An Arabidopsis downy mildew non-RxLR effector suppresses induced plant cell death to promote biotroph infection

Abstract Our understanding of obligate biotrophic pathogens is limited by lack of knowledge concerning the molecular function of virulence factors. We established Arabidopsis host-induced gene silencing (HIGS) to explore gene functions of Hyaloperonospora arabidopsidis, including CYSTEINE-RICH PROTEIN (HaCR)1, a potential secreted effector gene of this obligate biotrophic pathogen. HaCR1 HIGS resulted in H. arabidopsidis-induced local plant cell death and reduced pathogen reproduction. We functionally characterized HaCR1 by ectopic expression in Nicotiana benthamiana. HaCR1 was capable of inhibiting effector-triggered plant cell death. Consistent with this, HaCR1 expression in N. benthamiana led to stronger disease symptoms caused by the hemibiotrophic oomycete pathogen Phytophthora capsici, but reduced disease symptoms caused by the necrotrophic fungal pathogen Botrytis cinerea. Expressing HaCR1 in transgenic Arabidopsis confirmed higher susceptibility to H. arabidopsidis and to the bacterial hemibiotrophic pathogen Pseudomonas syringae. Increased H. arabidopsidis infection was in accordance with reduced PATHOGENESIS RELATED (PR)1 induction. Expression of full-length HaCR1 was required for its function, which was lost if the signal peptide was deleted, suggesting its site of action in the plant apoplast. This study provides phytopathological and molecular evidence for the importance of this widespread, but largely unexplored class of non-RxLR effectors in biotrophic oomycetes.


Introduction
Oomycetes include some notorious plant pathogens that severely reduce global crop yield and cause enormous economic loss every year. To date, oomycete pest management relies on inbreeding of RESISTANCE (R) genes and chemical plant protection. In this context, the occurrence of new virulent pathogen genotypes that overcome R gene-mediated resistance or chemical crop protection jeopardizes food security (Fry, 2008;Cohen et al., 2015;Delmas et al., 2017). Thus, there is an urgent need for developing innovative, sustainable strategies to control oomycete pests. However, a lack of understanding of pathogen virulence at the molecular level restricts this goal.
In oomycetes, classical forward or reverse genetics approaches remain challenging due to di-or polyploidy, and due to the fact that many oomycetes are obligate biotrophs, like the downy mildew pathogen Hyaloperonospora arabidopsidis infecting the model plant Arabidopsis (Coates and Beynon, 2010). Obligate biotrophs are impossible to grow in axenic culture despite numerous attempts (McDowell, 2014), and thus genetic transformation of these organisms is not achievable. Alternative approaches to investigating the function of pathogen effectors and other types of virulence genes in obligate biotrophs, which do not rely on pathogen transformation, are therefore highly appreciated. Hyaloperonospora arabidopsidis is one of the most used model pathogens to investigate Arabidopsis innate immune response to obligate biotrophs and was ranked the second most important oomycete pathogen by researchers in terms of scientific and economic relevance (Kamoun et al., 2015). It is highly adapted and specialized to its sole natural host plant, Arabidopsis, and infection frequently occurs in wild Arabidopsis plants (McDowell, 2014;Agler et al., 2016).
Genome sequencing of H. arabidopsidis uncovered a large repertoire of over 100 putative effector genes suggesting an extensive resource to suppress plant immunity (Baxter et al., 2010) and to enable host cell reprogramming for pathogen accommodation and propagation (Thordal-Christensen et al., 2018). Oomycete effectors are typically classified by sequence features into RxLRs, Crinklers, necrosis-inducing like proteins, elicitins, and if no further sequence homology is apparent, cysteine-rich (CR) proteins (Cabral et al., 2011). Current research in oomycete effectors focuses on RxLRs that are typically translocated into host cells and are relatively easy to predict in silico (Anderson et al., 2015). Hyaloperonospora arabidopsidis probably employs RxLRs to modulate plant immunity, too (Fabro et al., 2011;Pel et al., 2014). Nevertheless, a defined molecular function of only just a few oomycete effectors has been reported, mainly through ectopic expression in planta (Caillaud et al., 2013;Wirthmueller et al., 2018). In addition, non-RxLR effectors presumably contribute to virulence, as well. Nevertheless, H. arabidopsidis non-RxLR CR protein effectors remain functionally uncharacterized, despite the fact that they comprise some of the most highly expressed H. arabidopsidis genes during infection (Cabral et al., 2011;Asai et al., 2014).
Artificial expression of double-stranded RNA (dsRNA) in host plants can lead to silencing of complementary genes in their pathogens and pests, a strategy known as host-induced gene silencing (HIGS) (Baum et al., 2007;Mao et al., 2007;Koch and Kogel, 2014). Indeed, HIGS is a powerful method of choice for reverse genetics in plant-associated organisms with no transformation protocols available, such as root knot nematodes, mycorrhizal fungi and biotrophic pathogens, like powdery mildew and rust fungi (Nowara et al., 2010;Helber et al., 2011;Pliego et al., 2013;Dinh et al., 2014;Yin and Hulbert, 2018). Regarding oomycetes, an initial HIGS approach in Arabidopsis failed to knockdown gene expression of Phytophthora parasitica although HIGS small interfering RNA accumulated in the plant (Zhang et al., 2011). Nevertheless, HIGS was successfully introduced in Solanum tuberosum (potato) against the hemibiotrophic pathogen Phytophthora infestans and in lettuce against the downy mildew pathogen Bremia lactucae, conferring plant disease resistance (Govindarajulu et al., 2015;Jahan et al., 2015). Conversely, silencing of the RxLR-type avirulence gene Avr3a1 by HIGS allowed infection of resistant tobacco by Phytophthora capsici (Vega-Arreguín et al., 2014), highlighting the power of HIGS to enable functional gene studies in plant-oomycete interactions. Recently, HIGS was suggested to induce gene suppression of infecting fungal and oomycete pathogens by plant endogenous small RNAs in Arabidopsis and in cotton, proposing a novel RNAbased plant defence mechanism (Zhang et al., 2016;Cai et al., 2018;Hou et al., 2019;Hou and Ma, 2020). In this report, we used Arabidopsis HIGS for targeted gene knockdown of the H. arabidopsidis CYSTEINE-RICH (HaCR)1 and ectopic plant expression of HaCR1 to explore the function of this non-RxLR CR effector protein in plant-pathogen interactions.

