Cell death is induced by the expression of XopL_Xe but not XopL_Xcc

As mentioned, the activity of XopL from Xe 85–10 (XopL_Xe) is easily monitored in the context of transient expression assays due to a robust cell death reaction that is visible 5–7 days post inoculation (dpi); a reaction that requires a functional E3 ligase domain (XL-box) plus the LRR, the putative protein interaction domain [12]. Using the same model system it was independently reported that cell death was not induced by XopL from Xcc 8004 (XopL_Xcc) [16], a first indication of possible divergence in XopL activity between different xanthomonads. To confirm these results in our experimental conditions, translational XopL-GFP fusions were generated and expressed side-by-side in N. benthamiana leaves. Cell death was recorded at 5 dpi (Fig 1A); our results confirmed the results of Singer et al. [12] and Yan et al. [16]. We observed a strong cell death reaction in tissue overexpressing XopL_Xe but not XopL_Xcc, despite similar protein levels (Fig 1B). This supports the hypothesis that the two XopL homologs have different activities in planta.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. E3 ligase activity is conserved among XopL homologs from different xanthomonads.

(A) Agroinfection of N. benthamiana leaves to express GFP-myc and XopL proteins from Xanthomonas strains Xe 85–10 (XopL_Xe) and Xcc 8004 (XopL_Xcc) fused to a C-terminal GFP (OD of 0.8). Plant reactions were monitored 5 dpi. (B) Western blot analysis of protein extracts isolated 2 dpi from samples depicted in (A). One ‘not inoculated’ (NI) sample was included as a negative control. Signals were detected using a GFP-specific antibody (*). The left side of the blot shows protein mass in kDa. Amido black staining is shown as a loading control. (C) Schematic of the UbiGate plasmid used to test for E3 ligase activity of XopL proteins. Components of the A. thaliana ubiquitination machinery (affinity-tagged for detection by western blot), UBIQUITIN 10 (HIS:Ub), UBIQUITIN ACTIVATING ENZYME 1 (UBA1) and UBIQUITIN CONJUGATING ENZYME 28 (GST:UBC28) were expressed from a single, IPTG-inducible plasmid together with xopL coding sequences. Each translational unit was equipped with an independent ribosome binding site (red). (D-F) Western blot analysis of protein extracts isolated from E. coli expressing different XopL proteins. Controls were samples lacking the E3 ligase (E3-) or the E2 enzyme (E2-). The XopL protein tested is indicated above the designated lanes. (D) Ubiquitin was detected using the P4D1 antibody. Polyubiquitin chains are indicated by a black line. E2-ubiquitin (Ub+E2) conjugates are visible in E3- control sample (purple line). (E) C-terminally tagged XopL proteins were detected with a myc-specific antibody (expected size indicated with ‘*’). In some cases, autoubiquitination is detectable (Ub+XopL). (F) N-terminally-tagged E2 enzyme was detected using a GST-specific antibody. Samples were run on different gels to clearly visualize ubiquitin and other proteins. Protein mass is expressed in kDa.

https://doi.org/10.1371/journal.ppat.1011263.g001

Enzymatic activity of four XopLs is conserved

As plant cell death induced by XopL_Xe is known to depend on its E3 ligase activity [12], we wondered whether XopL_Xe and XopL_Xcc differ in enzymatic activity. For this, ubiquitination assays were performed with XopL_Xcc and XopL_Xe. As a general test for E3 ligase conservation, XopLs from two additional species were also included: XopL from X. citri pv. citri (XopL_Xac), which causes citrus canker disease and is closely related to Xe (non-vascular pathogens), and the more distantly related Xoo (XopL_Xoo), the causal agent of leaf blight on rice (vascular/systemic pathogen). xopL coding sequences from Xoo strain PX099A and Xac strain 306 were synthesized and codon-optimized for N. benthamiana. For the sake of comparability (see Materials and Methods), XopL_Xe and XopL_Xcc codon optimized versions were also synthesized for this experiment. E3 ligase activity was tested in the UbiGate system, a modular system for the expression of the plant ubiquitination machinery in E. coli [20]. Using UbiGate one can screen for E3 ligase activity of a protein of interest without any protein purification steps. The ability of XopL_Xe to efficiently generate polyubiquitin chains in vitro was previously demonstrated by Singer et al [12]. In this publication the LRR was found to be dispensable for E3 ligase activity in vitro and only the XL-box is required. As a proof of concept, we tested several XopL variants for their ability to generate ubiquitin chains in UbiGate: wild-type XopL, XopL_m, a variant including the N-terminus, α-helical region and the LRR (NTαLRR, 1–450 aa) and a truncation with the XL-box alone (451–660 aa). UbiGate assays showed that when co-expressed with the plant ubiquitination machinery in E. coli (ubiquitin, E1 and E2 enzymes; plasmid map in S1A Fig) XopL_Xe-dependent ubiquitin chain formation is observed, as evidenced by laddering of ubiquitin in samples with XopL_Xe and this laddering is not observed in the E2- or Ub- controls (S1B Fig). Just as in experiments by Singer et al., [12], expression of the XL-box also results in extensive ubiquitin laddering, while this does not occur in case of XopL_m or NTαLRR (S1B Fig). Interestingly, XopL_m appears to have some minimal residual enzymatic activity, indicated by a small smear in the ubiquitin blot between 100–180 kDa. The residual E3-ligase activity detected seems to occur mostly as autoubiquitination, as the smear is also visible in the α-myc blot (S1C Fig). XopL_m expression does not lead to the formation of free ubiquitin chains in lower molecular weight range (15–40 kDa) or high molecular weight ubiquitin chains which occur in the presence of the fully active wild-type and XL XopL derivatives (S1B Fig). Comparison of the results derived from the UbiGate system and from in vitro assays performed with purified XopL protein [12] confirmed that the UbiGate system is suitable for the analysis of XopL E3 ligase activity. Having validated our approach, UbiGate was then used to test for E3 ligase activity of XopLs from the four xanthomonads and revealed that they were all enzymatically active (ubiquitin ladder in all XopL-containing samples; Fig 1D). While this was already known for the closely related XopL_Xoo [17], this is new for XopL_Xac and XopL_Xcc. The UbiGate results also suggest that E3 ligase function is conserved, even among distantly related XopLs, and that differences in phenotypes of plant tissue expressing XopL_Xe and XopL_Xcc are unlikely due to differences in enzymatic activity.

XopL_Xe tag position impacts cell death

Next, we examined the in planta subcellular localization of XopL homologs. As mentioned above, localization studies have independently shown that XopL proteins from different Xanthomonas species can differ in their localization [13, 14, 17, 18], offering a possible explanation for differences in XopL-mediated plant phenotypes.

XopL_Xe was chosen as a starting point for in-depth localization experiments because the functionality of different florescence protein fusions could be easily evaluated using cell death as a readout which is readily observed when XopL is tagged with a C-terminal GFP (Fig 1A). To ensure that localization did not differ depending on tag position, N-terminal fusions were also tested. C- and N-terminal GFP fusions to XopL_Xe were tested side-by-side with the untagged XopL protein, as well as myc-GFP and the E3 ligase mutant (XopL_m) as negative controls (Fig 2A). These experiments resulted in strong cell death reactions at 5 dpi in tissue expressing untagged XopL or C-terminally-tagged XopL_Xe. However, cell death was severely reduced in case of an N-terminal fusion. As expected, expression of XopL_m and GFP controls showed no visible cell death. Although reduced cell death by the GFP-XopL fusion could be due to reduced protein level (Fig 2B), the difference in phenotype to untagged XopL might also indicate reduced/altered activity of GFP-XopL_Xe.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. XopL_Xe-induced cell death and MT-localization are negatively affected by an N-terminal fluorescence tag.

