The two Z. tritici strains “I05” (INRA09-FS0813, Mat1-1) and “I07” (INRA09-FS0732, Mat1-2), sampled in 2010 from STB lesions on wheat cv. Soissons in Grignon, France (48510 N, 1580 E), were crossed by co-inoculating adult plants with an equiproportional suspension of parental blastospores. After ascosporogenesis, 167 offspring individuals were collected from yeast-like colonies on Petri dishes placed upside down above wheat debris fragments to collect discharged ascospores as described in Suffert et al. (2016). The population of the 167 individuals resulting from the cross is hereafter referred to as “I05×I07”.

148 chosen randomly among the I05×I07 progeny strains, and the two parental strains were phenotyped over three replications on 16-day-old seedlings of the wheat cultivar ‘Renan’ grown in a growth chamber as previously described in Langlands-Perry et al. (2022).

The strains were precultured in YPD (Yeast extract Peptone Dextrose), composed of 1% yeast extract, 2% peptone, 2% glucose and 95% distilled water, 10 days prior to inoculation, and each preculture was then grown on a PDA solid culture medium (potato dextrose agar), as described in Langlands-Perry et al. (2022). 150 mL of blastospores were prepared in advance from these cultures, each with a concentration of 10⁶ ± 1.10⁵ spores.mL^-1, kept at -80°C between 1 to 3 months to be used for each of the three replicated inoculations. Before inoculation, a drop of Tween 20^® was added per 15 mL of inoculum to ensure adherence of the inoculum to the leaf.

The day before the inoculation, only three plants were kept per pot. On the first true leaf (about 3 cm from the base) of each plant, a surface of 7.5 cm in length was delineated with two black marks from a felt tip. Inoculation was carried out with cotton swabs, one per inoculum, in six passages (3 times back and forth), within the marks. After inoculation, the pot was covered with a plastic bag closed off at the top with a paper clip to not only create a water-saturated atmosphere, which encourages infection (Shaw, 1991; Boixel et al., 2022), but also to isolate the pots from one another to prevent cross contamination. The paper clips were removed after three days, a 72 h incubation period being the time it takes for the fungus to reach the mesophyll, which is necessary to the rest of the colonisation process (Kema et al., 1996). To optimise conditions for the survival of the inoculated leaf and to homogenise the quantity of light received by each leaf, new leaves were cut 2 to 3 cm above the first node 14 days post-inoculation (dpi).

The obtained data sets were analysed with the R software (R Core Team, 2019). For each trait an analysis of variance (ANOVA) was performed with the following model:

Where Y i j is the trait which is being studied, µ is the mean value for this trait,   I i is the individual genotype, r j the replication, I r i j the interaction and ϵ i j the residual. For the following analyses, I r i j was included in the residual.

The following hypotheses were tested after the variance analyses.

ϵ   ∼   N ( o , σ 2 )

c o v ( ϵ , ϵ ′ ) = 0  Homoscedasticity (homogeneity of v a r ( ϵ ) )

Broad-sense heritability is defined by the following formula:

Where H 2 is the heritability, σ g 2 the genotypic variance and σ e 2 the residual variance.

Two ANOVAs were carried out as 30 individuals were missing from the experiment for replicate 3. The first ANOVA included the data for all three replicates but with the 30 individuals missing from replicate 3 removed. The second ANOVA included the data for all individuals but only for the replicates 1 and 2.

The correlation between traits was studied using the Bravais-Pearson correlation.

The 167 progeny and two parental strains were grown over 7 days in a 250 mL Erlenmeyer flask containing 100 mL of YPD inoculated with 30 µL of inoculum. Growth chambers were fixed at 17°C, a hygrometry of 70%, constant agitation at 160 rpm and under neon lights (two Osram L 58W/840 Lumilux Cool White tubes). Then, spores were washed and transferred into 50 mL Falcon tubes to be lyophilized for 24 to 30 hours. The dry samples were ground in liquid nitrogen with a mortar and pestle. DNA was extracted following a phenol/chloroform-based protocol adapted from Fagundes et al. (2020). After drying, extracted DNA samples were suspended in 250 µL of Tris buffer at 10 mM. Sample purity was verified with a Nanodrop (Desjardins and Conklin, 2010) measurement while concentrations were measured with a Qubit 2.0 (Anon, 2011). All samples were subjected to an electrophoresis to verify that they were not degraded.

