C for 2-3 days before being moved to a climate chamber with long-day conditions (16 h light:8 h dark). After 7 days, the plates were photographed, and hypocotyl length was measured using ImageJ as described previously (De Wit et al. 2016).
The Arabidopsis PCM-overexpression lines were generated by amplifying the coding sequence of genes At2g32190 (PCM1), At2g32200 (PCM2), At2g32210 (PCM3), At3g22231 (PCM4/PCC1), At3g22235 (PCM5), At3g22240 (PCM6), At1g05340 (PCM7) and At1g56060 (PCM8/ ATCYSTM3) from accession Col-0. The PCM genes were part of a recent paper by Xu et al. (2018) and were named differently in the present study, as clarified in Suppl. Table S1. The primers used for cloning are also listed in Suppl. Table S1. The DNA sequence of the PCR fragments was verified and then cloned using Gateway® cloning (Invitrogen) in the pENTR vector, and subsequently in the pFAST-GO2 Gateway® (Shimada et al. 2010) compatible binary vector under control of the CaMV 35S promoter, followed by sequence verification. Binary vectors were transformed into Agrabacterium tumefaciens strain C58C1 containing pGV2260, which was used to transform accession Col-0 with the floral dip method (Clough and Bent 1998). Transformants were selected by growth on ½ MS plates containing DL-Phosphinothricin BASTA, and resistant T 1 seedlings were transplanted to soil for seed production. T 2 and T 3 lines were selected for single insertion of the transgenes using BASTA resistance. Experiments were performed using homozygous T 3 or T 4 seeds.

RNA-seq library preparation and sequencing
The experimental design of the RNA-seq time series experiment with SA-treated Arabidopsis leaves has been described previously (Hickman et al. 2019). In brief, the rosettes of 5-week-old Arabidopsis accession Col-0 plants were dipped into a solution containing 1 mM SA (Mallinckrodt Baker) and 0.015% (v/v) Silwet L77 (Van Meeuwen Chemicals BV), which was added as a surfactant. For mock treatments, plants were dipped into a solution containing 0.015% (v/v) Silwet L77. The sixth leaf (counted from the oldest to the youngest) was harvested from four individual SA-or mocktreated plants at each of the following time points post-treatment: 15 min, 30 min and 1, 1.5, 2, 3, 4, 5, 6, 7, 8, 10, 12 and 16 h. Total RNA was extracted using the RNeasy Mini Kit (Qiagen), including a DNase treatment step in accordance with the manufacturer's instructions. RNA-seq library preparation and sequencing was performed by UCLA Neuroscience Genomics Core (Los Angeles, CA, USA). Sequencing libraries were prepared using the Illumina TruSeq RNA Sample Prep Kit, and sequenced on the Illumina HiSeq 2000 platform with single read lengths of 50 bases.
For the comparison of the PCM1-OX, PCM5-OX and PCM7-OX lines with wild-type Col-0, two mature leaves (developmental leaf number six and seven) were harvested from two 5-week-old plants per genotype, resulting in two biological replicates. RNA-seq library preparation and sequencing was performed by the Utrecht Sequencing Facility (Utrecht, Netherlands). Sequencing libraries were prepared using the Illumina Truseq mRNA Stranded Sample Prep Kit, and sequenced on the Illumina NextSeq 500 platform with read lengths of 75 bases.

RNA-seq analysis
Quantification of gene expression from RNA-seq data was performed as described previously Hickman et al. 2017). Reads were mapped to the Arabidopsis genome (TAIR version 10) using TopHat version 2.0.4 (Trapnell et al. 2009) and aligned reads summarized over annotated gene models using HTseq-count (Anders et al. 2015). Genes that were significantly altered over time in response to SA in comparison to the mock treatment were identified using a generalized linear model implemented with the R statistical environment (www.r-proje ct. org). Genes that were differentially expressed between Col-0 and PCM1-OX, PCM5-OX, or PCM7-OX were identified using DESeq2 (Anders and Huber 2010;Love et al. 2014).

Identification of uncharacterized gene families
Protein sequences of the 630 SA-responsive DEGs with unknown/uncharacterized function (based on gene annotations retrieved from TAIR version 10 (retrieved in 2016) were run through OrthoMCL with default parameters (www. ortho mcl. org) (Li et al. 2003). JackHMMER (www. ebi. ac. uk/ Tools/ hmmer/ search/ jackh mmer) was then used to identify additional paralogs belonging to the groups identified with OrthoMCL. The phylogentic tree of PCM homologs was generated using PLAZA v4.0 (https:// bioin forma tics. psb. ugent. be/ plaza/) with the PCM1 gene as a query (Van Bel et al. 2018).