Plant materials and cultivation
Arabidopsis wild type (WT) Col-0, HIGS construct transformants and Hyaloperonospora arabidopsidis HaCR1 overexpression lines were cultivated under long day conditions in a growth chamber (16 h light: 8 h dark, 22 °C and 150 µmol m −2 s −1 photon flux density). Fourteen-day-old seedlings were used for H. arabidopsidis inoculation.
Arabidopsis effector overexpression lines were cultivated under short day conditions in a walk-in growth chamber (8 h light: 16 h dark, 22 °C and 150 µmol m −2 s −1 photon flux density). Five-to six-week-old plants were used for bacterial inoculation.
Wild tobacco (Nicotiana benthamiana Domin) plants were grown in a walk-in growth chamber under long day conditions (16 h light: 8 h dark, 22 °C and 275 µmol m −2 s −1 photon flux density) for 4 weeks prior to Agrobacterium tumefaciens-mediated transformation.

Plasmid construction
For HIGS constructs targeting HaCR1, HaACT A, HaA1E, or HaDCL1, target gene fragments of 334, 311, 267, and 256 bp length, respectively, were amplified from cDNA using home-made Phusion DNA polymerase. The DNA stretches were tested for off-targets in Arabidopsis and H. arabidopsidis cDNAs using the Si-Fi2.1 tool (http://labtools.ipkgatersleben.de/index.html) and have a maximum of two off-target small RNAs, as opposed to hundreds of effective on-target small RNAs.
RNA hairpins were cloned under the control of the strong proLjUBI promoter using the previously described and validated Golden Gate based RNAi plasmid assembly kit, containing the Arabidopsis AtWRKY33 intron 1 and the 35S terminator (Binder et al., 2014). Yellow fluorescent protein (YFP; mCherry for green fluorescent protein (GFP)-RNAi hairpin) was used in the final expression vector as an in planta transformation marker, and Agrobacterium tumefaciens AGL1 was transformed with completed vector constructs via electroporation.
Plasmid constructs for in planta expression were also made using the plant Golden Gate plasmid assembly kit (Binder et al., 2014). The coding sequence of HaCR1 was obtained by PCR amplification of H. arabidopsidis cDNA with Phusion High-Fidelity Polymerase (New England Biolabs, Frankfurt, Germany). The HaCR1 coding sequence lacking the signal peptide was amplified with home-made Taq DNA polymerase, as Phusion polymerase did not result in any amplification. Taq amplicons were blunted using Phusion DNA polymerase. All PCR products were validated by Sanger sequencing (LMU Genomics service unit, Planegg, Germany) before expression vector assembly.
The binary expression vector was assembled by ligation of the C-terminal GFP-tagged full-length or signal peptide-deleted HaCR1 sequences under the control of the proLjUBI promoter. As a control, a vector expressing only GFP was constructed. A list of primers used for the construction of plasmids is provided in Supplementary Table S1 at JXB online.

Arabidopsis transformation
Arabidopsis Col-0 plants were transformed by the floral dip method with A. tumefaciens strain AGL1, as described previously (Clough and Bent, 1998). Transformants from effector overexpression experiments were selected by kanamycin resistance on ½ MS agar plates with 1% sucrose and 50 mg l −1 kanamycin, as described previously (Harrison et al., 2006). Transformants expressing HIGS constructs were selected at the seedling stage by YFP fluorescence using a fluorescence stereo microscope. All experiments were performed on transgenic Arabidopsis plants in the T 2 generation.

RNA isolation, cDNA synthesis, and quantitative PCR
For DNA or RNA analysis, five Arabidopsis leaves from infected plants were pooled into one biological replicate, frozen in liquid nitrogen, and ground to powder using steel beads and a bead mill (MM400, Retsch, Haan, Germany). RNA was isolated using a modified cetyltrimethylammonium bromide-based protocol (Bemm et al., 2016). DNA digestion was performed on 1 µg total RNA using RNAse-free DNAse I (Thermo Fisher Scientific, Vilnius, Lithuania) after the manufacturer's instructions. For cDNA synthesis, SuperScript III (Thermo Fisher Scientific) and oligo-dT primers (50 µM) were used, following the manufacturer's instructions. Gene expression was determined by quantitative PCR (qPCR) using the EvaGreen master mix (Metabion, Planegg, Germany) or primaQUANT SYBRGreen Mastermix (Steinbrenner Laborsysteme, Wiesenbach, Germany) and a qPCR cycler (QuantStudio5, Thermo Fisher Scientific). For normalization of quantification values, H. arabidopsidis ELONGATION FACTOR 1α (HaEF1α) was validated as a reference gene using 40S ribosomal protein S3A (HaWS021) and β-TUBULIN (HaTUB) genes (Yan and Liou, 2006) (see Supplementary Fig. S1). For expression analysis of Arabidopsis genes, AtACTIN2 (AtACT2) was used as reference gene (An et al., 1996). Stable expression of AtACT2 was validated by correlating with the expression of AtTUBULIN (AtTUB). AtTUB was used in combination with AtUBQ10 as reference genes when was AtACT2 subjected for gene expression analysis itself. Differential expression was calculated using the 2 −ΔΔCt method (Livak and Schmittgen, 2001) and the reference gene(s) used for normalization are detailed in the figure legends. All primers with annealing temperature are listed in Supplementary Table S1.

