The root‐invading pathogen Fusarium oxysporum targets pattern‐triggered immunity using both cytoplasmic and apoplastic effectors

Summary Plant pathogens use effector proteins to promote host colonisation. The mode of action of effectors from root‐invading pathogens, such as Fusarium oxysporum (Fo), is poorly understood. Here, we investigated whether Fo effectors suppress pattern‐triggered immunity (PTI), and whether they enter host cells during infection. Eight candidate effectors of an Arabidopsis‐infecting Fo strain were expressed with and without signal peptide for secretion in Nicotiana benthamiana and their effect on flg22‐triggered and chitin‐triggered reactive oxidative species (ROS) burst was monitored. To detect uptake, effector biotinylation by an intracellular Arabidopsis‐produced biotin ligase was examined following root infection. Four effectors suppressed PTI signalling; two acted intracellularly and two apoplastically. Heterologous expression of a PTI‐suppressing effector in Arabidopsis enhanced bacterial susceptibility. Consistent with an intracellular activity, host cell uptake of five effectors, but not of the apoplastically acting ones, was detected in Fo‐infected Arabidopsis roots. Multiple Fo effectors targeted PTI signalling, uncovering a surprising overlap in infection strategies between foliar and root pathogens. Extracellular targeting of flg22 signalling by a microbial effector provides a new mechanism on how plant pathogens manipulate their host. Effector translocation appears independent of protein size, charge, presence of conserved motifs or the promoter driving its expression.


Introduction
Plant pathogens cause severe damage in agriculture, resulting in average yield losses of 20% worldwide (Strange & Scott, 2005;Fisher et al., 2012). Central to their parasitic lifestyle is the ability to manipulate their host. Bacteria, fungi and oomycetes, but also nematodes and insects, employ small, secreted proteins called effectors for this manipulation (Lo Presti et al., 2015;Toruno et al., 2016;Uhse & Djamei, 2018). Revealing the host processes targeted by effectors is key to understanding the virulence strategy of a pathogen.
Plants employ a two-layered immune system to ward off biotic threats. Recognition of microbe-derived molecules (called PAMPs, for pathogen-associated molecular patterns) triggers activation of a generic defence response (Cook et al., 2015;Couto & Zipfel, 2016;Gust et al., 2017). Classical PAMPs are bacterial flagellin (and its derived peptide flg22) and fungal chitin (Boller & Felix, 2009;Shinya et al., 2015). PAMP perception is mediated by cell-surface localised receptors, such as the flagellin receptor FLS2. Upon flg22 binding, FLS2 associates with the coreceptor BAK1 and forms a signalling complex (Schulze et al., 2010). Chitin perception in Arabidopsis depends on the receptor-like protein and -kinase (RLP/RLK) LYK5 and CERK1 (Shinya et al., 2015). PAMP recognition induces a series of physiological events, including changes in ion fluxes, production of ROS, activation of protein kinases and transcriptional reprogramming. Altogether, this pattern-triggered immunity (PTI) limits colonisation by most potential pathogens. The second layer of plant immunity can be activated upon recognition of pathogen-produced effectors by resistance genes that typically encode either intracellular NB-LRR or cell-surface RLK/RLPtype receptors. Effector perception by resistance proteins generally induces a strong immune response, often involving host cell death (Jones & Dangl, 2006).
Bacterial effectors frequently target PTI by interfering with intracellular signal transduction (Dou & Zhou, 2012). Many oomycete effectors also act inside plant cells to repress PTI, although the underlying mechanisms are less well understood (Fabro et al., 2011;Zheng et al., 2014). By contrast, known fungal effectors typically interfere with PAMP perception rather than PTI signalling. They do so by, for example, binding chitin or beta-glucan fragments in the apoplast, hiding them for the immune receptors (Lo Presti et al., 2015;Wawra et al., 2016). Known intracellular functions of fungal effectors involve manipulation of metabolic processes, or of transcriptional regulators (Djamei et al., 2011;Plett et al., 2014;Tanaka et al., 2014;Wessling et al., 2014). Recent evidence has indicated that fungal effectors can repress intracellular PTI signalling as well (Di et al., 2017;Irieda et al., 2019;Navarrete et al., 2019), although it remains to be shown whether this represents a generic strategy for fungi.
To target intracellular PTI signalling, effectors have to enter plant cells. Whereas the delivery of bacterial effectors via the type III secretion system is relatively well understood, how eukaryotic effectors reach the plant cytoplasm remains enigmatic (Petre & Kamoun, 2014). Translocated oomycete effectors carry well defined motifs close to their N-terminus, notably the RxLR-dEER and the LFLAK motif, that are critical for targeting them into host cells. Although the underlying mechanism is controversial (Whisson et al., 2007;Kale et al., 2010;Schornack et al., 2010;Wawra et al., 2017), these motifs are useful for bioinformatics-assisted identification of potential cytoplasmic effectors in these pathogens. In contrast to oomycete effectors, conserved sequence motifs have not yet been described for cytoplasmic fungal effectors, except for an Y/F/WxC motif specific for a family of powdery mildew effectors (Godfrey et al., 2010). Thus, discriminating cytoplasmic from apoplastic fungal effectors is currently not possible based on sequence analysis, and the properties explaining effector translocation into host cells remain to be elucidated.
The concept of PTI and its suppression by intracellularly acting effectors is largely based on the study of foliar pathogens. Roots are exposed to both beneficial and pathogenic microbes, a condition that has likely shaped the immune system in roots (De Coninck et al., 2015;Hacquard et al., 2017;Zhou et al., 2020). Currently, it is poorly understood which type of defence responses are activated in roots, and whether root pathogens use similar mechanisms for their suppression as those employed in aboveground interactions. Fo is a widespread root coloniser that can cause disease on more than 120 different plant species, many of agricultural importance (Edel-Hermann & Lecomte, 2019). Whereas colonisation of the root surface and the cortex is generally symptomless, pathogenic isolates invade the vasculature and grow upwards in the xylem vessels causing disease (Michielse & Rep, 2009). Individual strains have a narrow host range that is potentially determined by their effector repertoires . Initially, a set of 14 candidate effectors from a tomato-infecting strain (Fo f.sp. lycopersici; Fol ) was identified by proteomic analysis of the xylem sap of infected tomato plants, and hence these are named Secreted in xylem (Six) proteins. They all carry a signal peptide, are relatively small (<300 amino acids) and the encoding genes locate in the genome close to a transposable element (Schmidt et al., 2013;Gawehns et al., 2015). Three of them are recognised as avirulence factors: Six3 (Avr2) is recognised by an NB-LRR, whereas Six1 (Avr3) and Six4 (Avr1) are recognised by an RLK and an RLP, respectively (Houterman et al., 2009;Catanzariti et al., 2015;Catanzariti et al., 2017). Even though several Fo effectors were shown to contribute to virulence, an underlying mechanism has been proposed for only two of them: Avr2 from a tomato-infecting strain (Fo f.sp. lycopersici; Fol) targets PTI, whereas Six5 promotes cell-to-cell movement of Avr2 (Di et al., 2017;Cao et al., 2018). Thus, it remains an intriguing question how this hemi-biotrophic pathogen uses its effector repertoire for successful root and xylem colonisation.
Here we report the systematic identification, and characterisation, of candidate effectors from an Arabidopsis-infecting Fo strain, Fo5176. To reveal their cellular site of action, effectors were targeted to either the cytosol or the apoplast of N. benthamiana leaves and their effect on a typical PTI response was assessed. In addition, translocation of effectors into host cells during Arabidopsis root colonisation was monitored. Our work identifies PTI as a common target for Fo effectors. Furthermore, effector translocation into host cells is not determined by their promoter or untranslated region (UTR) sequences, nor by general protein properties such as size and charge.

