Aspergillus fumigatus can display persistence to the fungicidal drug voriconazole

Aspergillus fumigatus is a filamentous fungus that can infect the lungs of patients with immunosuppression and/or underlying lung-diseases. The mortality associated with chronic and invasive aspergillosis infections remain very high, despite availability of antifungal treatments. In the last decade, there has been a worrisome emergence and spread of resistance to the first line antifungals, the azoles. The mortality caused by resistant isolates is even higher, and patient management is complicated as the therapeutic options are reduced. Nevertheless, treatment failure is also common in patients infected with azole-susceptible isolates, which can be due to several non- mutually exclusive reasons, such as poor drug absorption. In addition, the phenomena of tolerance or persistence, where susceptible pathogens can survive the action of an antimicrobial for extended periods, have been associated with treatment failure in bacterial infections, and their occurrence in fungal infections already proposed. Here, we demonstrate that some isolates of A. fumigatus display persistence to voriconazole. A sub-population of the persister isolates can survive for extended periods and even grow at slow rates in the presence of supra-MIC (Minimum Inhibitory Concentration) of voriconazole and seemingly other azoles. Persistence cannot be eradicated with adjuvant drugs or antifungal combinations and seems to reduce the efficacy of treatment for certain individuals in a Galleria mellonella model of infection. Furthermore, persistence implies a distinct transcriptional profile, demonstrating that it is an active response. We propose that azole persistence might be a relevant and underestimated factor that could influence the outcome of infection in human aspergillosis. IMPORTANCE The phenomena of antibacterial tolerance and persistence, where pathogenic microbes can survive for extended periods in the presence of cidal drug concentrations, have received significant attention in the last decade. Several mechanisms of action have been elucidated, and their relevance for treatment failure in bacterial infections demonstrated. In contrast, our knowledge about antifungal tolerance and, particularly, persistence are still very scarce. In this study, we have characterised the response of the prominent fungal pathogen Aspergillus fumigatus to the first line therapy antifungal voriconazole. We comprehensively show that some isolates display persistence to this fungicidal antifungal and identify various potential mechanisms of action. In addition, using an alternative model of infection, we provide initial evidence to suggest that persistence may cause treatment failure in some individuals. Therefore, we propose that azole persistence is an important factor to consider and further investigate in A. fumigatus.

tolerance and, particularly, persistence are still very scarce. In this study, we have characterised the 56 response of the prominent fungal pathogen Aspergillus fumigatus to the first line therapy antifungal 57 voriconazole. We comprehensively show that some isolates display persistence to this fungicidal 58 antifungal and identify various potential mechanisms of action. In addition, using an alternative model 59 of infection, we provide initial evidence to suggest that persistence may cause treatment failure in 60 some individuals. Therefore, we propose that azole persistence is an important factor to consider and 61 further investigate in A. fumigatus. 62

INTRODUCTION
The same definition for fungi

Tolerance
Ability of all cells of a susceptible, genetically isogenic strain to survive, and even grow at slow rates, for extended periods in the presence of drug concentrations that are greater than the MIC.
In the presence of fungistatic drugs, capacity of all or high proportion of the cells of a susceptible isolate to grow in the presence of drug concentrations that are greater than the MIC.

Persistence
Ability of a small subpopulation of cells of a susceptible, genetically isogenic strain to survive, and even grow at slow rates, for extended periods in the presence of drug concentrations that are greater than the MIC.