(A) Analysis of cell death induction following agroinfection of N. benthamiana leaves to express untagged XopL_Xe and XopL_m (E3 ligase-inactive, H584A, L585A, G586E) N- or C-terminally fused to GFP and GFP alone (OD₆₀₀ of 0.4). Plant reactions were monitored 5 dpi. (B) Western blot analysis of protein extracts isolated 2 dpi from the experiment depicted in (A). Protein signals were detected with a GFP-specific antibody (*). The left side of the blot shows protein mass in kDa; # indicates a non-specific signal. Amido black staining is included as a loading control. (C-G) Confocal microscopy pictures of lower epidermal cells 2 dpi from the experiment depicted in (A). Images depict the subcellular phenotypes of (C) XopL_Xe-GFP, (D) a lower expressing cell from the same inoculation spot as in (C) (imaging with the LSM 780 required a laser intensity of 16.8% laser power vs. 10% used in (C), with identical gain settings), (E) GFP-XopL_Xe, (F) XopL_m-GFP, (G) GFP-XopL_m and (H) myc-GFP. GFP-labeled protein fusions are visible in magenta, plastid autofluorescence in yellow. Nuclei are labeled with ‘n’, blue arrows label mobile dots, white arrows point out microtubules and orange arrows label cytosolic strands. Scale bars are 20 μm for all images except (D), where the scale is 10 μm.

https://doi.org/10.1371/journal.ppat.1011263.g002

XopL_Xe labels MTs

Confocal microscopy of XopL_Xe-GFP fusions expressed in N. benthamiana leaves showed that, as previously reported [18], the C-terminal fluorescence protein fusion localizes to both mobile dots (blue arrows), as well as to the nucleus and cytosol (Fig 2C), a pattern that the N-terminal fusion also shows (Fig 2E). The expression of both N- and C-terminal XopL fusions also induced the typical plastid clustering phenotype that we previously characterized in detail [18]. The E3 ligase mutant, XopL_m, also localized to the nucleus and cytosol but additionally labeled filaments (Fig 2F and 2G). Both C- and N-terminal tagged XopL_m localized similarly, albeit less clearly for the N-terminally tagged version. It is known from our published work that the filaments labeled by XopL_m are microtubules (MTs) [18]. As expected, the GFP control was localized to the cytosol and nucleus (Fig 2H).

While at first glance it appeared that wild-type XopL_Xe and XopL_m proteins differed in their localization pattern, upon closer examination a few MTs were also visibly labeled by XopL_Xe in some cells (Fig 2C–2E, white arrows). It should be noted that in agroinfection assays epidermal cells are a mosaic of high expressing and lower expressing cells. In cells with low XopL_Xe expression MT-labelling by XopL_Xe was even more apparent (Fig 2D). This experiment told us three things: (i) wild-type XopL_Xe has a similar localization pattern to XopL_m, (ii) the reduction in visible MT labeling in cells with high XopL_Xe expression suggests that XopL_Xe may disrupt MTs and (iii) abundant labeling of MTs in cells expressing XopL_m compared with XopL_Xe suggests that MT disruption occurs in an E3 ligase activity-dependent manner.

To confirm the identity of XopL-labeled filaments and to determine the fate of MTs bound by XopL_Xe, XopL-mCherry fusions were expressed in leaves of an N. benthamiana transgenic line with labeled MTs (GFP-TUA6; TUBULIN ALPHA-6 from Arabidopsis thaliana [21]). C-terminal XopL fusions were chosen for further localization studies since both C and N-terminal fusions exhibited similar sub-cellular phenotypes (i.e., plastid clustering) and localization patterns. In addition, a reduction in cell death compared to untagged XopL_Xe and weaker MT binding of N-terminal fusions suggested that an N-terminal tag interferes with XopL activity. XopL_Xe as well as the XopL_m expression in tubulin-labeled plants confirmed the co-localization of these proteins with MTs (Fig 3), which does not occur when cells express the mCherry protein alone (S2 Fig). In addition, as suspected, wild-type XopL_Xe clearly reduced MT number compared to XopL_m (Fig 3A compared to 3D) and the mCherry control (S2A Fig). Intensity plots across the few remaining MTs demonstrate the overlap of the XopL_Xe-mCherry and GFP-TUA6 signals (Fig 3K and 3L), which further confirms that XopL co-localizes with MTs. In contrast, control intensity plots in cells that did not express XopL (Fig 3J) or that expressed mCherry (S2G Fig), did not show co-localization. Together with data shown in Fig 2, this confirms that XopL_Xe both localizes to MTs and triggers an E3 ligase-dependent decrease in MT abundance after agroinfection. These findings provide a plausible explanation as to why XopL_Xe localization appeared to be limited to nucleus, cytosol, and dots in previous experiments [13, 18]. The decrease in MT number is not likely due to a decrease in tubulin subunits because α and β-tubulin levels were stable at 1 and 2 dpi; this time point is well within the timeframe of MT destruction (Fig 3M–3O).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. XopL_Xe destabilizes MTs but does not degrade tubulin.

(A-I) Confocal microscopy pictures of lower epidermal cells of GFP-TUA6 (labels MTs) transgenic N. benthamiana leaves. Leaves were agroinfected (OD₆₀₀ of 0.4) to express (A-C) XopL_Xe-mCherry and (D-F) XopL_m -mCherry. Samples were harvested at 2 dpi for microscopy. The GFP channel is visible in white (MTs), the mCherry channel in magenta (XopL derivatives). (C) and (F) are merged GFP and mCherry channels. Plastids are visible in cyan, examples of MTs are labeled with white arrows. Scale bars are 20 μm. Magnifications from (C) are shown in (G-I); yellow lines represent the location of intensity plot measurements. (J-L) Intensity plots. (J) Shows a plot of a neighboring cell without XopL_Xe expression (negative control). (K-L) Measurements across filaments in a XopL_Xe-expressing cell show co-localization. The signal from the chlorophyll channel was included as a negative control (cyan lines). (M-O) Western blot analysis of protein extracts isolated from N. benthamiana tissue that was not inoculated (NI) or agroinfected (OD₆₀₀ of 0.4) to express the myc-mCherry control (Ch) or XopL derivatives (XopL_nt = untagged, XopL_Xe and XopL_m with C-terminal myc tags) 1 and 2 dpi. (M) Expression of proteins fusion was confirmed via detection with a myc-specific antibody. Expected size of myc-tagged proteins is marked with ‘*’. (N and O) The abundance of tubulin monomers was tested using antibodies specific to endogenous α and β tubulin. Expected tubulin size is marked with ‘<‘. The left side of the blot shows protein mass in kDa. Ponceau and amido black staining of tubulin blots were used as loading controls.

https://doi.org/10.1371/journal.ppat.1011263.g003

Taxol partially rescues MTs from XopL_Xe

To determine whether MTs could be rescued from XopL_Xe-mediated depolymerization we treated agroinfected tissues with the MT stabilizing agent taxol. Taxol reversibly binds to MTs via β-tubulin, altering its conformation to increase the rigidity and stability of filaments, independent of GTP binding, which is normally required for polymer growth [22]. Taxol treatments (dissolved in DMSO, see Materials and Methods) were administered at very early time points (4 h) following agroinfection, and microscopy was performed 2 dpi, when XopL_Xe expression typically has eliminated most MTs. The result of taxol treatments on XopL_Xe-inoculated tissues was dramatic. MTs were indeed partially rescued and those that remained were highly bundled and unorganized, often forming loops (S3A Fig), all clearly labeled by XopL_Xe-mCherry (S3B and S3C Fig). DMSO controls at the same time point showed almost no MTs, as expected (S4A–S4C Fig). Treatment of the XopL_m and mCherry controls with taxol clearly showed an increase in MT density and in the number of parallel filaments relative to the control (XopL_m taxol treatment in S3D Fig compared to DMSO in S4D Fig; mCherry control treated with taxol in S3K Fig compared to DMSO in S4G Fig) illustrating the effectiveness of the taxol treatment. Taxol treatments of XopL_Xe expressing cells resulted in a distinct ‘spaghetti-like’ conformation of MTs when compared to XopL_m (compare S3A and S3D Fig). Taken together, taxol treatments not only provide further evidence for MT association by wild-type XopL_Xe, but differences in MT appearance between XopL_Xe- and XopL_m-expressing tissues treated with taxol also suggests that there are E3 ligase-dependent alterations to MT dynamics and/or structure that are not completely overcome by taxol.

MT localization is not conserved between XopL_Xe and XopL_Xcc

Having confirmed MT binding of XopL_Xe in planta using multiple constructs, plant genotypes and taxol treatments, the next step was to determine if the same pattern could be observed in tissues expressing XopL_Xcc. For this purpose, translational XopL_Xcc-mCherry fusions were expressed in the GFP-TUA6 plants, allowing the simultaneous examination of protein localization and MT abundance. Like XopL_Xe, XopL_Xcc was localized to the nucleus and the cytosol (Fig 4B), however, in contrast to XopL_Xe, it did not label MTs, nor did it trigger MT depolymerization (Fig 4A–4E). In addition, XopL_Xcc did not induce plastid clustering as was seen with XopL_Xe (Fig 4B compared to Figs 2C and 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. XopL_Xcc does not localize to MTs.