Plasmid construction
For producing Avi-tagged effectors in Fo5176, the binary vector pRW2h (Houterman et al., 2008) was modified as follows: an oligonucleotide (oligonucleotide sequences are listed in Supporting Information Table S1) containing the coding sequences of a single HA unit followed by the AviTag TM was inserted into a XbaI-and BglIII-digested vector. The oligonucleotide contained a single AscI site preceding the HA-AviTag TM . Next, the FEM1 promoter was replaced by c. 1 kb promoter fragment directly 5 0 of the Fo5176 SIX1 gene, using HindIII and XbaI sites. The obtained vector was named pRW2h-Six1p-HB.
For expressing the effector genes in planta, the binary vector pBIN61 (SLDB3104; Tameling et al., 2010) was modified as follows: restriction sites were added to the aforementioned HA-AviTag oligonucleotide, which was then inserted into the XbaI and XmaI sites of SLDB3104. The obtained vector was named pBIN61-HB.
The coding sequence of effector genes was amplified by PCR, using as template cDNA obtained from Fo-inoculated Arabidopsis seedling roots (to be described later). The primers introduced AscI restriction sites to both ends of the PCR product, allowing its ligation into the AscI site of pRW2h-Six1p-HB and pBIN61-HB, respectively. The obtained plasmids allowed production of effectors that are C-terminally fused to the HA-AviTag.

Research
New Phytologist N. benthamiana as described previously (Ma et al., 2012) with following modifications: Overnight cultures were resuspended in infiltration medium (10 mM MgCl 2 , 10 mM MES, 200 µM acetosyringone, pH 5.8) and infiltrated into leaves of c. 4-wk-old N. benthamiana plants at an OD 600 of 0.5. A strain containing the silencing suppressor P19 was co-infiltrated at an OD 600 of 0.5.

Generation of stable transgenic Arabidopsis
Arabidopsis Col-0 plants expressing BirA were transformed with agrobacteria containing the binary vector pBIN-dspFoa2-HB using the floral dip method. Transformants were selected on medium containing 40 mg l À1 kanamycin. Single insertion lines were selected according to segregation ratio analysis, and putative transformants were confirmed by PCR and by Western blot. Two independent transformants differing in dspFoa2 accumulation (named 5-1 and 6-6) were selected. As the Col-0 BirA line was transformed with dspFoa2, this line was also used as the wild-type control in all experiments.