Phylogenetic analysis
Conserved protein domains and motifs were analysed with InterPro (https://www.ebi.ac.uk/interpro/). Sequences of group I and II CR proteins from H. arabidopsidis were obtained from the NCBI GenBank (accession numbers JF800102-JF800110). The draft genome sequence of the Noco2 single spore isolate Noks1 was obtained from NCBI GenBank (accession number PRJNA298674). A phylogenetic tree and sequence alignment were constructed with CLC Main Workbench 7.6.4 (https:// digitalinsights.qiagen.com/), with default settings for the alignment, and the tree was constructed using neighbour joining and Jukes-Cantor distance measurement. A cysteine-rich protein from Phytophthora parasitica (PpCR; F443_03861) was used to root the tree.

Transient Nicotiana benthamiana transformation
Agrobacterium tumefaciens strain AGL1 was grown for 2 d at 28 °C in LB medium with appropriate antibiotics. Bacteria were harvested by centrifugation at 4000 g and incubated in induction buffer (10 mM MES-KOH pH 5.6, 10 mM MgCl 2 , 150 µM acetosyringone) for 1-2 h. The OD 600 was adjusted to 0.5 for each construct to perform pathogen assays and 0.25 for protein localization experiments. Leaves were infiltrated using needleless syringes and plants were replaced in the growth chamber under the same conditions.

Phytophthora/Botrytis pathogen assay
Two days after A. tumefaciens infiltration, N. benthamiana plants were inoculated with the respective pathogen by adding two Ø 0.5 cm agar plugs with mycelium per leaf. Images were taken with a camera and lesion sizes were measured with Fiji/ImageJ software (https://imagej.net/Fiji).

Cell death suppression assay
HaCR1-GFP, ∆SP-HaCR1-GFP, or GFP plasmids were co-transformed with the effector AvrE1 cloned from Pseudomonas syringae pv. tomato that elicits cell death in N. benthamiana (Badel et al., 2006). Agrobacterium tumefaciens cell concentration of all constructs was equally adjusted to a final OD 600 of 1.0. Infiltration was performed on 4-week-old N. benthamiana plants. Each individual construct was injected into the same leaf at separate areas (1.5 cm 2 ). Pictures of the leaves were taken 5 d post-infiltration and analysed by mean grey value counts using the Fiji/ ImageJ software (https://imagej.net/Fiji).

Epifluorescence and confocal microscopy
Overview pictures of N. benthamiana leaves were taken using a M165 FC epifluorescence stereomicroscope (Leica microsystems) with a GFP/ DsRED filter. Confocal laser-scanning microscopy of N. benthamiana leaves was performed with an upright SP5 confocal laser scanning microscope (Leica Microsystems) and imaged using an HCX IRAPO L256/0.95W objective (Leica Microsystems). For image acquisition, the resolution was set to 1024×1024 pixels and the frame average to 4. Fluorescent tags were excited using an argon laser at 20% power. GFP was excited with a 488 nm laser line and detected at 500-530 nm, cyan fluorescent protein (CFP) was excited with a 458 nm laser line and detected at 465-505 nm.

Collection of apoplastic wash fluid and apoplastic protein isolation
Six-week-old N. benthamiana plants were transformed with A. tumefaciens, as described above. To isolate apoplastic wash fluids we adapted and modified a published protocol for the isolation of apoplastic fluids and vesicles from Arabidopsis (Rutter et al., 2017), describing here the modifications. Two days after infiltration, the leaves were detached and the leaf surface was gently washed with ultrapure water. The leaflet was cut along the midrib and damaged areas were excised. The leaf stripes were washed in ultrapure water for 5 min to remove cytoplasm contamination from the cut surface. The leaf pieces were vacuum infiltrated with apoplastic wash buffer (20 mM MES, 2 mM CaCl 2 , 0.1 M NaCl, pH 6.0 with NaOH) for 4 min with a desiccator and the vacuum slowly removed within 4 min. The apoplastic fluid was then collected via centrifugation for 15 min at 250 g and 4°C. The isolated apoplastic wash fluid was split and one part directly used for apoplastic protease activity measurement. The other part was used for total apoplastic protein extraction. Therefore, proteins were collected by trichloroacetic acid and acetone precipitation and dissolved in 5× protein loading dye (225 mM Tris-HCl pH 6.8, 450 mM dithiothreitol (DTT), 5% SDS, 50% glycerol, 0.05% Bromphenol Blue).

Plant protease activity assay
Plant protease activity of isolated apoplastic wash fluid was determined using the Pierce fluorescent protease assay kit (Thermo Fisher Scientific) following the manufacturer's instructions for samples with low pH. Fluorescence was determined using a microplate reader (Tecan, Männedorf, Switzerland) with excitation and emission wavelengths of 485 nm and 538 nm, respectively. Protease activity was normalized on total protein content determined by Coomassie Brilliant Blue staining and quantified using Fiji/ImageJ (https://imagej.net/Fiji) as previously described (Miller, 2010).