Confocal microscopy pictures of lower epidermal cells of GFP-TUA6 (labels MTs) transgenic N. benthamiana leaves. Leaves were agroinfected (OD₆₀₀ of 0.4) to express XopL proteins translationally fused to a C-terminal mCherry. Samples were harvested 2 dpi for microscopy. The GFP channel is visible in white (MTs), the mCherry channel in magenta (XopL derivatives) and plastids in cyan. (A-B) Plant cell expressing XopL_Xcc-mCherry. Scale bars are 20 μm. Insets from (A) and (B) (area outlined with a white box) are magnified in (C) and (D), respectively. (F-G) show a cell expressing XopL_Xac-mCherry and (I-J) XopL_Xoo-mCherry. (L-M) is the mCherry control. Scale bars for (C-M) are 5 μm. Examples of MT labeling are indicated by white arrows. (E, H and J) Fluorescence intensity plot across MTs to test for co-localization of GFP-TUA6 with (E) XopL_Xcc, (H) XopL_Xac and (K) XopL_Xoo. Magenta lines show the intensity of the XopL-mCherry signal, black represents GFP-TUA6. (N) Western blot analysis of protein extracts isolated 2 dpi from tissue used for microscopy in (A-M). Signals were detected using an mCherry-specific antibody (*), with exception of XopL_Xac () which was not detectable. Samples that were not inoculated ‘NI’ or expressing mCherry ‘Ch’ were loaded as controls. The left side of the blot shows protein mass in kDa. Ponceau staining shown as a loading control.

https://doi.org/10.1371/journal.ppat.1011263.g004

MT localization is conserved among XopLs from Xoo and Xac

The inability of XopL_Xcc to bind and trigger disassembly of MTs raised the question of whether MT-association/destabilization is a conserved characteristic of XopLs from other xanthomonads or specific to XopL from Xe 85–10. Therefore, codon-optimized XopLs from Xoo and Xac were fused in frame with a C-terminal mCherry for localization studies in GFP-TUA6 stable transgenic N. benthamiana plants. At 2 dpi, MTs were already degraded in the presence of XopL_Xac and XopL_Xoo, but not in the presence of mCherry (Fig 4F–4M). In some cases, a few MTs remained and were labeled by XopL_Xac and XopL_Xoo, demonstrating again that they also localize to MTs (white arrows, Fig 4G and 4J and intensity plots in Fig 4H and 4K). As previously reported, XopL_Xoo was also cytosolic [17] and appeared to be nuclear-excluded (S5B Fig), whereas XopL_Xac was cytosolic and nuclear (S5A Fig). This means that, while nuclear localization varied, MT binding and destabilization is a conserved feature of several XopLs originating from Xanthomonas species closely related to Xe, but not the more distantly related Xcc. A simple explanation for why previous reports on XopL_Xoo [17] and XopL_Xe [13] overlooked MT localization is that MTs were likely dismantled at the time points chosen after agroinfection (2–3 dpi). The conservation of MT localization among XopLs from multiple Xanthomonas species suggests the potential functional importance of this characteristic. In contrast, but equally exciting is the finding that the more distantly related XopL_Xcc does not exhibit MT localization, thus providing a unique chance to evaluate how the differences in localization may have evolved and how it affects effector function.

Unfortunately, codon optimization of XopLs for N. benthamiana resulted in very poor expression levels (Fig 4N). While proteins were sufficiently expressed for microscopy, expression was patchy and low with little signs of macroscopic cell death (S5C Fig), even in the case of XopL_Xe and XopL_Xoo, where strong cell death has been well documented in agroinfected N. benthamiana [12, 17]. For this reason, codon-optimized XopLs were not used to evaluate cell death phenotypes going forward.

The XL-box from XopL_Xe is not required for MT binding

As mentioned, XopL_Xe is composed of an α-helical region (3 alpha helices; herein referred to as ‘α’), LRR domain (LRR) and XL-box (XL) which have been crystalized, as well as an unstructured N-terminus (NT), which harbors the type III translocation signal [12]. To get a rough idea of which domains are important for the MT binding, XopL was divided into two parts, as was done by Singer et al. [12]. Amino acids 1–450 (referred to as NTαLRR) and 451–660 (XL) were separately fused to GFP and expressed in N. benthamiana via agroinfection (constructs outlined in Fig 5A). Anticipating that, if the XL-box was responsible for MT binding, it would also have the capacity to degrade filaments, we also included an E3 ligase-inactive version (451–660 aa, H584A_L585A_G586E; XL_m). The result was clear: NTαLRR-GFP was sufficient for MT association, whereas the XL-box was not required (compare Fig 5C to 5F and 5G). XopL NTαLRR was further divided into two parts, the NT (1–136 aa) and the αLRR (137–450 aa). Since the LRR domain is generally believed to be responsible for protein substrate binding, it was surprising that very little binding was observed for the αLRR (Fig 5E) and αLRR_XL_m (Fig 5H) derivatives. There was also no MT association in the case of NT expression (Fig 5D). Only the combination of NT plus the αLRR was sufficient to confer MT binding comparable to the full-length XopL_m protein (Fig 1E). All proteins were detectable at the correct size in Western blots (Fig 5L).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Comparative sequence analysis of XopL homologs identifies a proline-rich-region (PRR) important for MT localization.

(A) Domain structure of XopL_Xe and derivatives used to functionally analyze the PRR. Domains crystalized by Singer et al. (2013) are denoted with a white bar, LRR repeats by dark gray arrows. Amino acid exchanges in the XL-box of XopL_Xe to generate XopL_m are indicated. Single domain derivatives are annotated in yellow, multiple domain in orange, and ‘extension’ (ex) constructs in blue. Amino acids added to the αLRR_XL in ex1 and ex2 are shown at the N-terminus of the corresponding construct. Amino acid positions included in each derivative are shown in brackets. (B-K) Confocal microscopy pictures of lower epidermal cells of wild-type N. benthamiana leaves. Leaves were agroinfected (OD₆₀₀ of 0.4) to express (B) a GFP control, (C) NTαLRR, (D) NT, (E) αLRR, (F) XL, (G) XL_m, (H) αLRR_XL_m, (I) ex1_m; cell with MT localization, (J) ex1_m; cell without MT localization and (K) ex2_m. Subscript ‘m’ indicates a triple amino acid exchange in the E3 ligase domain (H584A, L585A, G586E) that renders the derivative E3 ligase-inactive. Samples were harvested at 2 dpi for microscopy. All XopL derivatives are tagged at the C-terminus with GFP (visible in magenta). Examples of MT association are indicated by white arrows. Scale bars are 5 μm. (L) Western blot analysis of protein extracts isolated 2 dpi from experiment in (B-K). Signals were detected using a GFP-specific antibody (*). The left side of the blot shows protein mass in kDa. (M) An alignment of XopL_Xcc (orange) and XopL_Xe (cyan) 3D structural models generated in AlphaFold2. The alignment includes the unstructured NT, α-helical (α1–3) and LRR regions. The cluster of prolines conserved in MT-localizing XopLs are labeled and represented in stick format. N- and C-termini are marked with N and C, respectively. Models do not include the C-terminal XL-box. (N) Alignment of XopL_Xcc, XopL_Xe, XopL_Xac and XopL_Xoo at the junction between NT and αLRR constructs. Amino acids are colored based on polarity (Geneious Prime), basic amino acids are in blue, and fuchsia highlights prolines. The boundary between the NT and αLRR constructs is annotated at amino acid 137. The sequence logo above the alignment shows sequence conservation at specific positions.

https://doi.org/10.1371/journal.ppat.1011263.g005

Comparative analysis of XopLs identified prolines contributing to MT binding stability of XopL_Xe

Differences in protein localization of XopL_Xcc and XopL_Xe prompted a comparison of amino acid sequences to narrow down the region of XopL_Xe that confers MT association. XopL_Xcc and XopL_Xe are 46.9% similar and 34.2% identical at the amino acid level, with most similarity lying within the LRR (70.5% similarity, 53.7% identity), followed by the XL-box (48.5% similarity, 32.7% identity) and then the α region (45.2% similarity, 29% identical). The LRR domain was previously predicted to be responsible for protein substrate recognition [12], and it was initially suspected it may be important for interaction with MTs, as opposed to the XL-box. While indeed the XL-box is not required, the finding that the αLRR was insufficient for XopL_Xe localization to MTs caused us to take a closer look at the NT sequences of these two proteins. NTs are the most variable region among XopLs and vary greatly in length between species [12]. AlphaFold2 structural predictions of the XopL_Xe and XopL_Xcc proteins resulted in very similar 3D models within the αLRR and highlighted the presence of a long unstructured NT in XopL_Xe that is absent from XopL_Xcc (Fig 5M). However, the shorter NT of XopL_Xcc is not sufficient to explain the lack of MT binding, since that of XopL_Xac is even shorter and is able to bind and destroy the MT network in a manner comparable to XopLs with extended NTs (XopL_Xe and XopL_Xoo; see alignment of all 4 XopLs in S6 Fig).