Reactive oxidative species (ROS) assay
ROS production was measured with a luminol/peroxidase-based assay as described previously (Felix et al., 1999). Briefly, N. benthamiana leaf discs (5 mm diameter) were collected 2 d after Agrobacterium infiltration and floated overnight on deionised water (MQ) in a Petri dish. The next morning, leaf discs were transferred to 96-well plates (Perkin Elmer, Waltham, MA, USA, white) containing 100 µl water per well, and immediately before measurement, 100 µl of a luminol/peroxidase solution was added to each well. Final concentrations were 250 µM luminol (Sigma), 10 µg ml À1 peroxidase (Sigma) and 200 nM flg22. To measure chitin-induced ROS production, a final concentration of 10 µM chitin hexaose (Elicityl SA, Crolles, France) and 250 µM luminol L-012 were used. Luminescence was recorded using 2-min intervals for 1 h using a plate reader (Synergy H1; BioTek, Winooski, VT, USA). To measure ROS in Arabidopsis, leaf discs from 5-6wk-old plants were collected, and the experiment was performed similarly as for N. benthamiana, except that luminol L-012 was used for both flg22 and chitin assays.

Pseudomonas syringae infection assay
Arabidopsis inoculation with P. syringae pathovar tomato (Pst) DC3000 was performed as described previously (Saijo et al., 2009), with following modifications: 5-6-wk-old plants were spray inoculated with bacteria at an OD 600 of 0.2, and bacterial titres were determined at 4 d post inoculation (dpi).

Transformation of F. oxysporum
Stable transformants of Fo5176 were generated by Agrobacterium-mediated transformation as described previously (Takken et al., 2004). Successful transformants were selected based on hygromycin resistance. For each construct, at least five independent transformants were used for inoculation of Arabidopsis seedlings, followed by Western blotting to detect accumulation of tagged effectors in the infected seedlings.

Arabidopsis inoculation with Fo
Fo51756 was grown on Czapek-Dox-Agar medium. To obtain spores, four agar plugs (each c. 0.5 9 0.5 cm) containing fungal mycelium were added to 100 ml liquid medium (3% sucrose, 100 mM KNO 3 and 0.17% yeast nitrogen base) and incubated on a shaker for 3 d at 25°C. Spores were harvested by filtration through a miracloth, washed and their concentration was measured using a cell counter. Arabidopsis seedlings were grown on vertically placed, square plates (12 9 12 cm, Greiner) containing 1% agar supplemented with Murashige and Skoog (MS) medium and 1% sucrose for 15-16 d in a climate chamber (14 h : 10 h, light : dark cycle, 21°C). For Fo inoculation, seedlings were transferred into horizontally arranged square plates (12 9 12 cm) that contained a 2-3 cm strip of agar at one end, whereas the rest of the plate was filled with liquid MS medium containing 10 5 Fo51756 spores ml À1 . The shoots of the seedlings were placed onto the agar strip, to prevent direct contact with the fungal spores. The plates were closed, aluminium foil was placed above the root containing parts and the plates were placed in a climate chamber (11 h : 13 h, light : dark cycle, 28°C) for 6 d. At this stage, the Fo-inoculated seedlings showed symptoms such as brown discoloration of the roots, and chlorosis of the leaf veins.

RNA isolation and cDNA synthesis
Total RNA was isolated from roots of Fo5176-inoculated Arabidopsis seedlings using TRIzol LS reagent (Ambion) according to the manufacturer's instructions. DNA was removed by oncolumn treatment with DNase (Thermo Scientific). cDNA was synthesised from 500 ng RNA using the RevertAid Reverse Transcription kit (Thermo Scientific) according to the manufacturer's instructions.

Streptavidin pulldown
Snap-frozen roots of Fo-inoculated seedlings were ground in liquid nitrogen and 100 mg was dissolved in 1.2 ml lysis buffer Ó2020 The Authors New Phytologist Ó2020 New Phytologist Trust New Phytologist (2020) 227: 1479-1492 www.newphytologist.com (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 10 mM EDTA, 10% glycerol, 10 mM DTT, 2% PVPP, 0.15% Nonidet P40 and protease inhibitor cocktail (Roche)). An aliquot of the supernatant was kept as input sample; the remaining material was incubated with 40 µl magnetic streptavidin beads (GE Healthcare, Chicago, IL, USA) and placed on a rotator for 2 h at 4°C. After washing three times, captured proteins were released by shaking the beads in 100 µl SDS loading buffer for 5-6 min at 96°C. As input, 20 µl of the lysate was loaded onto SDS gels, together with 10 µl of streptavidin-captured fraction.