Pseudomonas syringae pathogen assay
Liquid LB medium with rifampicin was inoculated with a single colony of Pst DC3000 and Pst DC3000 hrcC − and incubated overnight at 28°C. The bacteria were collected by centrifugation and diluted with 10 mM MgCl 2 to a final OD 600 of 0.0006. Leaves of 5-to 6-week-old Arabidopsis plants grown under short day conditions were infiltrated with the bacterial suspension, covered with a transparent lid and incubated under long day conditions. Two to three days after inoculation, leaf chlorosis of infiltrated leaves became visible and three leaf discs per biological replicate were harvested with a cork borer (Ø 0.6 cm). Leaf discs were homogenized in 10 mM MgCl 2 using a bead mill (MM400, Retsch) and two steel beads. A dilution series was plated on LB plates with rifampicin, and colony forming units were counted using a stereomicroscope.

HIGS is a powerful tool for functional gene studies in H. arabidopsidis
We used Arabidopsis HIGS in order to investigate the functional roles of genes in the obligate biotrophic plant pathogen H. arabidopsidis. As proof of concept, we chose four H. arabidopsidis candidate genes as HIGS targets, for which we presumed that gene knockdown would affect pathogen infection, namely the housekeeping gene ACTIN A (HaACT RNAi ), the CYSTEINE-RICH1 protein gene (HaCR1 RNAi ), an ALDOSE-1-EPIMERASE (HaA1E RNAi ) gene, and the type-III RNA endonuclease gene DICER-LIKE1 (HaDCL1 RNAi ). HaACT A (HpaG807716) is constitutively expressed in H. arabidopsidis and other oomycetes, and is a crucial component of the cytoskeleton (Ketelaar et al., 2012). HaCR1 (HpaG806256) and HaA1E (HpaG814621) are putative pathogenicity factors that are highly expressed in H. arabidopsidis during Arabidopsis infection (Asai et al., 2014). HaDCL1 (HpaG808216) is likely involved in biogenesis of H. arabidopsidis small RNAs, which we recently found to play an important role in suppressing plant genes for host infection (Dunker et al., 2020). The fungal plant pathogen Botrytis cinerea uses small RNAs for Arabidopsis plant infection, too (Weiberg et al., 2013), and HIGS against Botrytis DCLs indeed conferred disease resistance . To clone HIGS RNA hairpin transgenes (Fig. 1A), we chose target gene fragments that we predicted to not induce any off-target silencing either in H. arabidopsidis or in Arabidopsis using the Si-Fi2.1 tool (Lück et al., 2019). We confirmed the overall efficiency of our generated hairpin constructs by transient expression of a GFP RNA hairpin in leaves of the N. benthamiana line 16c stably expressing GFP (Ruiz et al., 1998) by A. tumefaciens infiltration. Transgenic GFP expression was clearly suppressed at local A. tumefaciens infiltration zones, as previously described (Kościańska et al., 2005), and release of repression by infiltration of a construct to co-express the viral RNAi suppressor protein p19 (Silhavy et al., 2002) verified GFP silencing via RNAi (see Supplementary Fig. S2). Therefore, we concluded that our constructs would effectively confer RNA silencing. Hence, we generated stable transgenic Arabidopsis lines expressing HIGS RNA hairpins in the ecotype Col-0. T 2 plants were selected and inoculated with the H. arabidopsidis isolate Noco2, which is virulent on Arabidopsis Col-0. We inspected infection phenotypes of WT and HIGS plants at 4 and 7 d post-inoculation (dpi) by light microscopy using the Trypan Blue staining method. Pathogen hyphae and haustoria were visible in all plant lines at 4 dpi, confirming successful infection ( Supplementary Fig. S3). At 7 dpi, local plant cell death was visible around the infecting hyphae in plants of a HaACT RNAi and two independent HaCR1 RNAi lines (Fig. 1B). Such H. arabidopsidis-induced local plant cell death, known as trailing necrosis, is associated with enhanced disease resistance against H. arabidopsidis infection (Uknes et al., 1992). Trailing necrosis also occurred, albeit to a lesser extent, in HaA1E RNAi plants ( Supplementary Fig. S4A), but not in WT (Fig. 1C) or in HaDCL1 RNAi plants ( Supplementary Fig. S4B). In HaACT RNAi and HaCR1 RNAi plants, trailing necrosis was accompanied by a reduction of H. arabidopsidis oospore production ( Fig. 1C; Supplementary Fig. S5). To examine the effect of HIGS on target gene expression, we determined transcript levels of HaACT A and HaCR1 in WT and HaACT RNAi or HaCR1 RNAi plants, respectively. We did not detect any target gene amplification by RT-PCR in non-inoculated HIGS plants, ensuring that the target gene-specific primers did not produce any signal derived from the HIGS hairpin construct (Supplementary Fig.  S6). Stable expression of the reference gene ELONGATION FACTOR 1α (HaEF1α, HpaG809424) was validated by quantitative reverse transcription (qRT)-PCR correlating their C t values with two other reference genes, 40S ribosomal protein S3A (HaWS021, HpaG810967) and HaTUB (a β-tubulin, HpaG814031), and of AtACt2 (At3G18780) by plotting the C t values against AtTUB (At5G62690) (Supplementary Fig.  S1A-D) according to the MIQE guidelines (Bustin et al., 2009). Gene silencing of HaACT A and HaCR1 was evident at 4 dpi in a HaACT RNAi line ( Supplementary Fig. S7A) and in two independent HIGS lines of HaCR1 RNAi (Fig. 1D). Neither target gene was suppressed at 7 dpi. (Supplementary Fig. S7B). The HaACT RNAi plants appeared smaller than WT plants ( Supplementary Fig. S8A), and thus we assumed an off-target effect on Arabidopsis ACTIN by the HaACT RNAi transgene. We determined the expression of the two Arabidopsis ACTIN genes, AtACT2 (At3G18780) and AtACT11 (At3G12110), showing the highest sequence similarity to HaACT A. The qRT-PCR analysis did not indicate any significant downregulation of AtACT2 or AtACT11 upon H. arabidopsidis infection at 4 dpi ( Supplementary Fig. S8B), rendering the connection between the HIGS construct and the plant growth phenotype unclear. We considered the possibility that such a growth phenotype in HaACT RNAi plants could have influenced the infection outcome with H. arabidopsidis. The transgenic Arabidopsis HaCR1 RNAi plants did not display any obvious pleiotropic effects, and we concluded that pathogen-induced plant cell death and enhanced disease resistance were due to HaCR1 silencing. With these data, we considered that HaCR1 was an important virulence factor of H. arabidopsidis to infect Arabidopsis.