In the presence of fungicidal drugs, capacity of a small fraction of cells of a susceptible isolate to survive, and even grow at slow rates, for extended periods in the presence of drug concentrations that are greater than the MIC. 100 Table 1. Current and proposed definitions for the phenomena of resistance, heteroresistance, 101 tolerance and persistence in pathogenic fungi (see discussion). 102

103
In fungi, tolerance has been often described as "trailing growth" and its relevance for infection has 104 been mostly disregarded. However, a recent landmark study in Candida albicans has characterised 105 tolerance as a distinct, strain-specific feature and provided evidence for its relevance in persistent 106 candidemia [17,21]. In Aspergillus, trailing growth is known to be prominent in the presence of 107 caspofungin [22,23] apparently due to single strain heterogeneity [24]. Moreover, at higher 108 concentrations of this drug some isolates can resume normal growth, an effect known as "Eagle" or 109 paradoxical effect (see review [25] for more information), which we have recently demonstrated as a 110 tolerance phenotype [26]. The possible relevance of the paradoxical effect in clinical practice is still 111 under debate, but there is evidence suggesting that it might be important [27], and concern has been 112 raised by some clinicians [28]. Remarkably, it seems that the phenomenon of tolerance in fungi is seen 113 with static drugs, e.g. azoles for C. albicans and echinocandins for A. fumigatus. However, the 114 possibility that A. fumigatus can display tolerance or persistence to azole antifungals had not been 115 previously investigated. Interestingly, in contrast to C. albicans and Cryptococcus neoformans, azoles 116 have been shown to have fungicidal activity against A. fumigatus [29][30][31], which implies that the 117 underlying mechanisms are likely different. Whilst these phenomena have so far only been 118 investigated in bacteria or yeasts, the approaches to detect and investigate azole tolerance or 119 persistence need to be tailored in filamentous fungi, like A. fumigatus, as these organisms form 120 multicellular hyphae. Using various complementary approaches, here we show that small 121 subpopulations of certain A. fumigatus isolates can survive and even grow at slow rates at supra- MIC 122 concentrations of the fungicidal drug voriconazole. 123

RESULTS 125
Some Aspergillus fumigatus isolates show persistence to voriconazole 126 To determine whether A. fumigatus can display tolerance or persistence to voriconazole, we examined 127 a collection of isolates consisting of 9 environmental, 10 clinical (gift from Prof Paul Dyer, collection 128 PD-47-XX, here shortened as PD-XX), 5 common laboratory strains and one resistant control (RC) 129 harbouring the TR34/L98H mutations in the cyp51A locus (Table  S1 in 130 https://doi.org/10.5281/zenodo.7021623), using disc diffusion assays. We evenly spread 4×10 4 131 conidia of the isolates on RPMI agar plates, placed a 6 mm disc in the centre, containing 10 μL of 132 voriconazole (0.8 mg/mL), and incubated them for 5 days. We observed variability in the sizes of the 133 inhibition halos, reflecting differences in susceptibility of the isolates (Figs. 1A and S1). As expected, 134 the RC strain did not show a proper inhibition halo (Figs. 1A and S1). In contrast, 15 of the 20 isolates 135 showed a clear and well-defined inhibition zone. Interestingly, we found that 5 isolates were able to 136 form colonies within the halo of inhibition, and 4 in 5 of these isolates (PD-9, PD-104, PD-254 and PD-137 266, Fig. S1) formed only a few small colonies, which might be indicative that a few conidia are able 138 to germinate and grow a little in the presence of a supra-MIC concentration of voriconazole ( Fig. 1A  139 and S1). The remaining isolate (PD-256) did not show any inhibition halo, suggesting that it is a 140 resistant isolate. 141 We then assessed the MICs of the original isolates and their derived colonies of the halo (CoHs). A 142 complication of working with a filamentous fungus is that the CoHs need to be grown on a new plate 143 in order to harvest spores (conidia) to be used for MIC determination. We therefore decided to assay 144 two different conidia-harvesting conditions to distinguish the different phenotypes: re-growing the 145 CoHs on solid RPMI in the absence or in the presence of a low concentration of voriconazole (0.12 146 µg/mL). It would be expected that for conidia grown in the presence of the drug, both resistance and 147 heteroresistance are detected as an increment in the isolate's MIC. In contrast, for conidia grown in 148 the absence of the drug, the reversible increase in MIC that is characteristic of heteroresistance would 149 be lost, whilst a stable genetic-based increment in MIC that defines resistance would be maintained. 150 Finally, persistence should not cause a change of the isolate's MIC independently of the presence of 151 the drug in the conidia-harvesting medium. Using conidia obtained from both conditions, absence and 152 presence of voriconazole, we found that the original isolate did not show an inhibition halo 153 Fig. 1A and S1) had a very high MIC (>8 µg/mL), demonstrating that it is a resistant isolate (Table 2). 