Determination of transcription factor binding motifs
Transcription factor-gene interactions were inferred from DAP-seq (DNA affinity purification sequencing) experiments, which provide the genome-wide binding profiles of in vitro expressed transcription factors (O'Malley et al. 2016). DAP-seq peaks for 349 Arabidopsis transcription factors with a FRiP (fraction of reads in peaks) score ≥ 5% were retrieved from the Plant Cistrome DB (O'Malley et al. 2016). DAP-seq peaks were used to infer representation of DNA-binding motifs in the promoters of the PCM genes. Motifs are grouped according to cognate transcription factor family.

Functional enrichment analysis
GO-term enrichment analysis on gene lists was performed using the GO term finder tool (Boyle et al. 2004). Where indicated, generic GO terms were removed from the analysis by limiting the maximum size of functional categories to 1500 genes.

Construction of YFP-tagged PCMs and visualization by confocal microscopy
For in planta localization experiments, cDNA extracted from Arabidopsis was used to amplify the coding sequences without the stop codon of PCMs using the primers listed in the Suppl. Table S1. The PCR products containing attB sequence were cloned into the Gateway pDONR221 vector, then the resulting entry vectors containing the PCM genes were recombined into the Gateway expression vector pB7WGY2, which contains the coding sequence of the Venus fluorescent protein (a variant of yellow fluorescent protein (YFP).
Competent cells of A. tumefaciens were transformed with the Gateway expression vector described in the previous paragraph made for protein localization. Transformed colonies were selected using the antibiotic resistance of the vector and rifampicin for which a resistance gene is present in A. tumefaciens. Single colonies were grown for 2 days at 28 °C in 20-mL LB medium under shaking conditions. After, the OD 600 was measured, the cells were pelleted and re-suspended to a final OD 600 of 0.5 with a ½ MS medium (Duchefa Biochemie) supplemented with 10 mM MES hydrate (Sigma-Aldrich), 20 g L −1 sucrose (Sigma-Aldrich), 200 µM acetosyringone (Sigma-Aldrich) at pH 5.6 and incubated in darkness for at least 1 h. The solutions were used to agroinfiltrated the abaxial side of 4-5-week-old Nicotiana benthamiana leaves using a 1-mL syringe. The plants were left to grow in normal light conditions and after 2 days leaf sections were taken from agroinfiltrated regions and visualized via confocal microscopy.
Microscopy was performed using a Zeiss LM 700 (Zeiss, Jena, Germany) confocal laser-scanning microscope. Fresh leaf material was prepared on a glass slide with cover slip. Excitation of YFP, plasma membrane FM4-64 dye (Sigma-Aldrich) and autofluorescence of chlorophyll were done at 488 nm. Light emission of YFP was detected at 493-550 nm and the red signal for the FM4-64 dye at 644-800 nm. Analyses of the images were performed with ZEN lite (blue edition).

Pathogen cultivation and bioassays
Hyaloperonospora arabidopsidis isolate Noco2 (Hpa Noco2) spores were harvested from infected (eds1-2 mutant) plants, eluted through Miracloth, and diluted in water to 50 spores µL −1 . For the disease bioassay, the leaves of 5-week-old plants were spray-inoculated with this spore suspension. Plants were subsequently placed at 100% RH, under short day conditions (9 h light: 15 h dark) at 16 °C. After 9 days, the spores from eight individual rosette plants were harvested in 5 mL of water and the number of spores per milligram of plant tissue (fresh weight of aerial parts) was counted using a light microscope. Spore counts in the wild-type, mutant and overexpression lines were compared using analysis of variance (ANOVA) followed by Tukey's multiple comparison tests.
Pseudomonas syringae pv. tomato (Pto) DC3000 was cultured in King's B medium supplemented with 50 mg L −1 rifampicine at 28 °C overnight. Bacteria were collected by centrifugation for 10 min at 5000 g and re-suspended in 10 mM MgSO 4 . The suspension was adjusted to OD 600 = 0.0005 and pressure infiltrated into three mature leaves of 5-week-old plants with a needleless syringe. After 3 days, leaf discs of 5-mm diameter were harvested from two inoculated leaves per plant, representing a single biological replicate. Eight biological replicates were harvested for each genotype. Subsequently, 500 µL of 10 mM MgSO 4 was added to the leaf discs, after which they were ground thoroughly with metal beads using a TissueLyser (Qiagen). Serial ten-fold dilutions were made in 10 mM MgSO 4 , and 30 µL aliquots plated onto KB agar plates containing 50 mg mL −1 rifampicin. After 48 h of incubation at 28 °C, bacterial colonies were counted. Statistical analyses were performed using ANOVA followed by Tukey's multiple comparison test for means of log 10 -transformed colony counts.
For powdery mildew assays, Arabidopsis plants were inoculated with powdery mildew (Golovinomyces orontii) at roughly 2.5 cm rosette size (radius) at 4-5 weeks after germination. G. orontii is adapted to infection of Arabidopsis (Kuhn et al. 2016) and was cultivated on susceptible eds1-2 plants. Inoculation was conducted by leaf-toleaf transfer of conidiospores. Leaves from five individual plants were collected at 48 h post inoculation and bleached in 80% ethanol at room temperature for at least 2 days. Prior to microscopic analysis, fungal structures were stained by submerging the leaves in Coomassie staining solution (100% v/v ethanol acid, 0.6% w/v Coomassie blue R-250; Carl Roth, Karlsruhe, Germany) twice for 15-30 s and shortly washed in tap water thereafter. The samples were analyzed with an Axiophot microscope (Carl Zeiss AG). The fungal penetration rate was determined as the percentage of spores successfully developing secondary hyphae over all spores that attempted penetration, visible by an appressorium (Haustorium index). Macroscopic pictures of G. orontii-infected plants were taken at 12 days post inoculation with a Coolpix P600 camera (Nikon, Tokyo, Japan). Susceptible Col-0 and the fully resistant mlo2-5 mlo6-2 mlo12-1 triple mutant (Consonni et al. 2006) served as positive and negative controls for infection by G. orontii, respectively. Haustorium index in the mutant and overexpression lines were compared using ANOVA followed by Tukey's multiple comparison test.