HaCR1 is a member of the H. arabidopsidis CR effector protein family
To seek the potential function of HaCR1, we performed in silico protein sequence analysis. The HaCR1 172-amino-acid sequence has a predicted 19-amino-acid secretion signal peptide, but no further predicted functional domains or motifs. Sixteen family members of the HaCR proteins were previously classified into group I and group II by their cysteine pattern, with HaCR1 belonging to group I (Cabral et al., 2011). We accomplished phylogenetic analysis on the group I and II HaCR proteins using a Phytophthora capsici CR protein to root the phylogenetic tree. Phylogeny analysis suggested separate clades of HaCRs, with HaCR1 forming one branch with its close homologues HaCR3 (HpaG813024) and HaCR4 (HpaG806254), and the second clade consisting in HaCR5 (HpaG814422), HaCR6 (Cabral et al., 2011), and HaCR7 (HpaG814216). Further HaCR clades were not explicitly reliable due to overall low sequence conservation (see Supplementary Fig. S9A). We did not detect HaCR2, a HaCR1 homologue that was previously reported in the H. arabidopsidis strain Waco9 (Cabral et al., 2011), in the genome sequence of Noks1, a single-spore isolate of Noco2 (Bailey et al., 2011). Waco9 HaCR2 and Noks1 HaCR1 share a 96.9% amino acid sequence identity and 98.4% sequence similarity, but HaCR2 comprises an additional 89-amino-acid insertion in the middle part of the protein (Cabral et al., 2011). Consistent with the absence of HaCR2 for the Noks1 genome sequence, we could not amplify a HaCR2 orthologue by RT-PCR. We therefore concluded that there is no HaCR2 orthologue existing in the strain Noco2. HaCR1 and its closest homologue HaCR3 (BLASTp E-value 9×10 -30 ) share 53.4% sequence identity and 61.1% sequence similarity ( Supplementary Fig. S9B) on the amino acid level. HaCR1 is unique to the species of H. arabidopsidis, because we did not find any HaCR1 homologue in another oomycete species by BLASTp search against the NCBI database (E-value cut-off ≤1). As HaCR3 shared also 68.2% of transcript sequence identity to HaCR1 (Supplementary Fig.  S10A), we sought to examine co-suppression of HaCR3 in HaCR1 RNAi plants upon H. arabidopsidis infection. We performed qRT-PCR for gene expression analysis and observed comparable HaCR3 transcript accumulation in WT and HaCR1 RNAi plants at 4 dpi ( Supplementary Fig. S10B), suggesting that HaCR1 RNAi was specific to knockdown HaCR1, but not HaCR3.

HaCR1 inhibits induced plant cell death and promotes infection by (hemi)biotrophs
In order to shed light on HaCR1 function during plant infection, we performed transient expression assays using N. benthamiana leaves. We cloned a full-length HaCR1 version and fused it with a C-terminal GFP tag (HaCR1-GFP), a C-terminal GFP-tagged HaCR1 version without its predicted signal peptide (∆SP-HaCR1-GFP), and GFP without any HaCR1 sequence as a negative control (GFP) for expression in N. benthamiana leaves ( Fig. 2A, B). Because HaCR1 knockdown by HIGS resulted in plant trailing necrosis, we hypothesized that HaCR1 might promote infection through supressing plant cell death. To test this hypothesis, we co-expressed the HaCR1-GFP or the ∆SP-HaCR1-GFP construct together with the P. syringae effector AvrE, a known trigger of plant cell death in N. benthamiana (Badel et al., 2006). Only HaCR1-GFP was able to dampen AvrE1-induced plant cell death in contrast to both ∆SP-HaCR1-GFP and GFP (Fig. 2C). To further substantiate the role of HaCR1 as a plant cell death inhibitor for plant infection, we inoculated HaCR1-GFP-infiltrated N. benthamiana leaves with the hemibiotrophic oomycete pathogen P. capsici or with the necrotrophic fungal pathogen B. cinerea. These two pathogens lack any homologous protein with sequence similarity to HaCR1 (no BLASTp hit with E-value ≤5). Phytophthora capsici generated significantly larger lesions in HaCR1-GFP expressing leaves, compared with ∆SP-HaCR1-GFP or GFP expressing leaves (Fig. 2D). In contrast, B. cinerea, the infection of which is supported by induced plant cell death (Govrin and Levine, 2000), produced significantly smaller lesions in HaCR1-GFP expressing leaves than in ∆SP-HaCR1-GFP or in GFP expressing leaves (Fig. 2E). Since HaCR1 does not contain any RxLR plant cell translocation motif, we postulated that it could be active in the plant apoplast. To inspect HaCR1 intercellular localization in plants, we co-expressed a CFP-fused protein of the known plant plasma membrane marker Lti6B (Kurup et al., 2005) with HaCR1-GFP, ∆SP-HaCR1-GFP, or GFP in N. benthamiana leaves. Confocal microscopy studies revealed overlapping GFP and CFP signals for HaCR1-GFP indicating co-localization with Lti6B and the presence of HaCR1 in both the plant apoplast and the symplast, while ΔSP-HaCR1-GFP or GFP indicated signals separate from Lti6B and seemingly located only in the plant symplast (Fig. 3A). These results demonstrated that HaCR1 was functional in suppressing induced plant cell death, and its signal peptide was crucial for this function.