154 CoHs formed by PD-254 and PD-266, which upon re-inoculation did not show an inhibition halo (Fig.  155 S1), showed an increased MIC compared to the parental isolate ( Table 2). The colony picked from PD-156 254 showed increased MIC when re-grown on both media with and without voriconazole, suggesting 157 that the CoH may have acquired a mutation that confers a stable resistance phenotype. In contrast, 158 PD-266, which already had an elevated MIC, only showed an increased MIC when re-grown on medium 159 containing voriconazole, suggesting that this strain may be heteroresistant. Finally, CoHs from isolates 160 PD-9 and PD-104 showed the same MICs as their original isolates after a passage in the absence or 161 presence of voriconazole (Table 2). This suggests that they are not resistant or heteroresistant 162 derivatives of the original isolates. Indeed, repetition of the disc assay with the original isolates and 163 with CoH re-grown in the presence of voriconazole, showed a similar level of colony appearance in 164 the original strains and their derived CoHs for PD-9 and PD-104 (whereas it grew to the edge of the 165 disc for PD-266, reflecting again a transient increase in its MIC) (Fig. S1). Although no formal 166 measurement of growth rate was performed, the clear difference in size between single persister (PD-167 9 and PD-104) and heteroresistant colonies (PD-266) after the same time of incubation ( Fig. S1) 168 indicate that the growth rate of persister colonies is reduced. Interestingly, all isolates showed a small 169 increase (one dilution) in MIC when re-grown in the presence of drug (Table 2). This may be due to 170 the development of conidia adapted to the growing environment in preparation for the subsequent 171 germination, an effect recently described in Aspergillus spp [32]. Next, we tested if CoHs could also be 172 detected using E-tests strips, which is a well-established method to measure the MIC and is thus more 173 quantitative than disc diffusion assay [33]. In agreement with disc diffusion assays, CoHs developed 174 inside the halo of inhibition created by E-tests on persister isolates (PD-9 and PD-104), but not on non-175 persister isolates (ATCC and PD-60) (Fig. S2A). Finally, to verify that the persistence phenotype is 176 stable, we sequentially passaged CoHs isolated from PD-9 and PD-104 on PDA media (without 177 voriconazole) and performed disc diffusion assays every two/three passages. The halo of the disc and 178 the relative level of persistence were maintained for 10 passages (Fig. S2B) A. fumigatus in more detail, we followed a protocol to detect tolerance/persistence in bacteria in 186 which the disc containing the drug, after a period of incubation, is switched, with another one 187 resistance [35,36], and in bacteria persistence has been shown to correlate with the evolution of 240 resistance [37,38]. However, we found again that all strains harboured a completely wild-type 241 sequence (not shown), suggesting that hmg1 is not related with voriconazole persistence. 242 In conclusion, we have observed that certain A. fumigatus isolates can survive and grow at slow rates 243 for extended periods of time in the presence of supra-MIC concentrations of voriconazole, therefore, 244 these isolates display persistence to voriconazole. 245 246 247 248

Figure 1. Certain Aspergillus fumigatus isolates display persistence to voriconazole. 249
A) In disc-diffusion assays (10 µL of 0.8 mg/mL voriconazole), the susceptible isolate ATCC46645 never 250 grew any colony in the inhibition halo, the resistant control (RC) isolate grew up to the edge of the 251 disc, as did the strain PD-256. The persister isolate PD-104 was consistently able to grow a few small 252 colonies. Plates were incubated for 5 days at 37°C. B) Inspection of the wells of a broth dilution assay 253 under the microscope 72 hours after inoculation revealed that the non-persister isolates ATCC and 254 PD-60 displayed only limited microscopic growth at the MIC concentration, and all conidia remained 255 non-germinated at higher concentrations. In contrast, the persister strains PD-9 and PD-104 showed 256 noticeable microscopic growth up to three fold (3x) the MIC concentration. Scale bar= 132.5 µm. C) 257 Full content of the well containing the maximum concentration of voriconazole (8 µg/mL) was plated 258 on rich media PDA plates and (CFUs) counted 48 hours after inoculation. Persister isolates grew 259 significantly more CFUs than non-persister isolates (PD-9 VS ATC46645C p=0.002 and PD-104 VS 260 ATCC46645 p=0.0331), demonstrating that these strains remain viable upon azole treatment for a 261 longer period. Three independent experiments with three biological replicates were performed, the 262 graph represents the means and SD, and data was analysed using the Brown- Initially, to rule out that persistence could be explained by basal differences in germination or growth, 277 we examined the germination and growth rates of two persister (PD-9 and PD-104) and two non-278 persister (ATCC and PD-60) isolates. We found that all isolates germinated at equivalent rate and ratio 279 ( Fig. S5B) and grew similarly on both solid ( Fig S5C) and liquid (Fig. S5D) RPMI media. Once excluded 280 differences in germination