Protein isolation and Western blot analysis
Proteins were isolated from N. benthamiana by grinding 20 mg leaf tissue in 100 µl lysis buffer (50 mM Tris-HCl pH 7.5, 2% SDS, 5 mM DTT, and 19 protease inhibitor; Roche), and centrifuging at 16 000 g for 25 min. Proteins were separated by SDS-PAGE using, in most cases, 13% gels. Western blotting was carried out using the semidry method on polyvinylidene difluoride (PVDF) membranes. Blots were probed with rat monoclonal anti-HA antibodies (clone 3F10; Roche) at a dilution of 1 : 5000. The secondary goat-anti-rat antibody (Pierce) was used at a dilution of 1 : 8000. The signal was visualised with the ECL kits (GE Healthcare, ECL prime or Thermo Scientific, Super Signal West Pico) according to the manufacturer's instructions.

Results
Determining the candidate effector repertoire of an Arabidopsis-infecting Fo isolate To gain insight into Fo effector-mediated host manipulation, we set out to identify candidate effectors in strain Fo5176, which is pathogenic on Arabidopsis (Thatcher et al., 2009). An effector prediction pipeline, encompassing 65 sequenced Fo isolates, revealed 62 candidates in this strain . Candidate effector prediction was based on the presence of a secretory signal peptide, a protein size of < 300 amino acids, and presence of inverted repeats of a transposable element in their promoter. Of the candidates, 19 showed a mosaic-like phylogenic distribution; a hallmark of known Fo effectors (clusters C, D, E and F in van . This set was further refined by excluding genes that were mis-annotated (1), are present in all pathogenic isolates (1), had an annotated function and/or domain (3), encoded a protein shorter than 60 amino acids (4) or presented an artefact of the prediction software (1). This resulted in a shortlist of nine candidate effectors, of which seven were successfully cloned and further characterised (Table S2). These seven candidate effectors include four SIX gene homologues present in Fo5176 (Six1, Six4, Six8 and Six9) and three novel genes, designated FoaEffector1, FoaEffector2 and FoaEffector3 (from this point forward Foa1, Foa2 and Foa3). In addition, one gene was included that was present in all the sequenced isolates, but showed high sequence variation, indicating that the gene may have undergone diversifying selection (Foa4). Fig. 1(a) shows a schematic representation of the identified candidate effectors. Without predicted signal peptides, their sizes ranged from 84 to 258 amino acids. All predicted proteins contain a paired number of cysteine residues, except Six1 (nine cysteines), suggesting that the ability to form disulfide bridges is a common feature for Fo effectors. The three largest candidate effectors, Six1, Six4 and Foa1, have a predicted N-terminal prodomain that is separated from the rest of the protein by a predicted Kex2 protease cleavage site and is followed by a central region with six similarly spaced cysteines. In addition, the prodomains of Six1 and Six4 contain predicted N-glycosylation motifs. Notably, Foa2 shares size and cysteine-spacing with FolAvr2 and FomAvr2, two sequence-unrelated effectors from tomato-infecting and melon-infecting Fo isolates, which are both recognised by intracellular NB-LRR proteins Schmidt et al., 2016). Six9, Foa3 and Foa4 are relatively cysteine-rich, whereas Six8 has only two cysteine residues. Taken together, although the candidate Fo5176 effector sequences share potentially structural homologies with each other, as well as with other Fo effectors, this information does not reveal insights into their function or localisation during host colonisation by the fungus.
Several Fo5176 candidate effectors suppress the flg22and/or chitin-triggered ROS burst in N. benthamiana To elucidate a possible virulence function of the identified candidate effectors we tested whether they could suppress a classical PTI output, the ROS burst induced by flg22; a 22 amino acid derivative of the bacterial PAMP flagellin (Felix et al., 1999). To this end, the candidate effector encoding genes were transiently expressed in N. benthamiana. To monitor whether the effectors act intracellularly or extracellularly, each candidate effector was expressed either with or without its endogenous signal peptide (indicated by the prefix 'sp' and 'dsp', for deleted sp). A C-terminal HA tag allowed detection of protein accumulation by Western blot analysis. Transient expression of both sp and dsp versions resulted in comparable accumulation levels and similar apparent molecular sizes for most effector proteins (Fig. 2a). Only dspFoa4 did not accumulate to detectable levels, and the abundance of dspFoa1 was somewhat reduced as compared with the secreted version. Six1, Six4 and Six9 expressed with the signal peptide migrated at a higher molecular weights than variants lacking the signal peptide, indicative for a posttranslational modification. In conclusion, all Fo effector candidates were successfully produced in planta, enabling assessment of a potential involvement in flg22-induced ROS burst in N. benthamiana.
Consistent with earlier reports, N. benthamiana leaves infiltrated with agrobacteria containing an empty binary vector responded to flg22 with a rapid and transient ROS burst (Fig. S1a;Zipfel et al., 2006). As a positive control for effectormediated suppression we used FolAvr2, which acts intracellularly to inhibit various flg22-induced PTI responses, including the ROS burst (Di et al., 2017). Indeed, as compared with the empty vector control, ROS levels were lower in the presence of Avr2 (Fig. 2b). Markedly, four out of the eight newly identified New Phytologist (2020) 227: 1479-1492 Ó2020 The Authors New Phytologist Ó2020 New Phytologist Trust www.newphytologist.com