Analysis of uncharacterized SA-responsive genes identifies a family of cysteine-rich transmembrane proteins
Recently, we used high-throughput RNA-seq analysis to profile genome-wide changes in mRNA abundance in Arabidopsis leaves following treatment with SA over a 16-h period. Analysis of these transcriptome data identified 9524 genes that were differentially expressed between mock-and SA-treated leaves (Hickman et al. 2019). Subsequent investigation of functional annotations associated with these differentially expressed genes (DEGs) revealed that 630 of these genes encode proteins of unknown or uncharacterized function. Because of the central role of SA in defense against pathogen infection, we hypothesized that among these genes would be genes with undiscovered roles in plant immunity. To simplify the analysis and functional interpretation of these uncharacterized genes, we first divided them in groups based on amino acid sequence similarity. To achieve this, we used OrthoMCL (Li et al. 2003), which is a tool for identifying homologous relationships between sets of proteins. This analysis resulted in a division of 101 groups of putative homologs, each comprising between two and nine members (Suppl. Data Set S1; Fig. 1a). Because we were specifically interested in genes that are involved in defense against pathogens, we analyzed gene expression behavior, using available gene expression data from Genevestigator (http:// www. genev estig ator. ethz. ch/) (Hruz et al. 2008). This pointed to a group of seven genes that were highly induced by a variety of immune elicitors and pathogens (Fig. 2a) and that were responsive to SA in our RNA-seq experiment (Fig. 2b). Their increased transcription was confirmed by an independent experiment in which genome-wide transcriptional changes as induced by exogenous SA treatment were analyzed using microarrays (Suppl. Fig. S2; Pajerowska-Mukhtar et al. 2012).
To identify all possible paralogs (including remote paralogs), the seven genes were used as queries in JackHMMER (Finn et al. 2015) (www. ebi. ac. uk/ Tools/ hmmer/ search/ jackh mmer). JackHMMER is a highly sensitive homology detection tool that can identify shared protein domains among matched sequences, as defined according to Pfam domains (Finn et al. 2015). This analysis led to the prediction of seven additional paralogs (Fig. 1b). Next, we quantified the degree of nucleotide sequence identity between the 14 proteins by constructing a nucleotide sequence identity First, SA-induced DEGs were grouped by DNA similarity using OrthoMCL. One pathogen-responsive group was subjected for further analysis using JackHMMER, followed by pair-wise similarity clustering, revealing a distinct family of eight homologous PCM genes. b DNA similarity matrix showing the 14 genes identified by the JackHMMER search. Red and blue indicate high and low similarity, respectively. Unsupervised hierarchical clustering identified a distinct group of PCM genes with high DNA similarity. c Amino acid sequence alignment of the eight PCMs. The conserved cysteine-rich transmembrane domain (CYSTM) is highlighted by asterisks. d The locations of the eight PCM genes on the Arabidopsis chromosomes (Chr1 to Chr5). The different color gene names reflect PCM gene distribution across chromosomes Page 7 of 17 102 matrix (Fig. 1b), which was followed by unsupervised clustering of the similarity matrix, leading to the identification of a distinct family of eight small genes (< 82 amino acids) with high nucleotide sequence identity (> 38%). All of the seven originally selected genes of unassigned function belong to this group, but also included now PCC1 (PCM4 in Fig. 1), which has a reported role in defense and is regulated by the circadian clock (Sauerbrunn and Schlaich 2004). One other member, CYSTM3 (PCM8 in Fig. 1), has very recently also been characterized and shown to negatively influence salt stress resistance (Xu et al. 2019). Suppl. Table S1 lists all of the PCM genes with their Arabidopsis Genome Initiative (AGI) number and alternative name. The genes in this family all encode short proteins (71-82 amino acids) with a conserved cysteine-rich transmembrane (CYSTM) domain, as predicted by the JackHMMER analyses (Fig. 1c, CYSTM domain indicated by asterisks). To reflect their regulation and enrichment for cysteine residues in the encoded proteins, this eight-member gene family was named Pathogen-induced Cysteine-rich transMembrane proteins (PCMs). The PCM gene family contains two distinct gene clusters; the PCM1, PCM2 and PCM3 genes are situated in tandem on Arabidopsis chromosome 2, while PCM4 (PCC1), PCM5 and PCM6 are tandemly arrayed on chromosome 3 (Fig. 1d). Furthermore, PCM7 and PCM8 (CYSTM3) are positioned at distant locations on chromosome 1. The expression behavior of the eight PCM genes is broadly along the lines of the three subgroups, showing overlap but also differences with members of the other subgroups ( Fig. 2a and b). This is in accordance with varying overrepresentation of different transcription factor binding DNA motifs in the promoters of the eight PCM genes (Fig. 2c). The remainder of this paper explores the significance of the PCM protein family and its three subgroups in plant immunity and development.