Samples were sequenced following the RADseq (Restriction site Associated DNA sequencing) strategy (Etter et al., 2011), on the Platform MGX (MGX-Montpellier GenomiX) on a HiSeq 2500 (Illumina) in paired-end 2*125nt mode. This type of sequencing enables one to target 1 to 10% of the genome via the use of a restriction enzyme and tagging of digested strands. Sequencing depth per sequenced locus is higher than classic sequencing, while the price is considerably lower. The restriction enzyme used was Pst1, following a previous study by Lendenmann et al. (2014), corresponding restriction sites were present throughout the IPO323 genome ( Figure S1 ).

A genetic map containing 18,316 SNP markers for 1,332 genetic bins covering the whole Z. tritici genome bar chromosomes 14 and 18 was built using the Multipoint ultra-dense software (MultiQTL Ltd, Haifa University, Israel) (details in Supporting information 2) and was used for QTL mapping.

A linkage analysis was carried out using the R/qtl software (Broman et al., 2003) version 1.42-8. This analysis included for each trait an initial Simple Interval Mapping (SIM), followed by a Composite Interval Mapping (CIM). Analyses were performed replication by replication. For SIM, 1000 genome-wide permutations were used to calculate the significant logarithm of odds (LOD) threshold. Were considered significant only QTL that showed P-values<0.05. The CIM was carried out with the QTL with the highest LOD used as a covariate. QTL intervals were evaluated with the LOD support interval with a drop in LOD of 1 and the “expandtomarkers” argument as true. QTL effects were calculated with the “effectplot” and “effectscan” functions. Possible epistatic interactions between QTL were looked into using the “addint” function.

A detailed version is presented in Supporting information 4.

In addition to the I05 and I07 de novo assemblies, the assemblies of 103 Z. tritici strains collected in France in 2009-2010 (BioProject: PRJNA777581) and 126 strains collected in France in 2018-2019 (BioProject: PRJNA881220) were also available to us, enabling us to look into the sequence diversity of the pathogenicity gene. We extracted gene sequences from genome assemblies of the 229 Z. tritici strains using the ncbi-blast+ software (Camacho et al., 2009). Prior to performing the population genetic analysis, we first verified population structure in our dataset, to ascertain that there is no inflation due to population structure. We constructed a protein sequence phylogenetic tree using the RaxML algorithm with the GAMMA JTT model and 100 bootstrap replicates (Stamatakis, 2014). Phylogenetic trees were visualized using iTOL (Letunic and Bork, 2016); https://itol.embl.de/ . We used the R package Popgenome (Pfeifer et al., 2014) to calculate sliding-window analyses of nucleotide diversity (π) of the effector gene, including ~500 bp upstream and downstream of the coding sequence. We used a window length of 20 bp and a step size of 5 bp. To verify whether the effector gene exhibits signatures of positive diversifying selection, we calculated the rates of ω, the ratio of nonsynonymous to synonymous mutational rates using the codon-based selection analysis codeML (Yang et al., 2004). The ratio indicates negative purifying selection (0< ω< 1), neutral evolution (ω = 1), or positive diversifying selection (ω > 1). We compared different evolutionary models and used the statistical likelihood ratio test (LRT) to determine the model that best fitted our data. The Bayes empirical Bayes method (BEB) was then used to evaluate the posterior probability of sites considered to have been positively selected.

A representation of the distribution of the different traits for all three replicates shows that they do not follow a normal distribution ( Figure S2 ), as confirmed by a Shapiro-Wilk normality test. For all traits, transgressive individuals are observed.

Bravais-Pearson correlation results show that all traits but NBS were correlated ( Figure 1 ). The most strongly correlated traits were S26 and AUDPCS with a correlation coefficient of 0.96, both traits linked to the sporulating area. AUDPCN and AUDPCG were also strongly negatively correlated with a correlation coefficient of -0.95. PYC was less strongly correlated overall than the other traits, it was however significantly correlated, with absolute correlation coefficient values ranging from 0.46 to 0.76 for PYC (NBS values excluded).