or growth, we tested the influence of the morphological stage on 281 persistence by incubating the RPMI plates inoculated with the fungus for 8 or 16 hours before adding 282 the drug to the disc. We found that the persister strains PD-9 and PD-104 were also able to form 283 colonies in the halo when drug was added 8 hours after the beginning of incubation ( Fig. 2A), 284 suggesting that persistence is not determined by the developmental stage. We could not draw 285 definitive conclusions for the 16 hours grown hyphae, as the background growth was too dense to 286 undoubtedly differentiate specific persister colonies ( Fig. 2A). Nevertheless, we could clearly detect a 287 growing colony in the voriconazole halo for the strain PD-104 ( Fig. 2A), proving that at least this strain 288 can display persister growth when short hyphae are challenged with the drug. Interestingly, it has 289 previously been reported that morphological status alters A. fumigatus susceptibility profiles against 290 various drugs including voriconazole [42], which supports the notion that persistence is a different 291 phenomenon, triggered by distinct mechanisms, and independent of MIC. 292 To determine if persistence is influenced by the growth medium, we performed a voriconazole disc 293 diffusion assay with the PD-104 isolate and the wild-type control ATCC46645 on PDA rich media or 294 AMM. The PD-104 strain showed persister growth on both media, and it seemed to be able to form a 295 higher number of COHs on PDA rich media (Fig. 2B). The ATCC46645 wild-type strain did not show 296 persistence on AMM, but surprisingly it did form colonies in the halo on PDA media (Fig. 2B). To 297 investigate this in more detail we repeated the disc assay using two different rich media, PDA and 298 Sabouraud, with more strains, CEA10, PD-9, and PD-60 ( Fig. 2C) and found that all strains were able 299 to form colonies in the halos. The MICs of the isolates, measured by broth dilution assay, were equal 300 on Sabouraud and RPMI (Fig. 2D), indicating that the effect of the drug is not reduced in rich medium. 301 Therefore, the ability of A. fumigatus to display persistence seems to be medium dependent, 302 suggesting that a rich nutrient environment favours survival at supra-MIC concentrations. This is 303 supported by a recent study that showed that a rich metabolic environment can promote azole 304 tolerance in Saccharomyces cerevisiae [43]. Nevertheless, it should also be considered that 305 voriconazole diffusion may be affected in solid rich media, which could be confounding this 306 observation. 307 Finally, we also considered that persistence might be affected by the age of the spores (all experiments 308 are performed with freshly harvested spores). To check this, we assayed disc diffusion assays using 309 10-weeks old spores. We did not observe remarkable differences with respect to fresh spores, 310 although in most of the experiments we could see one colony in the halo of the non-persister PD-60 311 ( Fig. 2E), suggesting that aged spores might have a slightly higher persistence potential, at least for 312 some isolates. 313 to be determined by the growth media. 316 A) Conidia of the different isolates were inoculated and incubated for 8 or 16 hours before the disc 317 containing voriconazole (10 µL of 0.8 mg/mL) was added to the RPMI plate. At 8 hours, when conidia 318 have germinated, the persister isolates PD-9 and PD-104 were still able to grow small colonies in the 319 inhibition halo, whilst the non-persister strains were not. At 16 hours, when conidia have already 320 formed hyphae, the background growth made impossible to distinguish colonies in PD-9, but a clear 321 one could be detected in PD-104. B) On Aspergillus minimal medium (MM), the persister isolate PD-322 104 was able to form colonies in the halo, but the non-persister strain ATCC was not. C) On rich media 323 PDA and Sabouraud (Sab) all isolates were able to form conidia in the halo. D) This happened even the 324 MIC of all isolates was the same in commonly assayed RPMI and the rich media Sab. E) 10-week old 325 spores displayed a similar persister phenotype as freshly isolated spores. Often a single colony could 326 be detected in the halo of inhibition on the non-persister PD-60, suggesting that aged spores might 327 have slightly more persistence capacity. 328 329 Aspergillus fumigatus persistence to voriconazole is independent of stress and cannot be inhibited 330 with adjuvant or antifungal drugs. 