Subcellular localization of PCMs
The characteristic CYSTM domain that resides in the PCM protein family is encoded by a total of 98 genes across 33 plant species (Suppl. Fig. S1). Transmembrane domains enable protein functions across membranes (Sharpe et al. 2010;Luschnig and Vert 2014) and are often conserved across kingdoms when the respective protein has a specialized function (e.g., photoreceptors in eyes of mammals and insects) (Fischer et al. 2004). To begin to characterize the PCMs, we determined their subcellular localization by fusing Venus YFP to the C-terminus of all eight PCM proteins and expressing these fusion proteins under the control of the constitutively active cauliflower mosaic virus (CaMV) 35S promoter. Expression of the empty vector (EV-YFP, resulting in free YFP) served as a control. The dye FM 4-64 was used as a membrane marker. The fusion proteins were transiently expressed in N. benthamiana leaves by infiltration with A. tumefaciens carrying the different constructs. Confocal microscopy analysis showed that for five of the PCM fusion proteins, namely PCM1, PCM2, PCM3, PCM4 and PCM5, the YFP signal overlapped with the fluorescent signal of the plasma membrane-localized FM 4-64 dye (Fig. 3). This suggests that these PCM proteins are plasma membrane localized. By contrast, in case of YFP-tagged PCM6, PCM7 and PCM8, the YFP signals were detected prevalently in the cytoplasm and the nucleus (Fig. 3), which could either be indicative of a non-membrane localization of these proteins or reflect undesired cleavage of the YFP label.

PCM coexpression analysis points to specificity in PCM function
Because genes with related biological functions often have similar expression patterns, a well-established method to investigate gene function is the construction and analysis of gene coexpression networks (Vandepoele et al. 2009). Using the eight PCM genes as query we generated PCM coexpression networks using publicly available microarray and RNA-seq datasets with the ATTED-II coexpression tool (Obayashi et al. 2017) (Fig. 4). The PCM coexpression network was enriched for genes associated with defense responses (P < 0.01; hypergeometric test) and included known defense-related genes, such as LURP1, ACD6, RLP36, NTL6, NAC61, NAC90, ZFAR1, PDR12, WRKY75 and MPK11, suggesting a role for the PCM protein family in plant defense. Within the PCM coexpression network, coexpression neighborhoods of members of the three PCM subgroups (Fig. 2) overlap. Interestingly, the coexpression neighborhood occupied by subgroup II (PCM4, PCM5 and PCM6) was distinct from that of all other PCM genes. Also, PCM7 was part of a relatively isolated coexpression subnetwork. On the other hand, PCM8 shared its coexpression neighborhood to a large extent with that of subgroup I (PCM1, PCM2 and PCM3). In sum, our coexpression network analysis suggests a role for all eight PCM genes in plant defense, but also highlights subnetworks, suggesting functional diversification and/or differential regulation of the PCM subgroups. This notion is further supported by the distinct gene expression behavior of the different PCM subgroups after treatment with pathogens or exogenous SA (Fig. 2a, b, and Suppl. Fig. S2) and the presence of different transcription factor binding sites in the promoters of the PCM genes (Fig. 2c).