Research
New Phytologist effector candidates also significantly reduced the flg22-induced ROS response, whereby Foa1, Foa2 and Foa3 conferred a nearly complete suppression (Fig. 2b). Since Foa2 and Foa3 inhibited the flg22 output when expressed without their signal peptide, they likely act intracellularly. Surprisingly, both effectors suppressed the ROS response to the same extent when expressed with their cognate signal peptide, indicating that they either remain inside, or they re-enter the cytosol after secretion. By contrast, Six1 and Foa1 only showed ROS suppression when expressed with their signal peptide, implying that these proteins acted extracellularly and have to be secreted to function (Fig. 2b).
Since flg22 is a bacterial PAMP and we used fungal effector protein candidates, we investigated whether these also altered responsiveness to the fungal PAMP chitin. Similar to flg22, chitin induced a rapid and transient ROS burst in N. benthamiana leaf discs infiltrated with agrobacteria containing an empty vector (Fig. S1b). As observed for the flg22 response, Foa2 and Foa3 strongly reduced chitin-induced ROS production, indicating that these two effectors target a shared plant signalling component that is required to respond to both PAMPs (Fig. 2c). Interestingly spFoa1 showed an opposite effect; it suppressed flg22-triggered, but enhanced chitin-triggered ROS (Fig. 2b,c). In conclusion, while two intracellular Fo effectors robustly interfered with both flg22-induced and chitin-induced ROS production, the extracellular spFoa1 effector differentially affected the ROS response to these PAMPs.
Although clearly accumulating, Six4, Six8, Six9 and Foa4 did not alter flg22-induced or chitin-induced ROS production. However, a marked chlorosis was observed in spSIX4-expressing leaf areas that was absent in sectors expressing dspSIX4 or in the empty vector controls (Fig. S2). This finding suggests an extracellular activity of Six4, other than suppressing PTI. In conclusion, half of the investigated candidate effectors of the Arabidopsis-infecting Fo strain negatively affected an early PTI signalling output in N. benthamiana leaves triggered by flg22 or chitin. One candidate triggers chlorosis, revealing a potential function in the host for these five proteins.

Foa2 impairs mitogen-activated protein kinase (MPK) activation and increases susceptibility to Pst
The transient expression screens in N. benthamiana revealed that multiple Fo effectors function as suppressors of early PTI signalling. To validate this finding, we focused on Foa2, which showed the strongest ROS burst suppression among the intracellular-acting Fo effectors. Transgenic Arabidopsis plants were generated that stably expressed dspFoa2. Two independent, single insertion lines were selected and Western blotting revealed higher Foa2 levels in line 6-6 than in 5-1 (Fig. 3a). First, it was tested whether flg22-triggered or chitin-triggered ROS generation was altered in the dspFoa2 Arabidopsis lines. Wild-type plants showed a clear ROS burst, whereas bak1-5 and cerk1-2 did not respond to flg22 and chitin, respectively, confirming the specificity of the assay (Fig. 3b). Both dspFoa2 lines showed a strongly reduced ROS accumulation in response to either PAMP as compared with wild-type plants, in line with the findings of the transient expression assays (Fig. 3b). Moreover, the level of suppression seemed proportional to the effector accumulation, pointing to a dose-dependent activity.

Research
New Phytologist chitin-triggered accumulation of phosphorylated MPKs was monitored by Western blot. Consistent with previous reports, two bands corresponding to MPK3 and 6 were detected in wildtype plants (Saijo et al., 2009). Their signal appeared 5 min after PAMP treatment and was further intensified after 15 min (Fig. 3c). The flg22 response was largely absent in bak1-5, whereas cerk1-2 plants showed MPK activation after 5, but not after 15 min of chitin exposure, indicating residual responsiveness of this line. Importantly, MPK activation was markedly reduced in dspFoa2 plants in response to both flg22 and chitin at both the 5 min and the 15 min time points (Fig. 3c). Altogether, these results showed that Foa2 inhibits both flg22 signalling and chitin signalling at the level of ROS production and MPK activation.
Next, we investigated whether the compromised PAMP responsiveness in dspFoa2 plants led to enhanced susceptibility to a pathogen. For this, Arabidopsis wild-type, two dspFoa2 lines and bak1-5 mutants were spray inoculated with the bacterial pathogen P. syringae pathovar tomato (Pst). At 4 d after inoculation bacterial titres were measured. In agreement with previous studies, bak1-5 plants showed more than a 10-fold higher bacterial count than wild-type plants, this increase is in line with a compromised PTI response (Fig. 3d) (Schwessinger et al., 2011). Importantly, both dspFoa2 lines showed similarly increased bacterial titres as bak1-5, suggesting that PTI is disabled by Foa2 (Fig. 3d). In summary, the stable transgenic dspFoa2 lines revealed that this effector targets a component in PTI signalling that is required for at least two different PAMP receptors, resulting in enhanced pathogen susceptibility. Furthermore, this result showed that an effector candidate selected by its PTI-suppressing activity in N. benthamiana, is also a potent PTI suppressor in the endogenous host of Fo.