331 Changing environmental conditions activate signalling cascades that trigger transcriptional adaptation 332 and cell wall alterations [44][45][46][47]. Therefore, we reasoned that the capacity of certain isolates to survive 333 and grow in supra-MIC concentrations of voriconazole might be influenced by environmental 334 stressors. Indeed, in C. albicans mutants and inhibitors of stress response pathways eliminate 335 tolerance [48]. To investigate this possibility, we analysed voriconazole persistence of PD-104 in the 336 presence of hypoxic (1% O2), oxidative (0.01% H2O2), osmotic (150 mM NaCl), membrane (0.05% SDS) 337 and cell wall (10 µg/mL CalcoFluor White) stress. Surprisingly, in contrast to C. albicans [21], we found 338 that most environmental conditions did not influence persistence in A. fumigatus (Fig. 3A). This 339 suggests that the underlying mechanism(s) of persistence in A. fumigatus are likely different from the 340 previously proposed mechanisms of tolerance in C. albicans. The only condition that influenced A. 341 fumigatus persistence was hypoxia, which could prevent growth in the halo (Fig. 3A). As we had 342 observed above that persistence was influenced by the growth medium, we wondered if hypoxia could 343 prevent persistence also on rich medium. As shown in Figs. 2B and 2C, all isolates were able to grow 344 colonies in the halo when grown on the rich medium YAG on normoxia (Fig. 3B). However, persistence 345 was eliminated under hypoxia for all strains, except for PD-104, for which it was reduced but 346 apparently not completely eradicated. Therefore, it seems that hypoxia reduces persistence, but the 347 nutritional composition of the medium also influences its impact on the phenomenon. 348 In C. albicans, fluconazole tolerance (but not resistance) can be prevented with the use of adjuvant 349 drugs that block general stress signalling pathways [21]. To further investigate if the underlying 350 mechanism of persistence may be different in A. fumigatus, we tested the effect of various drug 351 adjuvants that were previously shown to eliminate tolerance in C. albicans [21]: geldanamycin (0.8 352 µg/mL), an inhibitor of heat shock protein (Hsp90) [49,50], FK506 (4 ng/mL), an inhibitor of calcineurin 353 [51], H-89 (4 µg/mL), an inhibitor of the cAMP-dependent protein kinase (PKA) [52], rapamycin (6.25 354 µg/mL), an inhibitor of the mammalian target of rapamycin (mTOR) [53], and tunicamycin (10 µg/mL), 355 an inducer of the unfolded protein response pathway [54]. In contrast to C. albicans, the use of 356 adjuvant drugs did not prevent persistence of A. fumigatus isolates (Fig. 3C). We also tested lovastatin 357 (8 µg/mL) and simvastatin (2 µg/mL), as statins inhibit 3-hydroxy-3-methylglutaryl-coenzyme A 358 (HmgA) [55], an enzyme in the same metabolic pathway as the target of azoles, and these two were 359 previously shown to have antifungal activity against Aspergillus spp [56]. However, statins were also 360 not able to prevent persister growth (Fig. 3B). Finally, as efflux of antifungals has been proposed to 361 play a role in C. albicans tolerance [48], we tested if the efflux inhibitors milbemycin A oxim (8 µg/mL) 362 [57] or clorgyline (63.5 μg/mL) [58] can prevent A. fumigatus persistence. These compounds seemed 363 to be able to diminish the persister capacity of the strains, as only one CoH per plate could be detected, 364 and these colonies were exactly at the edge of the halo (Fig. 3C). In conclusion, adjuvant drugs cannot 365 prevent A. fumigatus persistence to voriconazole, and efflux inhibitors may affect this process and 366 deserve further study. 367 Next, we evaluated if persistence can be eradicated using combinatorial treatment with the other 368 classes of antifungal drugs in clinical use. However, neither amphotericin-B (1 µg/mL) nor caspofungin 369 (0.5 µg/mL) could prevent persistence (Fig. 3D). Interestingly, these antifungal drugs could also not 370 impede growth in the halo of the presumed heteroresistant strain PD-266 (Fig. 3D). Therefore, it 371 seems that combinatorial treatment with other antifungals cannot prevent persistence in A. 372 fumigatus. 373 hypoxia, and cannot be eliminated with adjuvant or combinatorial treatments. 376 A) Hypoxia was the only stress that could prevent persistence, whilst neither oxidative (H2O2), cell wall 377 (SDS or CFW) or osmotic (NaCl) stress influenced persistence. B) Hypoxia completely eradicated 378 persistence also on the rich medium YAG for ATCC, PD-60 and PD-9, but only reduced persistence for 379 the PD-104. C) The use of adjuvant drugs could not prevent persistence. D) Combinatorial treatment 380 with amphotericin-B and caspofungin did not prevent persistence. All plates were incubated with 10 381 µL of 0.8 mg/mL voriconazole added to the disc and the specific condition (stress, adjuvant or 382 combinatorial drug) as described in the text. Plates were incubated for 5 days at 37°C. All plates and 383 conditions were repeated in at least two independent experiments. 384

385
The phenomenon of persistence can be observed with other azole drugs. 386 To check if A. fumigatus can display persistence in the presence of other azoles, we firstly employed 387 the disc diffusion assay adding 10 µL of a 3.2 mg/mL itraconazole solution. We found that the isolates 388 PD-104 and PD-266 were able to form colonies in the halo (Fig. S6). Upon re-inoculation of a CoH (re-389 grown in itraconazole containing media), the isolate PD-104 showed an inhibition halo of the same 390 size and displayed similar level of CoH appearance, whereas the PD-266 isolate was able to grow on 391 the whole plate and did not show any inhibition halo. Therefore, this suggests that, as observed with 392 voriconazole, the isolate PD-104 is persistent and the isolate PD-266 is heteroresistant to itraconazole. 