To identify potential sequence differences within the NT that contribute to MT association, the sequences of the three MT-binding XopLs were compared with the non-MT localized XopL_Xcc. Closer examination of the boundary between the NT and αLRR region revealed that the N-terminal boundary chosen for the αLRR construct (position 137) excluded 5 amino acids (MPPPV) that were highly similar in sequence to that of the short NT of XopL_Xac (MPPTV) (Fig 5M). Interestingly, two of these amino acids are prolines which were 100% conserved in MT-associated XopLs but are conspicuously absent from XopL_Xcc. In addition, these amino acids are part of a larger proline-rich region (PRR) present in all MT-binding XopLs (Fig 5N).

Suspecting that the PRR may facilitate XopL MT binding, the αLRR_XL_m-GFP fusion was extended N-terminally to include the MPPPV motif (132–660 aa), henceforth referred to as extension 1 (ex1_m). Additionally, a second construct including the entire PRR (MPVRPRMPPPVαLRR_XL_m, 127–660 aa) was generated (ex2_m). Excitingly, ex1_m reconstituted MT binding when compared to the αLRR_XL_m, and some cells showed clear MT localization (Fig 5I). However, MT localization was not consistent, and some cells showed nearly no labeling (Fig 5J). In contrast, ex2_m reliably reconstituted binding, resembling XopL_m localization (Fig 5K). This means, that with the addition of only 11 amino acids the αLRR_XL_m was transformed from a non-binding to a MT-binding derivative. The discovery of a conserved PRR in XopL is particularly exciting since PRRs are known to stabilize direct interactions with MTs in both animal [23] and plant MT-associated proteins [24].

XopL_Xe binds MTs in vitro

The fact that XopL_Xe localizes to MTs in planta, combined with the presence of a conserved PRR specific to XopLs that associate with MTs led to the question of whether this localization pattern is the result of a direct interaction with MTs, or rather an association with an MT-bound protein. To test for direct MT binding, co-sedimentation assays were performed. In this assay, MTs are polymerized in vitro, the protein of interest is incubated with MTs, and the mixture is subjected to ultracentrifugation to pellet MTs together with bound proteins. Thus, binding to MTs is indicated by the enrichment of the protein of interest in the pellet fraction. It should be noted that the protein of interest must be soluble in low salt and room temperature since these conditions are optimal for tubulin polymer stability (see Materials and Methods). In the case of XopL_Xe this provided a challenge, as XopL_Xe was not soluble in low salt conditions in our hands. However, solubility was optimized by expressing a fusion of XopL_Xe to 6xHis-StrepII-SUMO, pH optimization and the addition of aminocaproic acid (see S7A Fig for details).

The co-sedimentation assay revealed that XopL is present in the supernatant both when MTs are absent and present, albeit the XopL signal in the supernatant is slightly reduced when MTs are included (Fig 6A left panel). More exciting, however, was that XopL was enriched in the pellet fraction in the presence of MTs when compared to samples without MTs (Fig 6A right panel); a strong indication for direct MT binding. As a control to ensure that MTs were indeed polymerized western blots were Ponceau stained. This showed that tubulin was only detectable in the pellet fraction (Fig 6A, lower right panel). Even in the ponceau staining XopL enrichment in the pellet fraction was visible. It should be noted that in some cases we did detect some XopL in the pellet fraction in the absence of MTs, especially in cases where more protein was used (S7B Fig). This was not surprising, given the low solubility of XopL at low salt and room temperature. However, even in these experiments, XopL was clearly enriched in the pellet fraction when MTs were present. Given that the MT co-sedimentation assay suggests that XopL binds directly MTs, the implications of a PRR that contributes to binding warranted further investigation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Prolines within the PRR are essential to MT-association of XopL_Xe.

(A-B) Western blot analysis of MT co-sedimentation assay utilizing (A) 5 μg of affinity-purified XopL_Xe (88.1 kDa) protein tagged with 6xHis-SUMO-strep or (B) the tag-alone control. Supernatant fractions are marked ‘S’ and pellet fractions ‘P’. Samples with MTs are marked ‘+’ and without by ‘-’. Recombinant protein was detected using a strep-specific antibody (*). The intensity of the XopL signal (abundance) is shown below corresponding lane. Ponceau stained membranes show tubulin at 55 kDa in the pellet fraction with MTs (<). The left side of the blots show protein mass in kDa. Experiments with XopL were repeated at least three times and the tag-only control experiment was repeated two times. (C) Quantification of MT-association of XopL_m derivatives from confocal images of agroinfected GFP-TUA6 lower leaf epidermis (2 dpi, OD₆₀₀ of 0.8). MT binding is expressed as ‘Relative MT fluorescence’ (see Materials and Methods). Treatments that are not significantly different (p>0.05) are labeled with the same letter (Kruskal-Wallis One Way Analysis of Variance on Ranks, pairwise Dunn’s post-hoc). Background colors are included for ease of comparison between key controls XopL_m (blue) and mCherry (red). (D) Quantification of confocal images to evaluate MT density following the expression of XopL_Xe derivatives in agroinfected GFP-TUA6 lower leaf epidermis (2 dpi, OD₆₀₀ of 0.8). Treatments that were significantly different than XopL_Xe are marked with black asterisks (* = p<0.05, ** = p<0.01, *** = p>0.001; Kruskal-Wallis One Way ANOVA on Ranks, Dunn’s post-hoc against XopL). ex2 proline to alanine exchanges that were significantly different than ex2 are marked with red asterisks (*** = p>0.001; Kruskal-Wallis One Way ANOVA on Ranks, Dunn’s post-hoc against ex2). Red box in the background is included for ease of comparison with the mCherry control (red). (C and D) Boxes represents first to third quartiles and the median is represented by a horizontal line. Whiskers represent the distribution of remaining data points. (E and F) Western blot analysis of protein extracts isolated from the experiments in (C) and (D), respectively. Signals of XopL derivatives and the mCherry control were detected using an mCherry-specific antibody (*). ‘NI’ and ‘Ch’ refer to not inoculated and mCherry control, respectively. Protein mass is expressed in kDa. (F) Amido black staining is shown as a loading control.

https://doi.org/10.1371/journal.ppat.1011263.g006

Quantification of MT association of XopL_Xe variants confirms the importance of prolines to MT association

To allow for the quantification of MT localization, XopL E3 ligase mutant derivatives described above were fused to mCherry and expressed in GFP-TUA6 plants. A method for quantification was devised using the open-source image analysis platform, Fiji [25, 26]. ‘Relative MT fluorescence’ was quantified by first measuring mCherry signals at MT and dividing by measurements taken in neighboring cytosol (for details, see Materials and Methods). It is worth noting that in cases where the background is high, such as αLRR-XL_m, weak MT labeling was observed in some cells but this was not reflected in measurements. Despite this, however, quantification largely reiterates what is seen in Fig 5. Individual XopL_Xe domains and αLRR-XL_m do not co-localize with TUA6-labeled MTs and are comparable to labeling by an mCherry control, whereas XopL_m and NTαLRR fluorescence signals were enriched along MTs when compared to the background (Fig 6C). Ex1_m binding was variable, but significantly enhanced compared to the mCherry control, and ex2_m enhanced binding further.

This approach was also used to evaluate the importance of the 7 conserved prolines in the PRR. Four ex2_m derivatives were generated with proline to alanine exchanges: P127A + P130A (ex2_m127/30), P133-35A (ex2_m133-35), P140A + P149A (ex2_m140/49) and a variant with all 7 exchanged (ex2_m7x). Expression of mCherry-tagged variants in GFP-TUA6 plants revealed that ex2_m127/30 and ex2_7x had the greatest decrease in MT binding, followed by ex2_m133-35, whereas ex2_m140/49 did not have a significant impact (Fig 6C). Protein sizes and stability were confirmed via Western blot (Fig 6E). In conclusion, the quantification and analysis of the MT association of proline exchange mutants further confirms that the PRR identified through sequence comparison of XopL_Xe and XopL_Xcc contributes to MT binding of XopL_Xe.