Both ANOVAs ( Table S3 ) showed that the genotype had a significant effect on phenotypes, though the significance was milder for PYC and NBS. They also showed that, overall, the replication had high statistical significance. This led to all subsequent analyses being carried out replication by replication. The traits that best performed in the statistical analyses were AUDPCG, AUDPCN and PYC with the ANOVA assumptions respected. For the other traits, the assumptions were not so well respected and heritability may not be optimally estimated. Heritability ranged from 0.45 to 0.59 for S26, AUDPCG, AUDPCN and AUDPCS, values that are reasonably high. The values were lower for S20, PYC and NBS, ranging from 0.16 to 0.36.

The linkage analyses enabled us to detect three repeatable QTL: Qzt-I05-1, Qzt-I05-6 and Qzt-I07-13, on core chromosomes 1, 6 and 13, respectively ( Table 1 , Figure 2 , Table S7 ). Qzt-I05-1 and Qzt-I05-6 were detected for all three replicates, while Qzt-I07-13 was detected only for replicates 2 and 3 ( Table 1 ). Qzt-I05-1 is in a sub-telomeric region of chromosome 1. It covers a 4.96 cM-long region on the genetic map spanning 138 kb based on the physical position of the flanking markers. The mean phenotypic variation explained by this QTL is 6.37% and the parental strain carrying the pathogenic allele for this QTL is I05.

Qzt-I05-6 covers a 13.48 cM-long region corresponding to a physical interval of 169 kb on chromosome 6. The mean phenotypic variation explained by this QTL is 24.91%, the highest among the three repeatable QTL detected, and the parental strain carrying the pathogenic allele for this QTL is I05. This QTL presents itself as being the most robust of the three.

Finally, Qzt-I07-13 is found at a sub-telomeric region of chromosome 13. It covers an 8.64 cM-long region corresponding to a physical interval of 186 kb. The mean phenotypic variation explained by this QTL is 6.93% and, contrary to the other two QTL, the parental strain carrying the pathogenic allele is I07.

Significant epistatic interactions were detected between Qzt-I05-1 and Qzt-I05-6 and between Qzt-I05-6 and Qzt-I07-13 ( Table 2 ). Qzt-I05-6 and Qzt-I07-13 had the highest and most significant epistatic interaction explaining on average 7.12% of phenotypic variation. This epistatic interaction explains why strains carrying the pathogenic allele for these two QTL led to the highest S26 values among the strains combining the pathogenic allele for two QTL ( Figure 3 ).

For further investigation, we chose to focus our efforts on Qzt-I05-6 which was the most robust QTL, with the strongest effect on phenotypic variation.

The interval defined by Qzt-I05-6 holds 36 genes annotated in Grandaubert et al. (2015). There is a 60 Kb-long TE-rich region in the middle of the QTL, with markers that came out as peak markers during the linkage analysis on either side. None of these 36 genes has any predicted functions; we therefore predicted functions with the InterPro database (http://www.ebi.ac.uk/interpro/search/sequence/) ( Table S8 ). We searched for genes that have characteristics of known effectors, as these are often involved in fungal pathogenicity. Out of the 36 annotated genes, two had a signal peptide. The first, named Zt09_6_00095 in the Grandaubert et al. (2015) annotation, comes out as being in the PTHR35523 family in the panther classification system (http://www.pantherdb.org/), which regroups cell wall proteins. As a component of the cell structure, this gene does not seem to be a good candidate. The second gene, Zt09_6_00123, is predicted to encode a FAD binding domain, linking it to an oxidoreductase process. This gene is a little large for an effector with a length of 1272 bp and a corresponding protein of 423 amino acids (aa), the general cut-off being set at 300 aa (Sperschneider et al., 2018). It has a predicted function and only one cysteine in its protein sequence. It does not seem to be a good candidate either.