Host cell entry of Fo effectors during Arabidopsis root colonisation
The PTI-suppressing activities of Six1 and Foa1 indicate that these two effectors act on an extracellular host target, whereas Foa2 and Foa3 function intracellularly and thus have to enter www.newphytologist.com plant cells during root colonisation by the fungus. To investigate this further, we assessed whether intracellular accumulation of Foa2 and Foa3but not of Six1 and Foa1is detectable in Fo colonised Arabidopsis roots. For this we adjusted an assay in which biotinylation of Avi-tagged effectors is monitored as a proxy for their presence inside living plant cells (Lo Presti et al., 2017). The Avi effector tag combined an HA epitope for detection by Western blot with a biotin ligase recognition peptide motif that is biotinylated by the biotin ligase BirA, a recombinant bacterial enzyme stably expressed in transgenic Arabidopsis (Deal & Henikoff, 2011). To compare uptake efficiency between different effectors in vivo, a synchronised infection and a relatively high amount of uniformly colonised roots are required. To generate such material, a semihydroponic system was developed in which sterile grown Arabidopsis seedlings were inoculated with Fo. First, we examined whether the fungus colonises the vasculature and causes typical disease symptoms in this setup. To monitor host colonisation a Fo reporter strain was used that carries a construct in which the promoter of the Fo5176 SIX1 effector gene was fused to the coding sequence of GFP. The green fluorescence observed upon Fo5176 pSix1::GFP inoculation revealed successful fungal colonisation of the root tip, outer tissue layers and the vasculature of the plants (Fig. 4b). In line with the observed root colonisation, 5-7 dpi the seedlings developed clear disease symptoms such as root browning and leaf chlorosis starting at the veins, which was followed by tissue necrosis at later stages (Fig. 4c). These symptoms were comparable with those observed in soil-grown plants inoculated with Fo (Thatcher et al., 2009), validating the setup of the in vitro bioassay.
To assess effector uptake during root colonisation, Fo5176 strains expressing individual Avi-tagged effectors were generated by Agrobacterium-mediated transformation. As the SIX1 promoter was found to be active throughout different stages of root colonisation (Fig. 4b), each effector was expressed from this promoter. In addition to the eight Fo5176 candidate effectors three controls were included: FolAvr2, which has been shown to function intracellularly (Houterman et al., 2009;Di et al., 2017), and two secreted Fo enzymes: a glycosyl hydrolase and an amidase that are both expected to function in the apoplast, thereby serving as negative controls for the uptake assay. Fo transformants were selected based on hygromycin resistance. Next, we tested whether tagged effectors were detectable in protein lysates of roots inoculated with the transformed fungi. For each construct three to five independent transformants were used to inoculate roots of Arabidopsis seedlings. Following protein isolation and separation on SDS-PAGE, effector presence was detected by Western blotting and probing with an anti-HA antibody. A specific band was detected for each tagged effector protein in at least one of the transformants, except for Six4 where none of the isolates showed a signal at the expected size (Fig. S3). In conclusion, except for Six4, the candidate effectors were readily detectable in protein extracts upon root colonisation by Fo, allowing assessment of their potential internalisation by the host cells.
To reveal which effectors were biotinylated during root colonisationa proxy for their uptake and accumulation inside host cells -BirA-expressing plants were inoculated with Fo strains producing Avi-tagged effectors. Roots were harvested at 6 dpi when the first disease symptoms in the leaves started to emerge, but widespread tissue necrosis was not yet apparent. Protein lysates from inoculated roots were subjected to streptavidin pulldown assays, followed by anti-HA Western blotting. An aliquot of the total lysate was blotted to verify accumulation of the tagged proteins in the root. To confirm that effector biotinylation solely relies on plant-produced BirA, wild-type Arabidopsis was inoculated with Fo5176 expressing Avi-tagged Six8. As expected, only when isolated from BirA plants, and not wild-type plants, Six8 showed binding to streptavidin (Fig. S4).
To compare the streptavidin pulldown efficiency between the different effectors and the controls, band signal intensities were quantified and their recovery after pulldown was normalised to the total accumulation levels of each protein (numbers shown above the streptavidin pulldown blots, Fig. 5a). Streptavidin beads effectively pulled down FolAvr2, whereas they did not pull down the negative control, the glycosyl hydrolase (Fig. 5a). This result showed that the assay is suitable to discriminate between host-translocated and apoplastic proteins. In addition to FolAvr2, a clear signal upon streptavidin pulldown was detected for Six8, Foa3, Foa2 and Foa4, indicating their uptake by host cells (Fig. 5a). This result is in line with the intracellular PTI-suppressing activity of Foa2 and Foa3, and with our finding that Six8 targets a plant transcriptional regulator (unpublished). The pulldown efficiency for Six9 was much lower than that of the putative cytoplasmic effectors, but higher than that of the apoplastic effectors, preventing to draw a firm conclusion on the potential localisation of this effector. Both presumed apoplastic effectors, Six1 and Foa1, showed a much weaker, yet detectable, signal after streptavidin pulldown (Fig. 5a). This signal could indicate that host cells internalised trace amounts of these two proteins, or that a minor fraction of extracellular protein was biotinylated, generating a weak nonspecific background signal. To address this concern, as an additional control mixed inoculations with strains expressing the translocated effector Foa3 and a negative control were performed. For this we used a secreted amidase from Fo, as this protein shows much higher total accumulation and thus provides a more suitable negative control. Although a weak signal after streptavidin pulldown was detected for the negative control, its signal intensity was much lower than for the translocated effector Foa3 (Fig. 5b). Altogether, the results of this translocation assay indicated that a subset of Fo effectors (Foa2, Foa3, Six8 and Foa4) accumulated inside host cells during infection, and that the putative apoplastic effectors (Six1 and Foa1) and extracellular enzymes did not, and remained extracellular.