HaCR1 might act as an apoplastic protease inhibitor to support infection
A previously described role of fungal CR proteins is the inhibition of apoplastic plant protease (Rooney et al., 2005). Therefore, we hypothesized that HaCR1 might function as a decoy to inhibit plant apoplastic proteases, too. To challenge this hypothesis, we measured the capacity of HaCR1 to interfere with the apoplastic plant protease activity in vitro. We collected apoplastic wash fluids from N. benthamiana leaves expressing either HaCR1-GFP or ∆SP-HaCR1-GFP. Comparative analysis of the total leaf versus the apoplastic proteins by SDS-PAGE and silver staining displayed a reduction of the intracellular protein ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) in the apoplast fraction indicating successful enrichment of apoplastic proteins, even though we could not entirely prevent cytoplasmic protein contamination, as RuBisCO and ∆SP-HaCR1-GFP were still detectable in apoplast samples (Fig. 3B). Indeed, the apoplastic wash collected from N. benthamiana leaves expressing HaCR1-GFP exhibited a significant reduction of plant protease activity determined by fluorescein isothiocyanate (FITC)-casein compared with ∆SP-HaCR1-GFP (Fig. 3C, D). This result further supported a function of HaCR1 in the plant apoplast.
To investigate the suppressive effect of HaCR1 on plant immunity in the native host Arabidopsis during H. arabidopsidis infection, we generated transgenic Arabidopsis plants expressing HaCR1-GFP or ∆SP-HaCR1-GFP under the strong constitutive Lotus Ubiquitin promoter (Maekawa et al., 2008). We recovered three independent Arabidopsis T 2 lines for HaCR1-GFP and two independent lines for ∆SP-HaCR1-GFP and verified ectopic expression of fusion proteins in seedlings by fluorescence microscopy and western blot analysis (see Supplementary Fig. S11A, B). None of the transgenic lines exhibited any obvious growth or morphological change (Fig. 4A). We pooled and germinated seeds of the corresponding transgenic lines and challenged seedlings with the virulent H. arabidopsidis Noco2. Disease progression was estimated by H. arabidopsidis housekeeping gene expression of HaACT A relative to plant AtACT 2 at 4 and 7 dpi. Moderate but significantly increased pathogen quantity was evident in HaCR1-GFP expressing seedlings, compared with ∆SP-HaCR1-GFP (Fig. 4B). Moreover, expression of the Arabidopsis salicylic acid (SA)-dependent immunity marker gene AtPR1 was significantly less induced in seedlings expressing HaCR1-GFP compared with ∆SP-HaCR1-GFP upon H. arabidopsidis infection (Fig. 4C). This finding supported a role of the full-length HaCR1 in plant immune suppression. The jasmonic acid-dependent immunity marker gene AtPDF1.2 did not exhibit any difference between HaCR1-GFP and ∆SP-HaCR1-GFP upon H. arabidopsidis infection ( Supplementary Fig. S12). To further explore if the HaCR1suppressive effect on SA-dependent immunity played a role during infection, we inoculated transgenic Arabidopsis lines either with the virulent bacterial hemibiotrophic pathogen Pseudomonas syringae pv. tomato (Pst) strain DC3000 or the avirulent mutant Pst DC3000 hrcC − lacking a functional type-III secretion system (Roine et al., 1997). Bacterial growth of DC3000 was significantly enhanced in HaCR1-GFP expressing Arabidopsis, compared with ∆SP-HaCR1-GFP. In contrast, bacterial population of hrcC − remained unaltered between the two different transgenic plant lines (Fig. 4D). These results further supported that HaCR1 is an apoplastic effector that impairs plant immunity against diverse biotrophic and hemibiotrophic plant pathogens.