PCM-overexpressing lines show enhanced resistance to (hemi-)biotrophic pathogens
To investigate the hypothesis that members of the PCM protein family play a role in plant immunity, transgenic Arabidopsis lines expressing the individual PCM genes under the control of the CaMV 35S promoter were generated. The transgenic PCM-overexpression (PCM-OX) lines were of unaltered size and did not show any obvious developmental abnormalities (Suppl. Fig. S3). RNA-seq analysis (see below) confirmed the overexpression status of the PCM-OX lines for genes PCM1 and PCM7, but not for PCM5 whose overexpression levels might have remained below the thresholds of statistical analysis (Suppl. Data Set S2). Because the PCM gene family responded to exogenous SA treatment ( Fig. 2b and Suppl. Fig. S2), the PCM-OX lines were screened for an altered level of resistance to two pathogens that are controlled by SA-dependent defenses: the obligate biotrophic oomycete H. arabidopsidis Noco2 (Hpa Noco2) and the hemi-biotrophic bacterium P. syringae pv. tomato DC3000 (Pto DC3000). For both assays, the performance of 5-week-old PCM-OX lines was compared to that of the wild-type (Col-0) and the super susceptible mutant eds1-2 of the same age. Apart from PCM6, overexpression of all other PCM genes led to reduced Hpa Noco2 spore formation when compared to wild-type plants (Fig. 5a). Pto DC3000 propagation was significantly decreased in the PCM1-OX, PCM2-OX, and PCM3-OX lines but not in the other lines (Fig. 5b). These findings suggest that the vast majority of PCM family members is positively involved in host defense against Hpa Noco2, while a protective effect against Pto DC3000 is only evident for the subgroup I of the PCM protein family comprising PCM1, PCM2, and PCM3.

Transcriptome analysis of PCM-OX lines reveals no upregulation of typical immune responses
To gain insight into the mechanisms underlying the enhanced disease resistance phenotype obtained by overexpression of the PCM genes, we analyzed the transcriptome of three PCM-OX lines, each representing a member of the three PCM subgroups: PCM1-OX (subgroup I), PCM5-OX (subgroup II), and PCM7-OX (subgroup III). These PCM members of the respective subgroups were representing the small differences between the different subgroups based on DNA and amino acid composition (Fig. 1b and c), chromosome position (Fig. 1d), induction by pathogens (Fig. 2a) and SA (Fig. 2b), enrichment in transcription factor binding motifs (Fig. 2c), inclusion in gene coexpression networks (Fig. 4), and effects on disease resistance (Fig. 5). RNA-seq analysis was performed on leaf tissue harvested from 5-week-old, non-treated plants. Differential expression analysis revealed that in the PCM1-OX, PCM5-OX and PCM7-OX lines 934, 873, and 515 genes, respectively, were differentially expressed in comparison to wild-type Col-0 plants (P < 0.05, fold change > 2) (Suppl. Data Set S2). Among the DEGs there were PCM1 in the PCM1-OX line and PCM7 in the PCM7-OX line, each showing a twofold log increase in transcript abundance. Notably, the list of DEGs comprised no other PCM gene in any of the overexpression lines, indicating that there is no compensatory regulation of other family members in this situation. There was considerable overlap between the expression profiles of the three PCM-OX lines (Fig. 6a). Of all DEGs, 27% were upregulated or downregulated in all three lines (in the same direction), whereas 44% were specifically up-or downregulated in a single overexpression line (Fig. 6b). More genes were downregulated (60%) than upregulated (40%).
The overlapping 131 upregulated DEGs shared by all three PCM-OX lines were not enriched for typical immunityrelated functions (Fig. 7 and Suppl. Data Set S3). Instead, the term 'circadian rhythm' was the most significantly enriched specific category, with additional enriched terms including 'regulation of multicellular organismal development' 'plant cell wall loosening', and 'response to red or far-red light'. The shared 214 downregulated DEGs by all three PCM-OX lines were associated with functional categories, such as 'rRNA processing', 'response to cytokinin' and There was also no enrichment of purely immunity-related categories DEGs that were specifically up-or downregulated in any of the PCM1-OX, PCM5-OX or PCM7-OX lines (Suppl. Data Sets S2 and S3). However, general terms like 'response to hormone' were overrepresented in up-and/or downregulated DEGs of different lines. 'Glucosinolate process' was detected particularly in the upregulated DEGs of the PCM7-OX line, while 'nucleolus' was enriched in both the up-and downregulated DEGs of PCM7-OX, as well as in downregulated DEGs of PCM5-OX (Suppl. Data Set S3). Overall, based on these data, we hypothesize that pathogeninduced PCM production contributes to an increased level of defense through an impact on developmental processes in the cells that may affect pathogen performance.