RNAseq data from Haueisen et al. (2019) enabled us to notice some reads that mapped to a position that was not annotated, right next to the central TE region. As pathogenicity genes can be found in regions such as this (Plissonneau et al., 2018), we looked further into the corresponding position. The RNAseq data produced by Haueisen et al. (2019) was used to make a read coverage file which was used in the integrative genomics viewer (IGV) software (Robinson et al., 2011) as a means of annotating a previously unidentified gene. On chromosome 6 of IPO323, this gene is positioned at 470,027-470,324 bp. It has two exons with coding sequence (CDS); the first has a predicted signal peptide (https://services.healthtech.dtu.dk/service.php?SignalP) (Teufel et al., 2022), and there is no predicted function or family for the protein (http://www.ebi.ac.uk/interpro/search/sequence/). It has an intron which is 61 bp long and has canonical splice site combination GT-AG (Kupfer et al., 2004; Frey and Pucker, 2020). The CDS is 237 bp long and encodes a 78-aa protein with 11 cysteines in its sequence (14% of the protein sequence). We identified four mutations between parental strains I05 and I07. In a recent, yet unreleased, annotation of the Z. tritici IPO323 genome sequence, specifically improved to detect genes encoding SSP, this gene was identified and named G_07189 (unpublished data). A BLAST search against the NCBI database (https://blast.ncbi.nlm.nih.gov/Blast.cgi) identified in other Z. tritici strains a second, longer protein, with a potential alternative methionine 32 aa before the predicted one. This longer, alternative protein has a cleavage site between the 47^th and 48^th aa, that is unlikely for a signal peptide. Searches for a GPI anchor did not return positive results. Therefore, the shortest isoform of 78 aa was retained for G_07189. The BLAST search did not yield any significant results in other species, indicating that this gene is species specific.

The study of the expression of this gene by RT-qPCR over the course of the infection revealed upregulation at 12 dpi in both I05 and I07 strains ( Figure 4 ), further consolidating its status as a very good effector candidate.

Additionally, comparison of the I05 and I07 assembled contigs revealed that the newly annotated gene lies in a highly dynamic region, with several presence/absence polymorphisms of TE between both parental strains ( Figure 5 ).

All knock-out, complementation and ectopic integration mutant strains were successfully generated following an ATMT protocol (Bowler et al., 2010). Three randomly selected mutants per construction were able to successfully infect ‘Chinese Spring’ and induce sporulation covering on average 99% of the inoculated leaf area ( Figure S4 ).

In the I05 genetic background, deleting gene G_07189 has no effect on the phenotype, while replacing the I05 allele (G07189_I05) by the I07 allele (G07189_I07) induces a suppression of sporulation ( Figure 6 ). Indeed, the I05 wild type protein induces a highly susceptible reaction on wheat cultivar ‘Renan’ (>90% sporulating leaf area) while ‘Renan’ becomes completely resistant when the I07 allele is introduced ( Figure 6 ). In the I05 genetic background, the I07 allele of G_07189 behaves as a classic avirulence gene inducing complete resistance in the host.

In the I07 genetic background, deleting the gene G_07189 also has a strong and significant effect on sporulation as I07 wild type induces 7% sporulating leaf area on average while the knock-out mutants I07_ΔG07189 induce 61% sporulating leaf area on average ( Figure 6 ). In the I07 genetic background, the effect of the I07 allele of G_07189 remains quantitative.

As the cloning experiments validated the effect of G_07189, we analysed its genetic diversity in the 229 French Z. tritici strains and searched for diversifying selection signatures. This analysis showed that G_07189 is highly conserved among these strains, regardless of their being collected in 2009-2010 or in 2018-2019, with non-synonymous polymorphism for only 7 amino-acid residues out of 78 ( Figure 7 ). With no collection period-related population structure, the strains carry alleles encoding fifteen isoforms of the G_07189 protein with four isoforms representing 86% of all strains including the I05 and I07 isoforms ( Figure 7 ). All of the sequence diversity observed for the gene is coded by the second exon, with π_{CDS2 =} 0.02254 while π_CDS1 and π_intron are around 0.0001 ( Table S9 ). The ratio between non-synonymous and synonymous mutations (ω=dN/dS) in the panel of strains was calculated at 2.118 for a one-ratio model M0 ( Table S10 ), a high value indicating diversifying selection, further validated by the selection M2 model having the highest InL value (-487.686, p<0.001) among the three models tested ( Table S10 ). According to the codon-based maximum likelihood approach, three residues are under significant diversifying selection (p<0.01) at positions 54, 71 and 72 in the protein sequence ( Figure 7 ). We therefore have a highly conserved gene that is under diversifying selection.