393 We then performed a broth dilution assay with our four well-characterised isolates (non-persisters 394 ATCC46645 and PD-60 and persisters PD-9 and PD-104) to calculate the MIC for itraconazole and 395 isavuconazole, and looked under the microscope at supra-MIC concentrations ( Fig. S7A and S7B). We 396 found that the isolates PD-9 and PD-104 showed slight growth at 2X MIC, whilst ATCC46645 and PD-397 60 did not (Fig. S7B). Moreover, we inoculated the entire content of wells containing the highest 398 concentration of the drugs (8 µg/mL) on PDA plates, and found that a detectable number of conidia 399 from the isolates PD-9 and PD-104 remained viable after 48 hours of incubation in the presence of the 400 azoles ( Fig. S7C). Hence, even if more experiments need to be done for a detailed characterization of 401 persistence to these azoles, these results suggest that the same isolates that are persisters to 402 voriconazole could also display persistence to itraconazole and isavuconazole. 403 The transcriptome of persister growth suggest that Galactosaminogalactan and high expression of 404 sterol biosynthetic genes may be relevant to establish persistence. 405 To better understand the heterogeneous nature of persistence, we compared the transcriptome of 406 PD-104 grown under conditions favouring persistence (above MIC), sub-MIC and in the absence of 407 drug. We inoculated the spores on top of a nylon membrane placed on an RPMI solid plate. We put 408 the disc with voriconazole (0.8 mg/mL) on the membrane and incubated for 5 days at 37°C. Standard 409 halos of inhibition formed on the membrane, and persister colonies appeared for the PD-104 strain, 410 but not for the A1160 strain (Fig. S8A). We harvested mycelium (avoiding conidia as much as possible) 411 from plates without a voriconazole disc (No Drug), from the area equidistant from the border of the 412 plate and the inhibition halo (Low Drug) and the colonies in the halo (Persistence) (Fig. 4A). We had to 413 combine persister colonies from 20 plates per replicate in order to obtain sufficient material for RNA 414 extraction. Additionally, we harvested mycelia from A1160 with No and Low Drug under the same 415 conditions. We performed RNA-seq of two biological replicates/condition and compared the 416 transcriptomes of the following conditions (i) A1160 reach significance for any of them, which is possibly due to the low number of genes included in the 447 analysis. However, we believe this analysis provided interesting clues and can help to direct future 448 research. For the up-regulated genes, the strongest enrichment was for the biological processes 449 galactose and aldehyde metabolisms (Fig. 4B) and the molecular function oxidoreductase activity 450 (  DEGs showed that secondary metabolism, ergosterol metabolism and transport were key upregulated 473 biological processes (Fig. 4C)   genes identified, as detailed in the text. In addition, the sites of sampling for the no drug (plate without  498 voriconazole disc), low drug (light green circle at mid-distance of the inhibition halo) and persistence 499 (colonies inside the inhibition halos) conditions are shown. B-E The most significantly upregulated GO 500 biological processes are shown using the REVINGO tool for GO data visualization [67]. The full list of 501 upregulated GO terms can be found in Tables S2-S5 and Table 3. B) PD-104 Persister only genes 502 (aligned to the A1163 genome), C) all genes upregulated in Persistence (aligned to the pangenome), 503 D) PD-104 Persister only genes, (aligned to the pangenome) and E) genes that are most upregulated 504 in persistence compared to normal response to the drug. 505 506 Bacterial persistence has been proposed to be a sub-population event due to stochastic high 507 expression of relevant genes [68,69]. Accordingly, we reasoned that those genes that are upregulated 508 in Low Drug VS No Drug and Persister VS No Drug comparisons, but also in Persister VS Low Drug might 509 reveal those genes that are important to adapt to presence of the drug, but also that can create 510 persistence with higher levels of expression. We identified 18 genes that appeared as upregulated in 511 all three comparisons ( Fig. 4A and Table 3), indicating that they have higher levels of expression in 512 persistence VS normal response to the drug. GO enrichment analysis revealed sterol metabolism as 513 the most significantly upregulated biological process ( Fig. 4E and Table S7 in 514 https://doi.org/10.5281/zenodo.7021623). Indeed, those few genes had a very significant enrichment 515 in the KEGG pathway "steroid biosynthesis" ( Table 3), suggesting that high expression of genes in the 516 sterol biosynthetic route (including cyp51A) may enable the sub-population of persisters to survive 517 and grow in supra-MIC concentrations. Additionally, these highly expressed genes were enriched in 518 KEGG pathways related with cytochrome P450 dependent drug metabolism (Table 3), which may 519 indicate that detoxification of azoles is also important for persistence. Finally, expression of the cdr1B 520 transporter (AFUA_1G14330), known to be associated with azole resistance [70], was also detected 521 to be higher in persistence (Table 3). This suggests that a higher capacity to efflux azoles may also be 522 important for the persister phenotype. Finally, we performed a protein functional association analysis 523 with these 18 genes using the STRING database and platform [66] to search for functional correlations. 