XopL_Xe-induced decrease in MT abundance is correlated with MT affinity

The above-mentioned test for MT-association of PRR truncations and alanine exchanges was conducted using E3 ligase-inactive variants, to avoid MT disassembly. To determine if the level of MT-association of these variants was sufficient to destroy MTs, new mCherry fusions were generated with active E3 ligase domains. These new fusions were expressed in GFP-TUA6 plants to monitor MT abundance. Quantification of MT number revealed that indeed the addition of the proline rich region (127–136 aa) onto the αLRR_XL was sufficient, not only to bind MTs, but to decrease MT number very close to the level of the WT XopL_Xe (Fig 6D). This experiment also shows that exchange mutants ex2_7x and ex2_127/30 resulted in significant MT rescue compared to XopL_Xe, whereas ex2 and ex2_133-35 and ex2_140-49 were not significantly different from wild-type XopL_Xe.

When relative MT fluorescence exhibited by XopL_m derivatives was plotted against the MT number following expression of the E3 ligase-active versions it revealed a negative correlation, i.e., the more binding a derivative exhibited, the less MTs remained (S8 Fig). This data not only suggests that the strength of XopL MT binding is correlated to the loss of MTs, but also emphasizes the contributions of prolines 127/30 to this activity.

NTα of XopL_Xe converts XopL_Xcc into a MT binding protein

The unstructured NT of XopL_Xe alone was not sufficient for MT localization in the experiments described in Fig 5, but this construct lacked part of the PRR region. Therefore, to test if the NT+PRR is sufficient for MT binding, the NT was extended C-terminally to include amino acids 1–149 (NT₁₄₉) and tagged with mCherry. However, this construct is insufficient for MT localization (Fig 7C), suggesting that structural components are missing that facilitate an interaction with MTs. This fits with data on PRRs in well characterized MT binding proteins from humans, such as Tau [27] and MAP2 [28], where PRRs are required to stabilize protein-MT associations but are not sufficient for binding alone. Interactions with MTs often require stretches of surface exposed basic amino acids to facilitate electrostatic interactions with acidic residues in tubulin [29]. Therefore, we re-examined the amino acid sequence of XopL_Xe downstream of position 149; the α region (made up of 3 helices) is enriched in basic amino acid residues when compared to XopL_Xcc, a feature that is notably conserved among MT-associated XopLs (S6 Fig). Electrostatic surface potential models of XopL_Xe and XopL_Xcc proteins predict visible structural differences in the surface of the α region (Fig 7G). Residues from XopL_Xe α2 and α3 are predicted to form a surface exposed, positively charged region that is absent in XopL_Xcc. We tested the importance of the α region for interactions with MTs by sequentially adding α-helices 1–3 to the 1–149 aa derivative of XopL to create constructs 1–164 aa (NTα1)-, 1–179 aa (NTα1–2)- and 1–210 aa (NTα)-mCherry fusions. Analysis of these constructs in GFP-TUA6 plants revealed very weak binding of NTα1 (Fig 7D), whereas extensions towards the C-terminus as in NTα1–2 and NTα incrementally increased in their affinity for MTs (Fig 7E and 7F). Quantification of MT-association of the NTα variants further confirms this finding, with only the NTα construct demonstrating MT-association approaching that of XopL_m (Fig 7K, p>0.05). These results suggest significant contributions of the α-helices of XopL_Xe in establishing a stable interaction with MTs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. The exchange of the NTα region of XopL_Xe and XopL_Xcc is sufficient to engineer a MT-binding XopL_Xcc.

(A-F, H-J) Confocal microscopy images of lower epidermal cells of GFP-TUA6 (labels MTs) transgenic N. benthamiana leaves. Leaves were agroinfected (OD₆₀₀ of 0.8) to express (A) mCherry and (B) XopL_m controls and (C-F) XopL derivatives: (C) XopL_Xe 1–149 aa (NT₁₄₉), (D) XopL_Xe 1–164 aa (NTα1), (E) XopL_Xe 1–179 aa (NTα1–2), (F) XopL_Xe 1–210 aa (NTα; α1–3). Scale bars are 15 μm. The mCherry channel is visible in magenta and GFP (MTs) is white. White arrows point out examples of MT labeling. The corresponding western blot is shown in S9 Fig. (G) Electrostatic surface potential comparison of the α-helical regions of XopL_Xcc and XopL_Xe modelled with AlphaFold2/PyMOL. Values are expressed on a color scale from -5 (red) to +5 (blue) kT/e. The right panel depicts an overlay of the electrostatic map of XopL_Xe with its ribbon structure to show the locations of α 1–3. Basic residues are represented in stick format. N-terminal ends of proteins are marked ‘N’ in all panels (H) Quantification of MT-association of XopL derivatives from confocal images of agroinfected GFP-TUA6 lower leaf epidermis (2 dpi, OD₆₀₀ of 0.8) depicted in (A-F). MT binding is expressed as ‘Relative MT fluorescence’ (see Materials and Methods). Treatments that are not significantly different (p>0.05) are labeled with the same letter (Kruskal-Wallis One Way Analysis of Variance on Ranks, pairwise Tukey post-hoc). Background colors are included for ease of comparison between key controls XopL_m (blue) and mCherry (red). Boxes represents first to third quartiles and the median is represented by a horizontal line. Whiskers represent the distribution of remaining data points. (I-J) Confocal microscopy image of XopL_Xe 1–210 aa translationally fused with XopL_Xcc 78–428 aa (NTα_XeLRR_Xcc) from the same experiment as depicted in (A-F). (I) GFP labeled MTs are in white, (J) the domain swapped XopL in magenta. White arrows point out examples of MT labeling. Scale bars are 15 μm. The corresponding western blot is shown in S9 Fig.

https://doi.org/10.1371/journal.ppat.1011263.g007

The finding that the NTα was sufficient for MT localization of XopL_Xe guided a domain swap experiment between XopL_Xe and XopL_Xcc to create a version of XopL_Xcc that can bind MTs. Hence, the NTα of XopL_Xe (1-210aa) was added to the LRR of XopL_Xcc (78–428) and fused with a C-terminal mCherry to produce the E3 ligase-inactive construct NTα_XeLRR_Xcc. NTα_XeLRR_Xcc clearly localized to MTs (Fig 7H–7J), re-iterating that the NTα and not the LRR or XL-box are needed for MT association.

XopL MT binding is correlated with cell death

As mentioned, one readout of XopL_Xe activity is that it induces cell death when expressed in N. benthamiana by agroinfection, a phenotype that was previously shown to rely on both the LRR and the XL-box together, likely due to the ubiquitination of an unknown plant protein [12]. The question arises, is the cell death phenotype linked to the decrease in MTs? To test this, we used the XopL_Xe E3 ligase-active truncations and amino acid exchange mutants to test for cell death induction. Cell death was monitored via red fluorescence scanning as described in detail by Xi et al. [30]. For these experiments, all derivatives were fused with a C-terminal GFP tag so that their expression would not interfere with detection of cell death in the mCherry channel. As previously published [12], XopL_Xe induced the strongest cell death, indicated by high gray values, whereas the GFP control showed almost no fluorescence (Fig 8A). Visual inspection of the leaves confirmed this (Fig 8C). In addition, constructs that strongly localized to MTs such as ex2, ex2_140/149 and to a lesser extent ex2_133-35, were most efficient at inducing cell death (Fig 8A). All derivatives were expressed at a similar level and were detectable at the expected size (Fig 8D). We therefore concluded that there was a strong link between MT-association of XopL_Xe and cell death induction.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Cell death is correlated with MT association of XopL.