Discussion
In this study, we used Arabidopsis HIGS for functional gene studies in the obligate biotrophic pathogen H. arabidopsidis. The short lifecycle, available cloning tools, and easy transformation of the host plant Arabidopsis enables the conducting of HIGS experiments in a relatively short time period. We applied HIGS to knockdown HaACT A, HaDCL1, HaCR1, and HaA1E in order to survey functional roles of these pathogen genes during host plant colonization. The HaCR1 RNAi , HaACT RNAi plants, and to a lesser extent HaA1E RNAi , exhibited trailing necrosis at H. arabidopsidis infection sites. In addition, HaCR1 RNAi and HaACT RNAi plants allowed reduced proliferation of oospores ( Fig. 1A-C; Supplementary  Fig. S4A), the sexual reproductive structure of oomycetes (Slusarenko and Schlaich, 2003). Both infection phenotypes are related to reduced disease, as comparable trailing necrosis symptoms had been observed when Arabidopsis was primed for immunity, or connected to Arabidopsis ecotypes infected with sub-compatible H. arabidopsidis strains (Uknes et al., 1992;Krasileva et al., 2011). The resistance response in HaACT RNAi plants also suggested that down-regulation of HaACT A was not compensated through functional redundancy by the paralogue HaACT B (HpaG809873), despite considerable sequence homology with the HaACT RNAi HIGS construct (see Supplementary Fig. S13A). The attenuated disease development in HaACT RNAi , HaCR1 RNAi , and HaA1E RNAi plants was not due to plant transformation or due to expression of non-self dsRNA in Arabidopsis, because HaDCL1 RNAi did not reveal any higher plant resistance or suppressed pathogen virulence ( Supplementary Fig. S4B). Why HaDCL1 RNAi plants did not reveal higher resistance despite the important role of pathogen small RNAs during infection (Dunker et al., 2020) remains to be investigated. One possible explanation might be functional redundancy of the two HaDCLs identified in the genome of H. arabidopsidis (Bollmann et al., 2016). Similarly, the mild phenotype expressed in the Arabidopsis HaA1E RNAi line might be explained by the presence of two paralogous and potentially functionally redundant genes: HaA1E-LIKE (HaA1EL, HpaG807738) and HaA1E-LIKE2 (HaA1EL2, HpaG807727). HaA1EL shares 66.3% amino acid identity and 79.6% amino acid similarity, while HaA1EL2 shares 86.6% amino acid identity and 90.3% amino acid similarity with HaA1E. HaA1EL and HaA1EL2 showed also 79.4% and 89.9% coding sequence identity, respectively. In addition, they revealed a considerable DNA sequence homology with our HaA1E RNAi HIGS construct ( Supplementary Fig. S13B). However, HaA1EL2 lacks an annotated open reading frame with a signal peptide and displayed very weak expression during infection. In contrast HaA1E and HaA1EL both comprise a predicted signal peptide and were previously found to be strongly expressed in H. arabidopsidis infecting Arabidopsis (Asai et al., 2014).
At the molecular level, target gene suppression of HaACT and HaCR1 by HIGS was evident at 4 dpi, but not at 7 dpi ( Fig. 1D; Supplementary Fig. S7). At this later time point, H. arabidopsidis had induced trailing necrosis of plant cells at the infection sites of HIGS plants. We suggest that plant cell death would lead to a collapsed haustoria-plant cell interface, which likely stopped the transport of RNAs from plants into the pathogen (Hudzik et al., 2020). Transgenic HaACT RNAi expressing Arabidopsis displayed pleiotropic effects, for instance slower plant growth, although we did not predict any Arabidopsis ACTIN as offtarget of the HaACT RNAi construct in silico, and the two closest orthologues of HpaACT A, AtACT2 and AtACT11, were not suppressed in the Arabidopsis HIGS line ( Supplementary Fig.  S8). Tracing back off-target effects would require plant RNA degradome analysis (Casacuberta et al., 2015), and was not further investigated as it would go beyond the scope of our study. With this experience, we propose to omit pathogen housekeeping genes as targets in HIGS studies although successful silencing would likely promote plant disease resistance. Since pathogen effector genes are unique and homologues do not exist in the host plant, HIGS against HaCR1 did not encounter any off-target problem. An interesting alternative to HIGS for targeted gene knock-down in H. arabidopsidis that is based on exogenous application of 5′ capped small interfering RNAs has been recently reported (Bilir et al., 2019). Applying Cellulose synthase A3 antisense RNAs to H. arabidopsidis conidia suspension inhibited spore germination on the leaf surface. Both, transgenic HIGS and external RNA treatment are innovative strategies to further explore gene functions of this pathogen.
To further investigate the role of HaCR1 during plant infection, we expressed HaCR1 in N. benthamiana and Arabidopsis. One obvious disease symptom in transgenic HaCR1 RNAi plants was the induction of local plant cell death suggesting that HaCR1 might be involved in cell death suppression. stained with Coomassie G250 was used as a loading control. (C) A representative picture of a tobacco leaf at 5 d after A. tumefaciens co-infiltration carrying either HaCR1-GFP, ∆SP-HaCR1-GFP or GFP together with a construct carrying the bacterial effector AvrE promoting cell death. This experiment was repeated three times with comparable results. Each experiment included three infiltrated leaves. Quantification of chlorosis symptoms was performed by measuring the mean grey value of the infiltrated area, with n=3. (D) Agrobacterium tumefaciens-infiltrated N. bethamiana leaves of HaCR1-GFP, ⊿SP-HaCR1-GFP, or GFP were inoculated with P. capsici, and pictures were taken at 2 dpi. Lesion size quantification on N. benthamiana leaves induced by P. capsici at 2 dpi, as determined by ImageJ with n≥20 lesions of n≥10 leaves. (E) Agrobacterium tumefaciens-infiltrated N. benthamiana leaves of HaCR1-GFP, ⊿SP-HaCR1-GFP, or GFP were inoculated with B. cinerea, and pictures were taken at 3 dpi. Lesion size quantification on N. benthamiana leaves induced by B. cinerea at 3 dpi, as determined by ImageJ with n≥20 lesions of n≥10 leaves. Letters in (C-E) indicate groups of statistically significant difference by ANOVA followed by Tukey's HSD test with P≤0.05. Such a function of HaCR1 was supported by the inhibitory activity on bacterial effector AvrE-induced plant cell death in N. benthamiana leaves (Fig. 2C). Of note, full-length HaCR1, but not a signal peptide-deleted version, was capable of suppressing plant cell death in this assay. AvrE expressed in N. benthamiana leaves was previously described to be localized at the cell plasma membrane (Xin et al., 2015), a possible contact compartment of in planta-expressed full