Involvement of PCMs in hypocotyl elongation
There was no clear link to plant immunity among the genes differentially expressed in the three PCM-OX lines assayed, while the association with developmental processes and light responses was obvious (Fig. 7). In all three PCM-OX lines, the HY5 and HYH genes, which are master regulators of light signaling (Van Gelderen et al. 2018) and also respond Fig. 5 Overexpression of PCMs enhances resistance to Hyaloperonospora arabidopsidis Noco2 and Pseudomonas syringae pv. tomato DC3000. a Quantification of Hpa Noco2 sporulation on 5-week-old wild-type (Col-0), eds1-2 and transgenic lines constitutively overexpressing individual PCM genes under the control of the CaMV 35S promoter (PCM-OX) at 10 days post inoculation (dpi) by spraying (n = 9-12). b Bacterial multiplication of Pto DC3000 in wild-type (Col-0), eds1-2 and PCM-OX lines at 3 dpi by pressure infiltration (n = 8). Means ± SE (error bars) are shown. Letters denote significant differences between genotypes (one-way ANOVA, Tukey's post-hoc test, P < 0.05) to pathogen infection (Genevestigator data) were upregulated (Suppl. Data Set S2). This prompted us to investigate morphogenic responsiveness of the PCM-OX lines. In shade-avoiding plants, such as Arabidopsis, perception of far-red light triggers morphological adaptations, such as elongation of the hypocotyl and petioles, to reach for better quality light (Ballaré 2014). The hy5 hyh double mutant, which is affected in HY5 and its closely related HY5 homolog (HYH), displays such elongated hypocotyl growth compared to wild-type plants when cultivated in white light (Van Gelderen et al. 2018). Unexpectedly, the hypocotyl length of the PCM1-OX, PCM5-OX, and PCM7-OX lines was also greater than that of the wild type, and the hypocotyl of PCM7-OX was even of the same size as that of the positive control, the hy5 hyh double mutant (Fig. 8). This points to a role for PCMs in modulating both growth and development. Possibly, the PCMs affect HY5 protein activity or stability, which is compensated by an enhanced expression level of the HY5 gene. Altogether, our data suggest dual roles for PCMs in defense and in photomorphogenesis.

Discussion
Despite more than two decades of research efforts focused on the model plant Arabidopsis, a significant fraction (over 13%) of genes found in this plant are not characterized to any extent (Luhua et al. 2013;Niehaus et al. 2015). Our analysis of SA-responsive genes in Arabidopsis leaves revealed that 630 genes encode proteins of unknown function. Using a protein homology search, we grouped these uncharacterized genes into 101 groups of paralogs that likely encode proteins with similar functions ( Fig. 1a; Suppl. Data Set S1). We validated whether such an approach could aid the functional annotation of groups of unknown genes. Therefore, we selected and further characterized a family of eight pathogen-induced cysteine-rich transmembrane proteins (PCMs). The PCM genes formed three subgroups, based on their nucleotide similarity, amino acid similarity, and chromosomal position (Fig. 1b-d). The expression profiles of the PCM members under different biotic stress conditions and SA treatment broadly followed that of the three subgroups, showing some overlap but also differences between the subgroups (Figs. 2a and b). This is in agreement with the commonalities and dissimilarities in transcription factorbinding sites detected in the promotor regions of the eight PCM genes (Fig. 2c), and the overlap or isolation of the coexpression networks of the PCM genes (Fig. 4).
Given the complexity of the PCM family and its overlap in coexpressed genes, we expected functional redundancy between members and thus resorted to using overexpression lines rather than knockout mutants for the functional analysis of this protein family. Overexpression of one PCM member of each of the three subgroups (PCM1-OX, PCM5-OX and PCM7-OX) revealed 27% overlap of all DEGs in all three lines, and 44% of all DEGs were specifically regulated in only one of the lines (Fig. 6). A function for the PCMs in defense was evidenced by the enhanced resistance of PCMoverexpression lines to the biotrophic pathogens Hpa Noco2 and Pto DC3000 (Fig. 5). Moreover, the overexpression of PCMs resulted in differential expression of genes related to light and development (Fig. 7, Suppl. Data Sets S2 and S3), and elongated hypocotyl growth of seedlings (Fig. 8), suggesting an additional role for PCMs in photomorphogenesis. Though we only used single PCM-OX lines in our sets of experiments, the shared phenotypes regarding DEGs, pathogen resistance and photomorphogenesis of the different tested PCM-OX lines suggest that the detected effects are authentic consequences of PCM function and do not result from unintentional chromosomal rearrangements at the location where the transgenes (PCMs) were inserted.