(A-B) Quantification of plant cell death via red fluorescence scanning of agroinfected leaves expressing XopL derivatives tagged with GFP 5 dpi (OD₆₀₀ of 0.8). (A) Cell death in leaves expressing MT-binding and non-binding derivatives of XopL_Xe or Arabidopsis proteins KTN1 and PHS1. Treatments that were significantly different than XopL_Xe are marked with asterisks (* = p<0.05, ** = p<0.01, *** = p>0.001; Kruskal-Wallis One Way ANOVA on Ranks, Dunn’s post-hoc test against XopL_Xe). (B) Cell death following expression of XopL_Xe, XopL_Xcc, and NTα_XeLRR_XccXL_Xcc; termed DS. Treatments that are not significantly different are marked with the same letter (p>0.05; One way analysis of variance, pairwise Bonferroni t-tests post-hoc). (A-B) Boxes represent first to third quartiles, while the median is marked by a horizontal line and whiskers show the distribution of remaining data points. (C-D) Examples of visible light and red-light photos of agroinfection spots recording cell death from the experiments in (A) and (B) respectively. (E) and (F) Western blot analysis of protein extracts isolated from tissue analyzed in experiments in (A) and (B), respectively. Signals were detected using a GFP-specific antibody (*). The left side of blots shows protein mass in kDa. Amido black staining was included as a loading control.

https://doi.org/10.1371/journal.ppat.1011263.g008

To explore the link between cell death and MT binding the next experiments aimed to quantitatively confirm data shown in Fig 1A which suggested that the non-binding XopL_Xcc does not cause cell death. Furthermore, a domain swap was used to generate an E3 ligase-active MT-associated XopL_Xcc variant (NTα_XeLRR_XccXL_Xcc; termed DS for short). All proteins were fused to a C-terminal GFP to allow cell death measurements via red fluorescence scanning. Confirming results from Fig 1A, XopL_Xcc showed little signs of cell death at 5 dpi when compared to GFP controls. In contrast, expression of the DS construct resulted in strong cell death reactions that were comparable to XopL_Xe and significantly higher than with XopL_Xcc (Fig 8B and 8C). Protein abundance of derivatives was similar (Fig 8E). Cell death at early time points (between 1–2 dpi) interfered with the ability to assess whether MTs were disassembled in these cells. These results strongly support the idea that XopL binding to MTs is linked to cell death, but also the idea that changing the subcellular localization of XopLs may be sufficient to alter their activity in planta.

MT destruction is not sufficient to induce cell death

Due to the importance of MTs for cell structure and general function, it is possible that the cell death caused by XopL is simply due to the lack of MTs in XopL-expressing cells. To test this hypothesis, MTs were disrupted via the overexpression of known MT disrupting proteins from A. thaliana: the MT severing protein KATANIN 1 (KTN) [31] and truncated versions of the tyrosine phosphatase PHS1, a protein involved in cytoskeleton rearrangement during stress [32]. Two phosphatase-dead variants of PHS1 were used: aa 1–700 and aa 85–700 (aa 85–700 lacks the kinase interaction motif), both of which were previously shown to depolymerize MTs [32]. MT disruption by these proteins was confirmed via over-expression in GFP-TUA6 plants (S10A–S10H Fig). This approach was chosen in place of chemical treatments, such as oryzalin, which sequesters tubulin to prevent MT formation [33]. Oryzalin fails to eliminate all MTs in our experimental setup [18] and is not effective over the time-frame desired (1–2 dpi). Surprisingly, even after 5 dpi, no significant cell death was observed when either PHS or KTN1 was expressed (Fig 8A and 8C). MT destabilization by XopL, therefore, is unlikely the cause of cell death. To determine whether the cell death was simply an additive effect of XopL E3 ligase activity and MT destruction occurring in the same cells, the αLRR_XL variant, which does not bind MTs, was co-expressed with the KTN1 and PHS_185-700 constructs. Quantification of cell death showed that the expression of the non-binding αLRR_XL and ex2_127/30 in cells where MTs were destabilized by PHS1 or KTN1 was insufficient to induce cell death comparable with XopL_Xe (S10 Fig). Taken together, the data suggest that MT-localized XopL_Xe activity is important for this phenotype.

Conservation of XopL PRR across Xanthomonas clades

In vitro MT co-sedimentation assays and in planta analyses of XopL truncations/amino acid substitutions suggest that the interaction of XopL with MTs is direct and not mediated by MT-associated proteins. XopL_Xe association with MTs requires the unstructured N-terminus, including a PRR, in concert with an α-helical region. Sequence comparisons of XopL proteins from 24 different species scattered throughout the Xanthomonas genus revealed that, while the NT varies in length, the PRR is a conserved feature among XopLs (S11 Fig). In this context it is also notable that the PRR co-occurs with a high density of basic amino acids (e.g., 19.6% in XopL_Xe vs. 9.9% in XopL_Xcc) within the α region. The conservation of the PRR and α-helical region in combination suggests that MT-localization is likely common to XopL proteins throughout the Xanthomonas genus. Interestingly, a strain of X. hyacinthi (CFBP1156) which shares a common ancestor with Xe, Xoo, Xac and Xcc [34, 35] harbors a XopL with 3 of 7 conserved PRR prolines (aligning with P133, 134, and 149 of XopL_Xe) and has relatively more basic amino acids in the PRR/α than XopL_Xcc (15.8% vs. 11.9%). While at this point it is not known whether XopL from X. hyacinthi associates with MTs, it appears that an ancestral form of the MT binding region was present in the common ancestor of Xcc and Xe, a characteristic that was likely lost as Xcc diverged.

So far, differences in XopL MT-binding ability are not obviously correlated to specific modes of pathogen entry or to the plant host species, as both Xac and Xe prefer to enter through stomata and proliferate locally while Xoo and Xcc enter primarily through hydathodes (although hydathode architecture of plant species differs) and move systemically through the plant via the xylem [2]. Xanthomonads typically have limited host ranges, and interactions with their hosts shape effector repertoires and activities, applying strong selective pressure for the evolution of T3Es that subvert effector-triggered immune responses [2].

XopL MT-binding mimics known eukaryotic MAPs

The presence of a PRR close to a region rich in basic amino acids, as in XopL proteins from xanthomonads, is reminiscent to the structure of known eukaryotic MT binding proteins (MAPs). PRRs in human MAPs, such as tau [27] and MAP2 [28] are required to stabilize protein-MT associations but are not sufficient for binding. In the case of tau, the PRR must act in concert with neighboring repeat regions harboring basic amino acids, with both protein regions interacting to facilitate MT binding [23, 36, 37]. It appears that XopL likely uses a similar mode for its association with MTs. This hypothesis is based on our finding that XopL derivatives lacking the PRR (αLRR_XL_m) are not MT-localized, nor are derivatives lacking the positively charged α region (NT), but together these regions are sufficient to bind MTs. It is worth noting, that while bacteria also possess tubulin homologs that assemble into filaments, bacterial tubulins lack acidic residues at the C-terminus that facilitate electrostatic interactions with kinesin motor proteins and MAPs [38]. Consistent with this, bacteria do not typically possess MAP homologs [39]. As far as we are aware, this is the first description of a MAP-like mechanism for MT binding described for any T3E. We suppose that XopL mimics features of both eukaryotic E3 ligases and MAPs.

A model for XopL interaction with microtubule-associated substrates

Agroinfection, the transient protein expression system chosen for this study, provided the ideal platform for comparative analyses of XopL homologs from different Xanthomonas species, facilitating both the visualization of XopL localization in plant cells coupled with a macroscopic readout for one of XopL’s activities, plant cell death. While it must be considered that during an infection with Xanthomonas XopL protein concentrations are much lower, it appears unlikely that a decrease in protein level would alter the ability of XopL to associate with MTs, which occurs both in vitro (with isolated MTs) and in planta. Taken together, the results of this study support a model in which XopL associates with MTs through the NTα region. Thus, the LRR, the predicted protein interaction domain, remains free for interactions with MT-associated substrates.

Interactions of XopL_Xe at the microtubule induce plant cell death

In the case of agroinfection-mediated expression of XopL_Xe in plant tissue, current data suggest that the modification of MT-localized targets, rather than MT disruption, triggers plant cell death. This is supported by the finding that no cell death occurs when MTs in plant cells are disrupted by other proteins, as shown here. A. thaliana KTN1 and PHS1, while destroying the MT network by 2 dpi, do not trigger macroscopic cell death, in contrast to XopL_Xe. Furthermore, XopL_Xe derivatives that are unable to bind MTs also fail to cause cell death, and co-expression of XopL mutants with KTN1 or PHS1 did not restore cell death. This suggests that there is an alternative, MT-associated, E3 ligase-dependent mechanism of plant cell death induction. Whether the cell death evaluated here is related to XopL’s virulence function or is the result of plant recognition and the initiation of programmed cell death triggered by XopL activity at MTs is enigmatic at this point.

Are MTs essential for autophagic flux?