Protein size and charge do not explain effector internalisation
Based on the presented data, the putative Fo5176 effectors could be classified as either cytoplasmic or apoplastic. Next, we investigated whether features could be identified that are specifically associated with either category. A conserved sequence pattern New Phytologist (2020) 227: 1479-1492 Ó2020 The Authors New Phytologist Ó2020 New Phytologist Trust www.newphytologist.com

Research
New Phytologist comparable with the RxLR-dEER motif of oomycete effectors or the Y/F/WxC motif of powdery mildew effectors was not apparent in the cytoplasmically localised Fo effectors. Calculated isoelectric points ranged from 6.07 to 6.81 for the cytoplasmic effectors and from 5.49 to 8.75 for the apoplastic effectors. This overlap indicated that protein charge was also not associated with the ability of these proteins to enter plant cells (Table 1). Notably, all five Fo effectors that accumulated intracellularly were 20 kDa in size, or smaller, whereas the apoplastic proteins were significantly larger. This finding may imply that uptake in the Fo Arabidopsis interaction is restricted to small proteins. To test this hypothesis, we used the Ustilago maydis cytoplasmic effector Cmu1 (Djamei et al., 2011), a 290 amino acid protein of similar size to the apoplastic Fo effectors Six1 and Foa1, in our root assay. Transgenic Fo5176 strains producing Avi-tagged Cmu1 were generated and used in the effector translocation assay. When produced by Fo during plant infection Cmu1 clearly accumulated, and could effectively be pulled down by streptavidin, indicating successful host cell entry of the protein (Fig. 5c). As a negative control for host cell entry, uptake of the secreted amidase was assessed. Even though it also accumulated to high levels in Fo-inoculated roots, the amidase was barely detected upon streptavidin pulldown, and thus mostly remained extracellular (Fig. 5c). Together, these results indicated that protein size per se does not determine effector uptake properties.

Discussion
This study describes a systematic characterisation of effectors of an Arabidopsis root-and xylem-infecting strain of Fusarium oxysporum. Four out of eight candidate effectors investigated were able to suppress the flg22-induced or chitin-induced ROS burst an early PTI responsein leaves of N. benthamiana. This result implied that these Fo effectors acted on (a) target(s) that were present in both root and foliar cells and were evolutionary conserved between plant species. The observed redundancy among effector activity implied that targeting of PTI components is important for pathogenicity of this vascular pathogen. Furthermore, more than half of the studied effectors were found to function inside host cells, and, in correspondence, these could be biotinylated by a plant-produced cytosolic BirA enzyme. Together, our findings show that several small proteins secreted by Fo during Arabidopsis root colonisation have properties that allow them to enter host cells and manipulate PTI. Six1, Foa1, Foa2 and Foa3 from Fo5176 showed suppression of the flg22-induced oxidative burst, a response that was detectable within minutes upon PAMP application, implying that Fo effectors intersected PTI signalling at an early step. The protein-derived flg22 and the oligosaccharide chitin are perceived by different classes of receptors and co-receptors, however their signalling converges on a subset of receptor-like cytoplasmic kinases (RLCKs) (Couto & Zipfel, 2016). Thus, it is conceivable that the intracellularly acting Avr2, Foa2 and Foa3 effectors impaired early signalling by targeting the cytosolic part of receptor complexes and/or the associated RLCKs. By contrast, Foa1 and Six1 acted in the apoplast, thus their ability to interfere with BirA seedlings inoculated with Fo transformants expressing Avi-tagged effectors or enzymes (indicated above the blots). Roots from 15 or 16 seedlings were harvested at 6 d post inoculation (dpi) and biotinylated proteins were isolated from total protein lysates (top panel) using pulldown assays with magnetic streptavidin beads (lower panel). Protein blots were probed with anti-HA antibodies. To allow direct comparison of the luminescence signals both blots were simultaneously exposed to the same film. The molecular weight (kDa) is indicated at the left side of each membrane. (b) Mixed and single inoculations with Fo transformants expressing the Avi-tagged translocated effector Foa3 and the Avitagged negative control amidase, followed by isolation of biotinylated proteins and anti-HA Western blotting. (c) Inoculation with an Fo transformant expressing the Ustilago maydis effector Cmu1. The intensities of specific bands on both the input and pulldown blots were quantified and the band intensity after pulldown was normalised to the one in the input (shown above the lower blots, given as % of total). Of note, streptavidin pulldown samples are c. five-fold enriched as compared with the input samples.

