Analysis of mutations in West Australian populations of Blumeria graminis f. sp. hordei CYP51 conferring resistance to DMI fungicides

Powdery mildew caused by Blumeria graminis f. sp. hordei (Bgh) is a constant threat to barley production but is generally well controlled through combinations of host genetics and fungicides. An epidemic of barley powdery mildew was observed from 2007 to 2013 in the West Australian wheatbelt (WA). We collected isolates, examined their sensitivity to demethylation inhibitor (DMI) fungicides and sequenced the Cyp51B target gene. Five amino acid substitutions were found of which four were novel. A clear association was established between combinations of mutations and altered levels of resistance to DMIs. The most resistant genotypes increased in prevalence from 0 in 2009 to 16% in 2010 and 90% in 2011. Yeast strains expressing the Bgh Cyp51 genotypes replicated the altered sensitivity to various DMIs and these results were confirmed by in silico protein docking studies.


Comparative growth rate assay of transformants 139
The growth rate of transformants was assessed using the Gen 5 data analysis software 140 (BioTek Instruments, Inc. Winooski, VT) where duplicate cultures of replicate transformants 141 were grown in SD GAL + RAF medium (SD medium) overnight at 30°C. One hundred 142 microliters of each overnight culture, at 10 5 cells ml -1 , was used to inoculate 3 wells 143 containing 200μl SD medium ±3μg ml -1 doxycycline. Cultures were incubated without light 144 at 30°C, and the optical density at 600nm (OD600) was measured every 15min for 12 days in a 145 Synergy TM HT Multi-Mode Microplate Reader (BioTek Instruments, Inc Winonski, VT). The 146 mean maximum growth rate for each strain ± doxycycline was determined on the basis of the 147 greatest increase in OD over a 2h period (Supporting Information Table S4). the heme cavity of the wild type and variant protein models was determined using Pocket-158 Finder (Leeds, UK) based on Ligsite (Hendlich et al., 1997). 159 160

DNA sequencing 162
A trial set of Bgh isolates were assessed for susceptibility on detached leaves to fungicides 163 then used in WA to control powdery mildew. Substantial variation in resistance was 164 observed. Due to the complexities of the phenotyping assay, we decided to sequence the 165 CYP51 gene first and then determine fungicide sensitivity of isolates from each genotype 166 class. 167 Primers were designed spanning both the coding and promoter region of the single CYP51B 168 gene (Becher & Wirsel, 2012) in Bgh (Supporting Information Table S3, Supporting  169 Information Fig. S1). The Bgh51wt DH14 sequence was used as a reference (Spanu et al., 170 2010). CYP51 was sequenced from 76 Australian isolates collected from 2009 to 2013, 171 including three from Eastern Australia. No indels were found in the promoter but two 172 synonymous and five non-synonymous changes were identified in the coding region (Fig. 2). 173 All Australian isolates carried previously seen the tyrosine to phenylalanine mutation at 174 amino acid position 136 (Y137F) (Wyand & Brown, 2005, Délye et al., 1998. All three 175 isolates from the east of Australia carried two synonymous changes at nucleotides 81 and genotypes were distinguished. Isolates collected in WA were either F137/T524 (genotype 2) 180 or F137/I304/G330/T524 (genotype 4) while isolates from other Australian states were either 181 F137/E172 (genotype 3) or F137 (genotype 1). Mutations I304 and G330 were consistently 182 found together in the same isolates ( Figure 2). 183 There was both spatial and temporal variation in the frequency of genotypes (Figure 1, 184 Supporting Information Table S1). All isolates collected in 2009 were wild type at CYP51 seasons; 99 of the 116 WA isolates collected in 2011 contained the T524 mutation. These 187 mutants were found in all major WA barley growing areas ( Figure 1). 188