Autophagy is a conserved eukaryotic degradation pathway which manages the content of the cytoplasm, ridding the cell of proteins, nucleic acids, faulty organelles and lipids, either through selective mechanisms or in bulk [40]. Cellular content is enclosed within a double membrane to form a vesicle (autophagosome), which is transported to the vacuole (in plants) or to lysosomes (animals) where its contents are degraded and recycled [40]. Autophagy plays a complex role in plant immunity and is a common target of T3Es from pathogenic bacteria, which either enhance or suppress this process [41]. XopL_Xe was recently reported to co-localize in planta to punctate structures with autophagy components AUTOPHAGY-RELATED GENE 8 (ATG8) and Joka2 and to inhibit autophagic flux [13]. MT localization of XopL is particularly interesting in this context as there are several lines of evidence to suggest that autophagy and MT dynamics are closely linked. In fact, Arabidopsis ATG8 isoforms are similar in sequence to known mammalian MT-associated proteins (MAPs) and are capable of direct MT binding in vitro and in planta [42, 43]. Tobacco ATG6, a component important for autophagosome assembly, also localizes to MTs and interacts with both β- and α-tubulin [44]. Most interesting is the finding that chemical or silencing approaches that decrease tubulin abundance or alter dynamics of MTs, inhibit stress-induced autophagy in both plant and animal cells [44, 45]. Taken together, this has led authors to speculate about the role of MTs in every aspect of autophagy from autophagosome biogenesis to transport or docking with the vacuole [42, 44, 45]. This brings about the question of whether the reported manipulation of autophagy by XopL_Xe [13] depends on its MT localization. Furthermore, one must consider whether the strong inhibitory effect of XopL_Xe on autophagic flux observed after agroinfection of N. benthamiana [13] is the result of MT destabilization.

Implications of MT manipulation by bacterial pathogens

There are many examples of animal bacterial pathogens that target MTs and MAPs. Human pathogens Shigella flexneri and Escherichia coli employ T3Es VirA and EspG to depolymerize microtubules at invasion sites to ease bacterial entry into host cells [46–49]. In contrast, Salmonella species deliver multiple T3Es to modulate MT-mediated vesicle cycling [49]. In addition to T3Es, animal pathogenic bacteria secrete many factors and toxins that target microtubules to facilitate entry, movement and colonization of cells [49]. Xanthomonas, by contrast, does not enter the cells of host plants, but colonizes the spaces between cells, suggesting a different function for MT modulation. Furthermore, there are fundamental differences in the MT function and regulation between plant and animal cells [50, 51]. Therefore, specific efforts must be made to understand how phytopathogens exploit MT-related cellular processes to their benefit.

Only a handful of T3Es from phytopathogenic bacteria have been identified that manipulate plant MTs or MAPs. One example is the Pseudomonas syringae effector HopZ1a which directly binds MTs [52] and alters MT dynamics in Arabidopsis thaliana, acetylating both tubulin [52] and the MT-stabilizing kinesin HINKEL [53]. Other T3Es target MAPs, such as MAP65-1 which is targeted by the P. syringae T3E HopE1 [54] and ACIP, targeted by the Xe effector AvrBsT [55]. In case of HopE1 and HopZ1a the impact on MTs and MAPs was linked with a decrease in the secretion of antimicrobial proteins and callose deposition [52, 54], but so far the exact role of MTs in these processes is unclear. Evidence is emerging that suggests that MTs play a role in signal transduction during plant immune responses [56], and as already mentioned, MTs play an unspecified role(s) in autophagy. Although the targeting of MT-related processes by T3Es points to their importance in determining the outcome of plant-pathogen interactions, there is still much ambiguity surrounding the exact role of MTs and MAPs in plant defense. In this regard the future study of effectors, including XopL_Xe, and their targets may serve to gain further insight into MT-dependent defense responses and answer questions surrounding fundamental plant processes.

Plant material and growth conditions

Wild-type Nicotiana benthamiana and stably expressing GFP labeled α-tubulin (GFP-AtTUA6; [21]) plants were grown in a greenhouse supplemented with light emitted by IP65 lamps (Philips) equipped with Agro 400 W bulbs (SON-T). Light intensity was approximately 130–150 μE/m² x s with light supplementation provided when light was below 100 μE/m² x s. Day length was 16 h at 26°C with a relative humidity of 60%, and the temperature at night was 19°C with a relative humidity of 40%.

Agrobacterium strains and inoculations

A. tumefaciens strain GV3101 (pMP90) was used for all expression assays in this study. Bacteria were grown in liquid yeast extract broth (YEB; Vervliet et al., 1975) supplemented with 100 ug/mL rifampicin, 15 ug/mL gentamicin, and appropriate antibiotics for plasmid selection (pGGA plasmids = 100 μg/mL spectinomycin, pICH47751 = 100 μg/mL carbenicillin), and cultivated in a tube rotator at 30°C overnight. Cells were harvested via centrifugation, re-suspended in A. tumefaciens infiltration medium (AIM = 10 mM MgCl2, 5 mM MES, pH 5.3 and 150 μM acetosyringone) and incubated at room temperature for 1 h. Bacteria were adjusted to an OD₆₀₀ = 0.4 of 0.8 prior to the inoculation of N. benthamiana lower leaves using a needleless syringe.

Taxol treatments

Treatments with Paclitaxel (taxol) (Cytoskeleton inc. Cat. # TXD01) were administered at 4 hpi with A. tumefaciens strains facilitating the expression of XopL_Xe, XopL_m or the mCherry control protein fusions. Paxlitaxel was dissolved in DMSO to a concentration of 1 mM then diluted in AIM without acetosyringone to a concentration of 20 μmol and administered to A. tumefaciens inoculation spots via a needleless syringe. DMSO controls were diluted in the same way. Microscopy was performed at 2 dpi unless otherwise stated.

Cloning

Unless otherwise stated XopL_Xe constructs were amplified from plasmids generated by Singer et al., [12]. XopL_Xcc was PCR-amplified directly from Xcc 8004 DNA. Derivatives of A. thaliana PHS1 (AT5G23720) and KTN1 (AT1G80350) were cloned from A. thaliana cDNA. KTN1 was amplified with primers harboring attB sites to facilitate Gateway cloning (Thermo Fisher Scientific). BP clonase II enzyme mix was used for a BP reaction into pDONR221. The pDONR221-KTN1 plasmid was then used as a template for further PCR reactions. PCR reactions on the templates described, used primers with BsaI or BpiI recognition sites to generate appropriate fragments for GoldenGate cloning [57].

For cloning of XopLs for UbiGate experiments and selected in planta assays XopLs from Xoo PX099a, Xac 306, Xcc 8004 and Xe 85–10 strains were synthesized as gene fragments with appropriate BpiI sites (GeneArt gene synthesis by Thermo Scientific; codon-optimized for N. benthamiana) and subcloned as level 0 CDS1ns modules (in pAGM1287) for use in the MoClo cloning system [58]. CDS1ns modules were used for the assembly of UbiGate plasmids or plant expression vectors. For in planta expression of XopL, CDS1ns modules were assembled in combination with the ’35S long’ promoter (pICH51266), a C-terminal mCherry tag (pICSL50004), and a 35S terminator (pICH41414) module into the plant expression vector pICH47751 (Amp^R) [58]. The mCherry control was assembled using the same 35S promoter and terminator modules but with the mCherry module pICSL80007 [58]. For UbiGate, codon-optimized XopL CDS1ns modules were assembled into pET-28GG (Kan^R) under control of the T7 promoter [20]. Final pET-28GG plasmids were assembled from modules published by [20]: pGG-193 (His-tagged A. thaliana UBIQUITIN 10) and pGG-194 (untagged A. thaliana UBIQUITIN-ACTIVATING ENZYME 1), in combination with modules generated in this study; UG36 (GST-tagged UBIQUITIN CONJUGATING ENYZME 28; assembled from pUG-UBC28 [20] and pAGM18012 [20]), and XopL modules (XopLs with C-terminal 4x Myc tags; assembled from XopL CDS1ns and pICSL50010 [58]).