Research
New Phytologist PRR function must rely on a different mechanism. Previously investigated examples of apoplastically acting fungal effectors were Cladosporium fulvum ECP2 that sequesteres fungal chitin oligomers that are released by the activity of plant-produced chitinases (de Jonge et al., 2010), and Avr4 that prevents chitin degradation (van den Burg et al., 2006). However, such a mechanism would not explain the effect of the apoplastic Fo effectors on flg22-induced signalling. A putative mechanism on how they interfere with PAMP signalling is that these effectors target apoplastic proteases that process and activate secreted plant peptides called Rapid Alkalinisation Factors (RALFs) (Masachis et al., 2016;Stegmann et al., 2017). Currently, it is unknown why Fo possesses an effector (SpFoa1) that targets flg22-induced, but not chitin-induced, signalling. It is conceivable that this effector interferes with recognition of, yet unidentified, Foderived PAMPs that share components with flg22 perception. Existence of at least one such PAMP was recently reported (Coleman et al., 2019). Having identified a set of four sequence-unrelated Fo effectors, acting at different locations and all counteracting PTI signalling, provides a unique set of molecular probes to dissect early immune signalling and to identify the proteins involved.
The observed redundancy of Fo effectors targeting the same process implied that interference with this process is a crucial element in host colonisation by this vascular pathogen. Intriguingly, a recent study revealed highly localised and damage-dependent activation of PTI in Arabidopsis roots (Zhou et al., 2020). Initial root surface colonisation by Fo is followed by its invasion of the xylem vessels that is likely to result in plant tissue damage and thus PTI activation, explaining the necessity for these PTI-suppressing effectors. Whether these are produced and secreted in a highly localised manner during certain stages of the interaction is currently unknown.
Numerous fungal and oomycete effectors have been shown to function inside plant cells (Lo Presti et al., 2015;Whisson et al., 2016). However, demonstrating effector uptake during pathogen colonisation remains technically challenging. Lo Presti and colleagues developed an assay that detected effector biotinylation as a proxy for an intracellular protein location in planta. This assay was developed to analyse the localisation of effectors upon infection by the leaf-invading pathogen Ustilago maydis (Lo Presti et al., 2017). We adapted this method to elucidate which Fo effectors and/or enzymes were internalised into plant cells during root colonisation. Only a very weak signal indicating biotinylation of the two apoplastic effectors (Six1 and Foa1) and the two secreted enzymes (glycosyl hydrolase and amidase) was observed. It is unclear whether this finding reflects a small fraction of apoplastic proteins that entered host cells, or rather BirA leaking from damaged cells into the apoplast. Importantly, a much stronger biotinylationindicating uptakeof the intracellularacting Avr2, Foa2, Foa3 and Six8 effectors was observed, supporting the specificity of the assay. In conclusion, the in planta biotinylation assay could be used to distinguish between intracellularly and extracellularly localised effectors of a fungal root pathogen. Whether Fo effectors are internalised by host cells from the xylem sap, from the root apoplast, or both, and whether close contact between the fungus and plant cells is required, as for instance in Magnaporthe oryzae (Khang et al., 2010), remain questions for future research.
Another intriguing question is which features of effector proteins determine their uptake by plant cells. No apparent consistent difference in size, charge or amino acid composition (e.g. number of cysteine residues) was detected when comparing cytosolic and apoplastic effector proteins. Also, no obvious motifs such as RxLR, CRN or CHX were discernible in the translocated effectors, implying another mechanism directing their fate in the host. M. oryzae and Phytophthora infestans are proposed to use an unconventional, Golgi-independent, secretory pathway for effectors that are destined for host cell entry (Giraldo et al., 2013;Wang et al., 2017). If Fo similarly uses a special secretory pathway for translocated effectors, then the information for entering this pathway must be embedded within the protein sequence, as all effectors were expressed from the same promoter and carry the same UTR sequences.
One mechanism resulting in an apparent specific uptake would be the selective retention of apoplastic proteins in the extracellular space, for instance by binding to the cell wall, and nonspecific uptake of all other proteins. Alternatively, translocated effectors might possess a special capacity that enable them to enter host cells. Our observation that all three intracellular-acting effectors (Avr2, Foa2 and Foa3) suppress the flg22-triggered ROS burst to a similar extent when expressed with or without their endogenous signal peptide are in support of this hypothesis. However, as these effectors do not share sequence homology nor contain obvious motifs or domains, their host cell entering activity might depend on hidden features in their structures, or alternatively they might use different entry mechanisms. Determining their 3D structures, as already done for the intracellularly acting Avr2 protein (Di et al., 2017), might reveal shared folds or elements that correlate with the ability of effector proteins to enter host cells. The set of effectors acting at different locations in the host characterised here provides a good starting point for such a structural study. Understanding the mechanism of how Fo effectors target plant immunity and are translocated from the extracellular spaces into host cells, may provide new leads for strategies to combat fungal plant diseases.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.    Table S1 Primers used in this study.