In this study we identified regions in the Z. tritici genome, which carry genes that contribute to quantitative pathogenicity towards T. aestivum. The cultivar used was ‘Renan’, known to carry at least three resistance QTL with different levels of resistance towards I05 and I07 on chromosomes 1D, 5D and 7B (Langlands-Perry et al., 2022). Additionally, the resistance QTL on ‘Renan’ show strain specificity between I05 and I07 and none of these QTL alone is able to confer total resistance to either I05 or I07, indicating that resistance is polygenic and quantitative in the ‘Renan’/I05-I07 interaction (Langlands-Perry et al., 2022). Pathogenicity towards ‘Renan’ is polygenic, with three QTL identified in this study, each contributing partially to the observed phenotypic variation. For two of these QTL, the parent carrying the pathogenic allele is I05, the parental strain known to be the most pathogenic on ‘Renan’. For the third, the parent carrying the pathogenic allele is I07, the least pathogenic parental strain on ‘Renan’. We showed that these QTL have varying effects on the phenotype, with Qzt-I05-6 having the strongest effect, and are subject to epistatic interactions, notably in the case of the Qzt-I05-6 and Qzt-I07-13 interaction. None of the detected QTL colocalizes with an already known avirulence gene in Z. tritici. Indeed, AvrStb6 is found on chromosome 5 (Zhong et al., 2017), AvrStb9 is on chromosome 1 but not in the region defined by Qzt-I05-1 (Amezrou et al., 2022) and Avr3D1 is on chromosome 7 (Meile et al., 2018). Looking at progeny strains that did not have any recombination within the QTL intervals, we are able to separate the different strains according to the QTL allele combination that they carry to observe the impact of these allele combinations on the performance of the strains on ‘Renan’ ( Figure 3 ). We looked into the data for maximal sporulation S26 as this trait led to the detection of all three QTL. S26 was always higher for strains with the pathogenic allele for several QTL compared with strains carrying the pathogenic allele for only one QTL, though the difference was not always statistically significant, indicating incomplete additivity of the QTL effects. None of the non-pathogenic alleles of the QTL induced a completely resistant phenotype such as that conferred by a major gene like Stb6 in the presence of AvrStb6 (Zhong et al., 2017). The only strains that were almost completely non-pathogenic were those that carried the non-pathogenic allele for all three QTL, demonstrating the quantitative and polygenic nature of pathogenicity and mirroring the quantitative and polygenic nature of resistance in ‘Renan’ (Langlands-Perry et al., 2022).

An effector-like gene (i.e. coding for SSP) was identified as a candidate to explain Qzt-I05-6. This gene had not been previously annotated but was supported by RNA sequences obtained from infected leaves (Grandaubert et al., 2015; Haueisen et al., 2019). Moreover, in a recent reannotation of the Z. tritici genome, which aimed at optimizing the detection of SSP (unpublished data), this gene was annotated and named G_07189. Additionally, qPCR performed to evaluate relative expression of the gene over the course of infection showed differential expression patterns, with an expression peak at 12 dpi further validating its status as a good effector candidate.

We were able to demonstrate the role of G_07189 in the pathogenicity of I05 and I07 on the cultivar ‘Renan’. The knock-out of this gene from I05 (most pathogenic parental strain on ‘Renan’) which generated I05_ΔG07189 strains, did not have a significant effect on the phenotype. The inclusion of the I07 (least pathogenic parental strain on ‘Renan’) allele for this gene, generating strains I05_ΔG07189+G07189_I07 and I05+G07189_I07, led to significantly lower S26 regardless of whether or not the I05 allele was present in the genetic background. The same phenomena were observed in the characterization of Avr3D1 (Meile et al., 2018), where the Avr3D1 knock-out in the background of the most pathogenic parental strain of the pair in that study (3D7) was unaltered compared to the wild-type. The knock-out however led to increased disease symptoms in the background of the strain with lower pathogenicity used in that study (3D1) compared to the wild-type.

This, corroborated with our results, suggests that G_07189 encodes an avirulence factor. Additionally we showed that the region in which G_07189 lies is TE rich and presents with TE presence/absence polymorphisms ( Figure 5 ), similar to what has been previously observed with AvrStb6 and Avr3D1 (Zhong et al., 2017; Meile et al., 2018). Moreover, while I07, which carries the avirulent allele for G_07189, is able to produce symptoms on ‘Renan’ (i.e. S26 averaging at 10%), I05_ΔG07189+G07189_I07 strains, which all carry this same allele, do not produce sporulation on ‘Renan’ (i.e. S26 averaging at 0.8%). G_07189 therefore leads to a quantitative phenotype in the I07 background, but to a qualitative, avirulent, phenotype in the I05 background. Rather than acting like a classic avirulence gene in an R/Avr interaction, G_07189 has a spectrum of effects depending on the genetic background of the strain. This is illustrated in Figure 3 , where the strains in the “none”, “13”, “1” and “1+13” categories, which all carry the avirulent allele for G_07189, have different effects, if not always significantly, on the phenotype. These different effects form a gradient depending on the QTL combinations. Such a gradient in phenotypes, or variation of magnitude of the effect of quantitative pathogenicity on the phenotype, was recently demonstrated to also occur depending on the allele of the Avr gene in an isogenic background (Meile et al., 2023).