For cloning of non-codon-optimized XopLs, derivatives and A. thaliana genes for expression in planta, a set of GoldenGate-compatible vectors of the pGG system (Schulze et al., 2012) were used. As described in Schulze et al. [59] all pGG vectors are preassembled, using the pBGWFS7 backbone with spectinomycin resistance for selection in Agrobacterium, with 35S promoter and terminator plus N- or C-terminal fluorescence tags. All proline exchange mutations were introduced into the XopL wild type or E3 ligase mutant sequences via PCR-based single/multiple-site mutagenesis [60]. Unless otherwise stated, BsaI/T4-ligase cut-ligations facilitated cloning into all vectors, including an extra ligation step to repair any internal BsaI sites [57]. For overexpression of SUMO-tagged recombinant proteins in E. coli restriction-free cloning [61] was used to introduce strep-XopL, and strep alone into the pETSUMO plasmid (Thermo Fisher). A 352 bp fragment (5453–5804 bp) directly following the SUMO tag up to the BamHI site was replaced with strep-XopL or strep alone (control construct) to create 6xHIS-SUMO-strep-XopL and 6xHIS-SUMO-strep constructs. RF-Cloning.org [62] was used to generate primers. Primer sequences are available upon request.

Confocal microscopy

Confocal microscopy was done using a Zeiss LSM 780, inverted AxioObserver, with a 63x water lens (C-apachromatx63/1.20 W Korr M27) or a Leica STELLARIS 8 inverted fluorescent microscope, with a 40x water lens (HC PL APO40x/1.10). In case of the Zeiss LSM780 laser lines of 488 nm and 561 nm were used to excite GFP and mCherry fluorophores, respectively, and channels were defined from 493–523 nm (GFP), 588–632 nm (mCherry) and 684–721 nm (chlorophyll). ZenBlack software controlled the microscope and image capture. In case of the Leica, fluorophores were excited with a white light laser, and channels were defined the same as described for Zeiss microscopy. LasX software controlled the microscope and image capture.

Image processing and analysis with Fiji

Confocal images were collected as z-stacks (.czi or.lif files), and images were processed using the ’Z-project’ function of Fiji [25, 26] to create an extended depth of focus. Fiji [25, 26] was also used to quantify MT association of fusion proteins and MT destabilization. For these experiments, mCherry labeled fusion proteins were expressed in GFP-TUA6 plants. To quantify MT labeling, MTs were located and marked with the line tool in the GFP channel; mean gray values were measured in the mCherry channel, thus quantifying the brightness of the mCherry signal in the location of the filament. To quantify the cytosolic background, a neighboring region parallel to the MT was also measured. %MT fluorescence was calculated by dividing MT brightness by the cytosolic background. %MT measurements were collected at random along 10 MTs in 15 cells per treatment. To quantify the number of MTs remaining in cells expressing fusion proteins, lines of 30 μm were drawn at random in cell lobes (avoiding the nucleus and cell boundaries) and MTs crossing this line were quantified using the ‘Cell Counter’ tool. One line was drawn per cell and at least 10 cells were measured per treatment. SigmaPlot was used for the statistical analysis of all data.

E. coli growth and overexpression

Chemically competent E.coli BL21 DE3 RIL cells were freshly transformed with plasmid DNA and cells were cultivated in 50 mL liquid lysogeny broth (LB; [63]) medium containing 50 μg/mL kanamycin overnight at 37°C (shaking at 140 rpm). The following day a main culture of 20 mL was started for UbiGate assays or 500 mL for SUMO tagged protein overexpression and purification. Cultures were inoculated at a starting OD_600nm of 0.05. At an OD₆₀₀ of 0.6–0.8, at which point protein expression was induced with isopropyl β-D-1-thiogalactopyranoside (1 mM final concentration) and 99% EtOH (3%). For UbiGate, cultures were incubated in a shaker (140 rpm) at 28°C for 3 hours. For SUMO-tagged XopL and control overexpression cells were incubated in a shaker (140 rpm) overnight at 16°C. Bacterial cells were harvested using a tabletop Eppendorf centrifuge (5810R) at 4000 xg with a swing-out rotor (A-4-62), and cells were frozen at -80°C until samples were processed.

UbiGate assays

The UbiGate protocol was modified from Kowarschik et al. [20]. E. coli cells were resuspended in 400 μL 1x PBS buffer and adjusted to an OD_600nm of 13 in a volume of 100 μL to normalize for differences. Cell suspensions were treated with 1 μL lysozyme (10 mg/mL stock) for 15 min and 4x freeze-thaw (samples were transferred between liquid nitrogen and a 30°C heating block). Lysates were then treated with 2 μL RNase and DNase (stock 1 mg/mL) and incubated for 30 minutes. 100 μL of 2x Laemmli buffer was added and samples were boiled at 95°C for 10 minutes. Samples were stored at -20°C until analysis.

Purification of proteins and MT spin-down assays

E. coli cells expressing 6xHIS-strep-SUMO-XopL and the tag only control (6xHIS-strep-SUMO) were resuspended in 25 mL of PBS + PMSF (1 mM), and subjected to a French press (1000 psi, 3 cycles). All following procedures were conducted at 4°C. Cell lysates were cleared via centrifugation (30 min, 4000 rpm in a Beckman Coulter Avanti J-25 centrifuge with a F15-6x100y rotor). Supernatants were subjected to ultracentrifugation (60 min, 28, 000 rpm in a Beckman Coulter Optima MAX Ultracentrifuge with a SW-32 Ti rotor) to sediment membranes, and supernatants were transferred to a 100 mL bottle and topped up to 100 mL with Ni-column binding buffer (20 mM HEPES (pH 7.5), 500 mM NaCl, 20 mM Imidazole) and transferred to a gravity flow Ni-column via peristaltic pump (50 μL/min). Econo-Pac Chromatography Columns (20 mL; BioRad) packed with Ni²⁺ nitrilotriacetic acid (NTA)-sepharose (GE Healthcare) were used. The column was washed with 5 matrix volumes of binding buffer, followed by an elution step (Elution buffer: 20 mM HEPES (pH 7.5), 500 mM NaCl, 500 mM imidazole); the purified protein was detected in the flow-through via Western Blot.

To prepare MT binding assays the buffer was exchanged to eliminate salt and imidazole, which could interfere with assays. For this, purified protein was loaded into cellulose membrane tubing for overnight dialysis (3.5 kDa cutoff) into four different buffers (volume was 1000x that in the membrane) for a solubility test (details in S7A Fig). Due to low XopL solubility in buffers without salt (required for the spin-down assay), the SUMO tag was not cleaved off. XopL was incubated in HEPES buffer containing 80 mM HEPES, 5 mM MgCl2 and 250 mM of aminocaproic acid. Immediately following dialysis, protein samples were subjected to ultracentrifugation to remove aggregates (55, 000 rpm, 20 min, 4°C). MT assays were performed using the MT Binding Protein Spin-down kit from Cytoskeleton, Inc. (Cat. # BK029), according to manufacturer instructions. Samples were stored in Laemmli buffer [64] at -20°C prior to analysis via SDS-PAGE and immunoblot or Coomassie staining. Protein amount was quantified using Fiji [25, 26].

Protein preparation and western blot

For the analysis of in planta protein level, agroinfected plant material from two independent plants was ground in liquid nitrogen, resuspended in 120 μL of 2x Laemmli, boiled for 10 min at 95°C, and stored at -20°C.

Proteins were analyzed via SDS-PAGE followed by staining with Coomassie Brilliant Blue [65], or immunoblot using the following monoclonal antibodies: α-Strep (1:5000, Sigma-Aldrich), α-P4D1 at (1:5000, Millipore), α-GFP (1:2000, Roche diagnostics) at 1:2000, α-c-Myc (1:2000, Roche diagnostics), DM1A (1:1000, Sigma-Aldrich), or the following polyclonal antibodies: α-mCherry (1:2000, Abcam) and α-GST (1:10, 000, Sigma-Aldrich). Secondary α-rabbit, α-mouse and α-goat antibodies (GE Healthcare) were coupled with horseradish peroxidase for visualization of proteins by enhanced chemiluminescence.

Protein sequence and structural alignments

Using the phylogenetic trees of the Xanthomonas species [34, 35] as guides, NCBI blastp with the αLRR_XL amino acid sequence of XopL_Xe were used to identify XopLs from representative strains throughout the genus. Multiple sequence alignments were conducted using Geneious Prime, Clustal Omega 1.2.2.

XopL_Xe and XopL_Xcc 3D structures were generated from full length XopL sequences utilizing AlphaFold2 [66, 67]. Protein structural alignments were conducted using Geneious Prime. Electrostatic surface calculations were performed using PyMOL’s APBS Electrostatics plugin [68] and results were visualized using PyMOL software (www.pymol.org).