524 All 18 genes were matched to proteins and a significant interaction (p<1.0e-16) was found, involving 525 a highly interactive nodule of 6 proteins related with steroid biosynthesis (Fig. S8C  broth dilution assays in which we inoculated each isolate (ATCC, PD-9, PD-60 and PD-104) in two 543 different lines and added purified GAG at a concentration of 100 µg/mL to one of them. Comparison 544 of the MIC with and without GAG demonstrated that this polysaccharide did not affect the resistance 545 profile of the isolates (Fig. 5A). After reading the MIC at 48 hours, we added 10 µg/mL of the dye 546 calcofluor white (CFW) to the wells and imaged the entire well (87 photos using a 20X objective) at 547 the blue emission channel. Images were merged (Fig. 5B) and the number of germinated conidia/short 548 hyphae in the whole well calculated as explained in materials and methods (Fig. 5B). By these means 549 we could calculate that ~2% of the PD-9 spore population and ~1% of the PD-104 conidia were able to 550 germinate at 2X MIC, and ~0.35% for both isolates in 3X MIC. Interestingly, addition of GAG enhanced 551 the number of persisters in PD-104 (from 1.09% to 2.80%, p=0.0267 at 2X MIC and from 0.41% to 552 0.75% p=0.038) but not in PD-9 (Fig. 5B). Next, we inoculated the four isolates in duplicated lines of 553 96-well plates containing a high concentration of voriconazole (4 µg/mL) in all wells, and added GAG 554 to one line of each isolate. The entire contents of wells were plated 48 hours after inoculation on PDA 555 rich plates to count the number of viable CFUs. As observed before (Fig. 1C), we found that the 556 persister isolates (PD-9 and PD-104) maintained viability of a significant number of conidia for an 557 extended period of time (Fig. 5C). Interestingly, GAG addition seemed to slightly increase the number 558 of grown CFUs for all isolates, although this increment was only significant for the PD-104 isolate (PD-559 104 VS PD-104+GAG p=0.0437). 560 These results suggest that a high level of GAG can potentiate persistence in some, but apparently not 561 all, isolates. Future investigations will aim to understand the underlying mechanism and to determine 562 why GAG addition is isolate-specific. to cover the whole well of broth dilution assays. The fungal material was stained with calcofluor white. 568 Representative wells of PD-104 with and without GAG at 2X the MIC are shown C) Counting of 569 germinated conidia/short hyphae in the entire wells of 2X (left) and 3X (right) MIC of broth dilution 570 assays. External addition of GAG significantly increased the number of germlings in PD-104 from 1.09% 571 to 2.80%, p=0.0267 at 2X MIC and from 0.41% to 0.75% p=0.038, two-tailed unpaired t-test. Two 572 independent experiments with three technical replicates were performed. The graphs represent the 573 means and SD. D) Plating of the entire content of wells containing 4 µg/mL of voriconazole shows that 574 external addition of GAG significantly increased the number of viable PD-104 cells that can be 575 recovered 48 hours after inoculation (p=0.0437, one way ANOVA with Tukey's multiple comparisons). 576 Two independent experiments with three biological replicates were performed. The graph represents 577 the means and SD. 578 579

Persistence can be detected in diverse A. fumigatus collections of isolates 580
To corroborate that certain A. fumigatus isolates display persistence to voriconazole, we decided to 581 screen two independent collection of isolates. Initially, we tested an environmental library of isolates 582 collected in the area of Manchester, UK [73]. We screened all 157 isolates for growth under the 583 microscope after 72 hours incubation in the presence of a high concentration of voriconazole (8 584 µg/mL). We found that 34 isolates were able to show a limited degree of growth, suggesting that they 585 might be persisters (Table S9 in https://doi.org/10.5281/zenodo.7021623). Next, we employed disc 586 diffusion assays with 0.8 mg/mL voriconazole and observed that 15 out of the 34 isolates were able to 587 form colonies in the halo (Table S9 in https://doi.org/10.5281/zenodo.7021623). Broth dilution assay 588 with the original isolates and the CoHs revealed that 10 of those isolates did not have increased MICs, 589 indicating that they were persister strains (Table S9 in https://doi.org/10.5281/zenodo.7021623). 590 Therefore, at least ~6% of this collection was persister to voriconazole. We further evaluated a 591 collection of 17 azole-susceptible clinical isolates obtained from TAU medical centre (Table S9 in  592 https://doi.org/10.5281/zenodo.7021623), which were prior shown not to have any mutation in the 593 cyp51A gene promoter or ORF, using the disc diffusion assay. We found that 6 out of 17 (35%) isolates 594 were able to grow small colonies in the inhibition halo (Fig. S9A). Colonies picked from within the halo 595 formed upon re-inoculation similar size halos, indicating that the MIC had not increased, and grew a 596 similar number of CoHs, suggesting that the isolates are persisters. 597 Therefore, evaluation of independent collection of isolates seem to always retrieve a significant 598 number of persister strains, demonstrating that this is quite a common phenomenon. 599