We are able to present conclusive evidence that an Avr gene can have a quantitative or qualitative effect on the phenotypes depending on the fungal genetic background. Pathogenicity is therefore not a strictly quantitative or qualitative variable, but rather fits somewhere on the continuum that they form. Recently, another example of an avirulence factor being involved in a quantitative interaction was proposed within the Leptosphaeria maculans-rapeseed pathosystem (Jiquel et al., 2021). A gene-for-gene interaction was demonstrated between LmSTEE98, a ‘late effector’ encoded by AvrLmSTEE98, and RlmSTEE98, encoded by the corresponding resistance gene. As a ‘late effector’, AvrLmSTEE98 is expressed during stem colonization which is considered to be polygenic and quantitative.

On the plant side, it has been previously suggested that quantitative resistance genes in plants could be weaker forms of major resistance R genes (Poland et al., 2009) and that most wheat resistance loci are matched by a specific effector (Plissonneau et al., 2018). We therefore propose that G_07189 could interact with the wheat resistance QTL identified in ‘Renan’ on chromosome 5D, designated as Stb20q, as this was the only QTL identified in ‘Renan’ which had specific resistance to I07 (Langlands-Perry et al., 2022). I07 carries the avirulent allele for G_01789, meaning that this gene could correspond to AvrStb20q.

In a classic R/Avr gene-for-gene relationship, disease resistance is not considered durable due to the high probability of adaptation by the fungus with mutations in the Avr gene (Niks et al., 2011; Hartmann et al., 2017; Meile et al., 2018; Cowger and Brown, 2019). Here, we observed no loss of G_07189 in the fungal populations or indeed any mutations in the effector-like features of the corresponding protein, i.e., the signal peptide and cysteine residues, inferring a potential fitness cost of such a loss. We found that the validated effector G_07189 is highly conserved among French Z. tritici strains and shows a signature of positive diversifying selection. This is also what was observed for Avr3D1, the only other cloned effector in this pathosystem known to be involved in quantitative interactions (Meile et al., 2018). In the latter case however, a high level of diversity was observed in fungal populations. G_07189 has the capacity to evolve to adapt to wheat resistance, as demonstrated by the diversifying selection signatures in three residues. The low diversity and fact that the virulent, or most pathogenic, isoform (I05), is not the most abundant in recent strains may therefore seem paradoxical. Two potential explanations for this paradox are, first, that the corresponding resistance gene, which we hypothesize to be Stb20q, has not been largely deployed in the French cultivar landscape, meaning that no selection due to this resistance has been imposed on fungal populations. The cultivar on which Stb20q was identified is ‘Renan’ (Langlands-Perry et al., 2022), mostly used in organic farming, and therefore not one of the majority cultivars in the landscape. The other explanation is that the selective pressure imposed on G_07189, despite being present, is low enough that adaptation is slowed down compared to what is observed with major R/Avr interactions. Indeed, the least pathogenic allele (I07) does not stop the fungus from completing its life cycle, as illustrated by the presence of pycnidia on the infected leaf surface. This second hypothesis fits in with the narrative that quantitative resistance is more durable than its qualitative counterpart.

An example of a highly conserved effector under diversifying selection is APikL2 in M. Oryzae (Bentham et al., 2021). Research has shown that a single amino-acid polymorphism is sufficient to evade host recognition, however, the evolutionary driver of this polymorphism was attributed to the expansion of the host target spectrum rather than immune receptor evasion (Bentham et al., 2021). Moreover, similar to what we observe with G_07189, diversifying residues are located at a specific part of the protein (Bentham et al., 2021). This example opens a new avenue of possibilities as to the molecular interactions in which G_07189 could be involved.