length HaCR1 with AvrE. However, the molecular mechanism of AvrE-induced cell death repression by HaCR1 is not clear, and needs to be further explored by identifying the molecular interactors of HaCR1. On the one hand, HaCR1 promoted disease caused by the oomycete hemibiotrophic pathogen P. capsici (Fig. 2D). We suggest that P. capsici profits from HaCR1-repressed plant cell death during the early biotrophic phase. This is in line with Avr1b from Phytophthora sojae that impaired plant cell death and promoted lesion formation of this hemibiotrophic pathogen (Dou et al., 2008). On the other hand, HaCR1 expression limited disease symptoms caused by the necrotrophic fungal pathogen B. cinerea (Fig. 2E), because this pathogen exploits and promotes plant apoptosis for infection (Veloso and van Kan, 2018). In this context, reduced Botrytis virulence was reported in plants expressing animal cell death suppressors (Dickman et al., 2001). Consistent with plant cell death suppressive activity, HaCR1 also promoted disease progression caused by other (hemi)biotrophic pathogens, P. syringae DC3000 and H. arabidopsidis itself (Fig. 4B, E). HaCR1 overexpression in Arabidopsis moderately promoted H. arabidopsidis disease, which might be explained by the high expression of endogenous HaCR1 in H. arabidopsidis during infection. Similarly, a previous study on HaRxLR effectors could detect only small positive effects on H. arabidopsidis infection suggesting a combined action of effectors to effectively suppress plant immunity (Pel et al., 2014).
To better understand the molecular function of HaCR1, we explored its peptide composition. HaCR1 contains a predicted secretion signal peptide but no further plant cell translocation domain indicating its function in the plant apoplast. In accordance, only full-length HaCR1 expression in plants suppressed induced plant cell death and promoted infection of (hemi)biotrophs, while a secretion signal peptide-truncated HaCR1 version expressed in plants lost these activities. This is in agreement with other apoplastic effectors found in fungal pathogens, such as Zymoseptoria tritici and Magnaporthe oryzae (Kim et al., 2013;Kettles et al., 2017). A conserved class of apoplastic effectors in fungi are the LysMs. These effectors act as decoys that prevent microbe-associated molecular pattern (MAMP)-triggered plant immunity, for instance by binding chitin oligomers and thereby hampering chitin recognition by plant pattern receptors (Kombrink and Thomma, 2013;Zeng et al., 2020). However, HaCR1, like all other members of the HaCR family, does not contain any predicted protein domain or motif, making a specific ligand binding rather unlikely. Instead, we found evidence that plant-expressed HaCR1 can interfere with apoplastic plant protease activity in vitro (Fig. 3D), similar to fungal CR proteins exhibiting protease inhibition activity (Rooney et al., 2005). The strict dependency of HaCR1 function on the presence of a signal peptide, and thereby its apoplastic localization, suggests a link between the cell death inhibition function and plant protease inhibition. In this context, several apoplastic plant proteases are crucial regulatory components of plant programmed cell death, and protease inhibitory effectors of fungi and oomycetes have been associated with inhibition of plant programmed cell death (Dickman and Fluhr, 2013;Salguero-Linares and Coll, 2019).
In general, small, apoplastic CR peptides containing no further sequence-conserved domains have been described in high numbers for oomycete and fungal pathogens suggesting a functional conservation in a wide range of pathogens (Sperschneider et al., 2018). Pathogen-secreted protease inhibitors or decoys prevent degradation of pathogen effectors or the release of MAMPs produced by plant proteases (Jiang and Tyler, 2012). Such a function of HaCR1 is supported by our results, because Arabidopsis overexpressing HaCR1 exhibited reduced AtPR1 induction upon H. arabidopsidis infection (Fig. 4C), and HaCR1 expression in plants promoted infection of (hemi)biotrophic pathogens H. arabidopsidis, P. capsici, and P. syringae.
Our data revealed an important role of a CR effector protein in host infection by the obligate biotrophic pathogen H. arabidopsidis. A next crucial step to understand the molecular mechanism of how HaCR1 suppresses pathogen-induced plant cell death will be to uncover its molecular interactors, which are likely to include plant apoplastic proteases or receptor-like proteins. This knowledge would be crucial to completely elucidate whether the dual function of HaCR1 in plant protease inhibition and cell death inhibition is directly or indirectly linked, or is independent. microscopy images (lower panel). (B) Total leaf protein and apoplastic wash fluid fraction visualized via a silver-stained SDS-PAGE indicated depletion of cytoplasmic proteins, such as RuBisCO, from the apoplastic wash fluid. Western blot shows detection of HaCR1 and free GFP in total leaf and apoplastic wash fluid from the same experiment. (C) Schematic overview of the apoplastic protease activity assay. Nicotiana benthamiana leaves were transformed by A. tumefaciens infiltration, apoplastic wash fluid was collected by centrifugation, and endogenous protease activity was determined by addition of FITC-casein. Upon protease activity, casein is hydrolysed and quenching of FITC fluorescence is released. (D) Protease activity in the apoplastic wash fluid was measured and compared with apoplastic proteins collected from N. benthamiana leaves expressing HaCR1-GFP or ∆SP-HaCR1-GFP. Protease activity was normalized to total protein quantities of apoplastic fluid samples. Each data point represents an independent experiment using eight leaves. *P≤0.05, significant difference by Student's t-test.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Expression correlation analysis of H. arabidopsidis and Arabidopsis reference genes by qRT-PCR.  HaACT expression was not detected (n.d.) before infection. (C) AtPR1 was quantified by qRT-PCR using AtACT2 and AtTUB as reference genes, and relative transcript levels were compared between Arabidopsis expressing HaCR1-GFP or ∆SP-HaCR1-GFP at 4 and 7 dpi with H. arabidopsidis. (D) Arabidopsis susceptibility to the virulent Pst DC3000 or the avirulent Pst DC3000 hrcC − was evaluated by counting colony forming units (cfu) at 3 dpi. Each data point represents cfu derived from three infected leaf discs. For (B, C), each experiment was performed at least with three biological replicates, and each biological replicate represented two technical repeats. For (B-D) *P≤0.05, significant difference by one-tailed Student's t-test.  Table S1. Primers used in this study.