DMI sensitivities of Bgh isolates 189
The sensitivities of 18 Bgh isolates (2 isolates from genotypes 2 and 3; 7 from 1 and 4) were 190 determined using detached barley leaves inserted into DMI-amended agar. The results varied 191 between genotype and fungicide ( Figure 3, Supporting Information Fig. S4). There was no 192 significant differences in the mean EC50 of S524 isolates (genotypes 1 and 3) or between 193 isolates with the T524 mutation (genotypes 2 and 4). Isolates of genotypes 2 and 4 were 194 found to have significantly higher mean EC50 values than genotypes 1 and 3 for most of the 195 DMIs tested. The mean EC50s for T524 genotypes ranged from 1.88 ug.ml -1 for triadimefon, 196 3.73 ug.ml -1 for propiconazole to 29.88 ug.ml -1 for tebuconazole. The estimated resistance 197 factors ranged from 3.41 for propiconazole to 17.6 for tebuconazole. However, for 198 fluquinconazole (used solely in WA in seed dressing formulations) mutant T524 genotypes 199 were marginally more sensitive (EC50 4.73 ug.ml -1 ; RF = 0.58). Unfortunately, due to 200 quarantine restrictions we were not able to phenotype the Bgh CYP51 DH14 isolate carrying 201 the wild type Y137 allele. 202

Heterologous expression in yeast 203
The Bgh51wt gene was synthesized and cloned into S. cerevisiae YUG37:erg11 with a 204 doxycycline repressible promoter. The S. cerevisiae Bgh51wt transformant was able to grow 205 in the presence of doxycycline (Supporting Information Fig. S2) and for most variants there 206 was no significant difference in the growth rates in the absence of doxycycline. Two S. 207 cerevisiae Bgh51 variants (pYES-Bgh51_Y137F/S524T/R330G and pYES-208 Bgh51_Y137F/M304I/R330G/S524T) had significantly lower rates and were therefore 209 removed from all further in vitro analysis. 210 The DMI sensitivities of S. cerevisiae strains expressing Bgh51 variants which restored 211 growth on doxycycline-amended medium were determined (Supporting Information Table 4) 212 and resistance factors were calculated (Table 1). Modest RFs were associated with the solo 213 K172E and M304I mutations. RFs for the S524T mutation varied from 0.5 for 214 fluquinconazole to 12.4 for propiconazole. The combination of F137 and T524 had much