Membrane association of CYSTM domain-containing PCMs
The PCMs are small proteins (< 84 amino acids) that contain a predicted cysteine-rich transmembrane C-terminus domain (CYSTM), which is a rare domain, but highly conserved among eukaryotic organisms. CYSTM domain-containing proteins are present in diverse species, including Arabidopsis, Caenorhabditis elegans, Candida albicans, Homo sapiens, Mus musculus, Oryza sativa, Saccharomysces cerevisae and Zea mays (Venancio and Aravind 2010). The molecular mechanism by which the CYSTM module functions is not clear yet, but the proteins appear to play a role in stress tolerance, for example, by altering the redox potential of membranes, thereby quenching radical species to protect the plant, or by affecting membrane-associated protein functions (Kuramata et al. 2009;Venancio and Aravind 2010).
The conserved cysteines may serve to interact with a ligand, e.g. other PCMs, which could result in homo-or heterodimerization as shown in yeast expression systems for several Arabidopsis PCM family members (Mir and León 2014;Xu et al. 2018). However, the PCMs can potentially also interact with other protein partners, as shown for PCM4/ PCC1, which interacts with its N-terminal part (cytoplasmfaced, non-CYSTM containing) with the subunit 5 of the COP9 signalosome at the plasma membrane. This may lead to post-translational control of multiple protein targets involved in diverse biological processes, such as light signaling, development, and immunity (Mir et al. 2013;Mir and León 2014).
We experimentally confirmed a tight association of PCM1-PCM5 with the cell periphery and the fluorescent FM4-64 marker (Fig. 3), suggesting that these proteins are anchored to the plasma membrane. This localization could potentially promote a change in local lipid composition, as shown for PCM4/PCC1 (Mir et al. 2013), and also affect the membrane structure. This notion is supported by changes in gene expression observed in the PCM-OX lines, which highlighted enrichment in GO terms related to 'response to lipid', cell wall modification', and 'regulation of development' (Fig. 7). Moreover, membrane alterations may block the invasion of intracellular pathogens like Hpa (Fig. 5a) that form an intricate interface with the host membrane. There . Means ± SE (error bars) are shown. Letters denote significant differences between genotypes (one-way ANOVA, Tukey's posthoc test, P < 0.05). Inset: representative pictures of 7-day-old seedlings. The experiment was repeated twice with similar results may also be consequences for membrane permeability or activity of (defense regulatory) proteins associated with the plasma membrane. While plasma membrane localization of PCM1-PCM5 was supported by experiments using YFPtagged PCMs in transiently transformed N. benthamiana leaves, this was not the case for PCM6-PCM8 (Fig. 3). The latter finding is consistent with a recent study by Xu et al. (2018) who also found cytoplasmic localization of these proteins using the same study system. However, these authors also reported cytoplasmic localization for PCM1, PCM2, PCM3 and PCM5, for which we detected solely plasma membrane localization, which is in line with the expectations based on the presence of the CYSTM domain (Venancio and Aravind 2010) and early reports on PCC1/PCM4 (Mir and León 2014). At this point, we cannot exclude that the nucleo-cytoplasmic localization of PCM6-PCM8 detected by us and others (Xu et al. 2018) is due to degradation of the PCM-YFP fusion protein in this experimental setup. Alternative experimental approaches, such as biochemical analyses, will thus be required to corroborate the subcellular localization of these proteins.

The function of CYSTM domain-containing PCMs in plant defense
The PCM4 gene is also known as PCC1 and has previously been identified as an early-activated gene upon infection with the bacterial pathogen Pto carrying the avirulence gene AvrRpt2 and to be controlled by the circadian clock (Sauerbrunn and Schlaich 2004). Microarray analysis of npr1-1 plants revealed that in addition to several pathogenesis-related (PR) genes, the expressions of PCC1/PCM4 and PCM6 were affected in this mutant (Wang et al. 2006). Later, PCC1/PCM4 was identified to be induced by UV-C light in an SA-dependent manner, potentially playing a role as activator of stress-stimulated flowering in Arabidopsis (Segarra et al. 2010). Transgenic plants carrying the β-glucuronidase (GUS) reporter gene showed that the expression of PCC1/ PCM4 in the seedling stage was confined to the root vasculature and the stomatal guard cells of cotyledons, but spread to the petioles and the whole limb of fully expanded leaves (Mir et al. 2013). PCC1/PCM4-OX lines showed enhanced resistance to Hpa (Sauerbrunn and Schlaich 2004; Fig. 5a), while RNAi plants were more susceptible to the hemi-biotrophic oomycete pathogen Phytophthora brassicae and more resistant to the necrotrophic fungal pathogen Botrytis cinerea when compared with wild-type plants (Mir et al. 2013). We confirmed that PCC1/PCM4-overexpression lines are resistant to Hpa and extend this finding to additional PCMs: Overexpression of PCM1, PCM2, PCM3, PCM5, PCM7, and PCM8 also provided protection against Hpa infection (Fig. 5a). This points to a common underlying defense mechanism that is activated by the PCMs, which might be related to an altered membrane environment as we discussed in the previous paragraph. This mechanism may also be responsible for the enhanced protection against Pto infection that we observed by overexpressing PCM1, PCM2, and PCM3 (Fig. 5b). The lack of effect on Pto of the other PCM-OX lines, however, also points to divergent effects of the different PCMs, which is corroborated by the partly distinct DEG sets of the PCM1-, PCM5-, and PCM7-OX lines (Fig. 6b). We also assayed the PCM1, PCM5 and PCM7 overexpressors for resistance to the biotrophic powdery mildew fungus Golovinomyces orontii but found that these lines displayed the same level of disease development (haustorium formation and macroscopic symptoms) as the wild type, whereas the triple mutant mlo2 mlo6 mlo12, serving as a positive control, was highly resistant (Suppl. Fig.  S4). It may be that the protection mechanism provided by the PCMs is not effective against this pathogen species, but it may also be that the strain that was used in the bioassays was so virulent that it could have overcome any quantitative resistance accomplished by PCM overexpression.
Notably, cultivated PCM-OX lines did not show any morphological abnormalities, such as dwarfism, and the RNAseq data of the PCM1-, PCM5-, and PCM7-OX lines did not reveal any evidence for the constitutive expression of typical defense-related genes (such as PR genes) that would explain the enhanced disease resistance of these plant. In the future, it will be of interest to elucidate the yet unrecognized mechanisms that contribute to this phenotype. Conditioned by the antagonistic interplay of defense-associated phytohormones (Leon-Reyes et al. 2010), plants with enhanced resistance to biotrophic pathogens often show enhanced susceptibility to necrotrophic pathogens. For instance, infection with hemi-biotrophic Pseudomonas syringae, which induces SAmediated defense, rendered plants more susceptible to the necrotrophic pathogen Alternaria brassicicola by suppression of the jasmonate (JA) signaling pathway (Spoel et al. 2007). It will thus be also interesting to explore how the PCM-OX lines perform upon challenge with necrotrophic pathogens.

Interplay between immunity and photomorphogenesis
Our transcriptome data revealed that the PCM1-, PCM5-, and PCM7-OX lines were primarily enriched for genes and biological functions related to circadian rhythm, light signaling, and growth and development ( Fig. 7 and Suppl. Data Set S2). The PCM4/PCC1 gene had previously been reported to respond to circadian rhythm and UV-C light, and to influence stress-induced flowering (Segarra et al. 2010). Here, we show that the PCM1-, PCM5-, and PCM7-OX lines exhibit elongated hypocotyl growth compared to wild-type plants (Fig. 8). This phenotype is shared with the hy5 hyh double mutant, suggesting that the PCMs promote photomorphogenesis. Several studies have addressed the connection between plant defense and light signaling; e.g. UV-C induces SA-dependent defenses, and high levels of far-red light (as in shade) repress defense responses to both pathogen and insects, as reviewed by Ballaré (2014). A recent paper by Nozue et al. (2018) reported that SA pathway genes are key components of shade avoidance; PCM4 and PCM5 were part of the downregulated gene set by high farred levels, and showed an altered expression level in shade avoidance syndrome mutants. Therefore, a double role for the PCMs in defense and photomorphogenesis is not unexpected. How the PCMs accomplish this dual function is not clear yet. Like discussed earlier, the PCMs might influence membrane structure and activity of proteins that reside in the membrane or that bind to PCMs, like the subunit 5 of the COP9 signalosome (Mir et al. 2013;Mir and León 2014). These diverse effects may independently influence defense and photomorphogenesis, but an interdependence between the two biological processes, where one is a consequence of the other, is also a possibility.
In conclusion, our approach led to the identification of the family of PCM proteins that carry the distinctive CYSTM module, and which have a broad biological impact on plant performance, as shown by enhanced protection against biotrophic pathogens and increased hypocotyl growth in PCM-OX lines. We elucidated some molecular effects of the PCMs by showing that the majority of the PCM members localize to the plasma membrane, that the PCM genes are responsive to SA and pathogen challenge, and that overexpression of PCMs leads to the induction of genes associated with light responses and development, but not to typical defense-associated responses.
Author contribution statement MCVV and SCMVW conceived and designed the research. MPM, RH, MCVV, NMN, and AR conducted the experiments. MPM, RH, RP and MCVV analyzed the data. MPM, RH, CMJP and SCMVW wrote the manuscript. All authors read and approved the manuscript.
Funding CAPES Foundation, Ministry of Education of Brazil (to MPM), Netherlands Organization for Scientific Research through the Dutch Technology Foundation (VIDI 11281 to SCMVW and VENI 13682 to RH), European Research Council (Grant 269072 to CMJP).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.