Structural modelling 218
Protein variants of all Bgh51 genotypes were modelled in silico (Supporting Information Fig  219   S5). The effect that each mutational change had on the volume of the heme cavity containing 220 the DMI binding site and the morphological changes to the cavity access channel were 221 determined (Table 2). Modest volume increases in binding cavity were observed with the 222 introduction of the solo mutations; a 17.7% increase with K172E and 39.6% increase in 223 volume with Y137F. Mutation S524T was an exception, with an increase in the volume of the 224 heme cavity by 73.2% compared to that of the wild type model. The combination of 225 F137/T524 gave an even more substantial increase in volume of 83.9%. Table 2 also shows 226 the estimated distances between amino acids Y226 (Y222) and S521 (S506), which span the 227 entrance to the channel leading to the DMI binding site. Modelling simulations predicted that 228 all Bgh CYP51 mutations observed in WA would cause a restriction in the diameter of the 229 access channel when compared to wild type Bgh CYP51. The most dramatic decrease was 230 observed with the introduction of the Y137F mutation, which caused a 28.5% decrease in 231 channel diameter compared to the wild type model. The combination of Y137F and S524T in 232 a single model did not result in a further significant restriction (Table 2). 233 234 Further morphological changes were observed that may impact DMI binding. In particular 235 conformation of a loop of beta turn running from S520 (S505) to F523 (F508) is markedly 236 different in the Y137F genotype to that of the wild type, with the result that it projects into the 237 cavity. A similar constriction is observed for the F137/T524 mutant (Supporting Information 238 Fig. S5b). However, in this case it is also accompanied by a substantial increase in cavity 239 volume (Table 2) the presence of the Y137F mutation with strong resistance to triadimenol. We were unable 275 import the wild type CYP51 Bgh isolate DH14 in Australia due to quarantine restrictions. 276 However Y137F expression in the heterologous yeast system showed only modest decreases 277 in sensitivity to most DMIs including triadimenol (Table 1). This suggests that Y137F would 278 lead to only small reductions in the DMI field efficacy. The ubiquity of Y137F in Australia 279 suggests two possibilities; (1) the limited fungicide use in eastern of Australia has been 280 sufficient to select for this mutation or (2) the wild type CYP51 genotype has never been A search was conducted on the CYP51 mutations in other fungal species reported as 283 conferring a reduction in DMI sensitivity. The Bgh51 amino acid sequence of Australian 284 genotypes was aligned with Z. tritici CYP51 (Figure 2). Mutational changes at the amino 285 acids 137, 304, 330 and 524 fall in regions conserved between Bgh and Z. tritici (Becher & 286 Wirsel, 2012). Amino acids 136 and 509 in Bgh51 correspond to 137 and 524 in Z. tritici 287 which have previously been correlated with alterations in DMI sensitivity (Cools et al., 288 2011). The combination of Y137F and S524T was associated with substantial RFs in both the 289 Z. tritici strains and the yeast transformants. This study did not test fluquinconazole or the 290 solo Y137F genotype in the yeast system. 291 In the current study the combination of Y137F and S524T encoded a CYP51 with a marked 292 decrease in sensitivity to tebuconazole and propiconazole in both the mildew strains and the 293 yeast system. This may be sufficient to account for the field failure ( Figure 3). Increases in 294 heme cavity volume and restriction of the access channel in Y137F/S524T protein models 295 correlate well with the significant RF obtained ( Figure 5). A high RF for propiconazole was 296 also observed for the Y137F/S524T Bgh CYP51 construct when expressed in the yeast 297 system. 298 Structural modelling suggests that there are two main mechanisms that underpin the 299 emergence of DMI resistance associated with mutational changes in Bgh51. The first 300 mechanism is similar to that observed in Z. tritici CYP51 (Mullins et al., 2011), where the 301 gross volume of the heme cavity increases with successive mutations (Table 2). There 302 appears to be a correlation between the increase in cavity volume and the RFs reported in 303 table 1. It is likely that any increase in heme cavity volume would perturb the orientation of 304 the DMI ligand and hence its binding to the heme. This therefore differentiates the smaller 305 DMI ligands such as tebuconazole and epoxiconazole. 306 The second mechanism at play provides a means of linking structural changes with 307 phenotypic changes in a measurable way. Changes in distances between specific pairs of 308 residues that border the cavity result in changes to the diameter of the access channel. The 309 limiting of the binding surface between Y226 and S314 (S312) appears to correlate well with 310 resistance to tebuconazole. The narrowing of the access channel between Y226 and S521 311 correlates particularly well, especially when tempered by consideration of the effects of each 312 variant on the cavity volume. This is demonstrated by the result obtained when the product of 313 the percent change in the heme cavity volume is multiplied by the percent change in the distance between Y226 and S521 ( Figure 5). All the variants that contain F137 demonstrate a 315 substantially reduced distance between Y226 and S521 (Table 2). When one of the 316 mechanisms is employed, moderate resistance factors are observed (F137 (access channel 317 narrowing); T524 (substantial increase in cavity volume)). Although, when both mechanisms 318 act together there is a strong correlation between the structural changes and the very high 319 resistance factors of the F137/T524 mutants in the presence of tebuconazole. The in silico 320 creation of Bgh51wt and mutant CYP51 protein variants opens the possibility of future 321 docking studies employing novel or unregistered DMI fungicides. This will allow the 322 prediction the effectiveness of any new product prior to in planta testing. Furthermore, we 323 can now recommend bespoke spray regimes depending on which Bgh51 genotype is present 324 in the field. 325 326 One of the major resistance strategies used for fungicides is to mix active compounds with 327 different MOA because isolates with mutations conferring resistance to one fungicide will 328 most likely still be sensitive to the second (Van Den Bosch et al., 2014). This strategy 329 requires that there is no positive cross resistance between the two fungicides and so generally 330 rules out mixtures of the same MOA. However some cases of negative cross-resistance 331 within a single MOA group has been shown with Z. tritici isolates which are highly resistant 332 to tebuconazole but fully susceptible to prochloraz (Leroux et al., 2007, Fraaije et al., 2007. 333 The negative cross-resistance shown in both the Bgh (Figure 3) and yeast expression studies 334 (Table 1)        GGCATCGTGGATTATCTACC -

± 582
a Values represent the greatest increase in OD600 (10 -2 ) in the absence (-DOX) or presence 583 (+DOX) of doxycycline over a 2h period and are the means of 12 independent replicates ± 584 standard deviations. Growth rates significantly different from the wild type construct (pYES-585 Bgh51wt) are given in bold. Numbers refer to amino acid positions in Z. tritici. 586 Supp Table 4. In vitro EC50 and resistance factors of T524 (T509) and S524 (S509) Bgh isolates when exposed 588 to currently registered triazoles in WA.