Persistence is not a lineage specific trait. 600
Recent studies suggest that the tandem repeats (TRs) in the promoter region of cyp51A, which cause 601 high levels of antifungal resistance, possibly have evolved in the environment, due to the use of 602 demethylase inhibitors (azoles) in the fields [74]. Interestingly, the isolates with the TR34/L98H 603 polymorphism have been found to be closely related [65,74], which suggest that this mechanisms has 604 evolved in a distinct lineage of the species. To investigate if the phenomenon of persistence could also 605 be lineage specific, we sequenced the genome of 23 of our isolates, 12 persisters and 11 non-persisters 606 (Table S9 in https://doi.org/10.5281/zenodo.7021623) and integrated their genome into a recently 607 published phylogenetic analysis [65]. We found that both persister and non-persister isolates 608 scattered through the entire tree, demonstrating that persistence is not a trait of a specific lineage of 609 A. fumigatus (Fig. 6). In addition, it is noteworthy that some persister and non-persister isolates seem 610 to be genetically very similar, as they are close in the tree (JA12-JA14-JA7-JA8 and also JA1-JA2, Fig.  611   6). This suggests that the genetic features that enable persistence are modest and/or that non-612 genomic features, as the transcriptional response, are important for isolate specificity. The genomes of newly sequenced persister and non-persister isolates, as characterised in this study 616 (Table S9 in https://doi.org/10.5281/zenodo.7021623), distributed scattered when included in the 617 previously generated phylogenetic tree [65]. Therefore, persistence is not a feature that has evolved 618 in a particular lineage. 619

In a Galleria mellonella infection model, a voriconazole treatment seems to be less efficient against 620 persister isolates in some larvae. 621
To evaluate if persistence may be relevant during antifungal treatment, we employed the Galleria 622 mellonella mini-host model of infection. This model has been successfully used to investigate the 623 efficiency of azole treatment against A. fumigatus [75,76]. In addition, a recent study has 624 characterised the pharmacokinetics of voriconazole in infected larvae, which allowed us selecting the 625 optimal dose to reach a high concentration of the drug in the haemolymph (above the MICs of the 626 isolates), but that is nearly completely removed in 24 hours [77]. 627 Initially, we performed a survival experiment with all four strains to investigate if they could have 628 different virulence potential. We infected larvae with 10 4 or 5×10 4 conidia of non-persister 629 (ATCC46645 or PD-60) or persister (PD-9 or PD-104) isolates and followed mortality for 10 days. All 630 the isolates killed larvae at a similar rate (Fig. S9B), demonstrating that they are similarly virulent. 631 Therefore, we aimed to use this model to investigate if persistence could potentially cause treatment 632 failure. We reasoned that a voriconazole treatment should very efficiently eradicate susceptible, non-633 persister strains, but it may not be able to eliminate persister strains with the same efficiency in all 634 individuals. We infected larvae with 10 4 (Fig. 7A) or 5×10 4 (Fig. 7B)  decreased fungal burden for all isolates ( Fig. 7A and B). To note, for voriconazole treated groups, the 639 mean of fungal burden was bigger for persisters than for non-persister isolates, although these 640 differences were not significant (ATCC VS PD-9 p=0.41 for 10 4 and p=0.57 for 5×10 4 , ATCC VS PD-104 641 p=0.27 for 10 4 and p=0.69 for 5×10 4 , Mann-Whitney test), due to the high variability in burden among 642 individuals infected with the persister isolates. Indeed, at both infectious doses several larvae infected 643 with persister isolates had noticeable greater burdens than all the other ( Fig. 7A and B), which we 644 speculate might indicate that in some individuals infected with persistent isolates treatment may be 645 less efficient in eliminating the fungus. infection with 5×10 4 conidia For all strains, fungal burden was greatly lower in larvae that had received 652 a voriconazole treatment (8 µg/mL) that in those who had not. The reduction in burden was bigger in 653 non-persister isolates than in persister isolates, although these differences were not significant. 654 Several treated larvae infected with persister isolates had noticeable higher burdens than the other: 655 A) 4/11 for PD-9 and 3/11 for PD-104 B) 3/11 for PD-9 and 4/11 for PD-104. Each graph displays the 656 combined data from two independent experiments. 657

DISCUSSION 659
The phenomenon of persistence to antimicrobials was first observed in bacteria 80 years ago [78,79]. 660 In the last decade, intensive research has permitted unravelling various underlying mechanisms in 661 diverse bacteria, and it has become clear that persistence can cause treatment failure and lead to the 662 development of resistance [80,81]. However, our insight into antifungal persistence and tolerance in 663 fungal pathogens is still in its infancy. 664 We have recently shown that all conidia in a tolerant A. fumigatus isolate are able to grow at high 665 concentrations of caspofungin [26]. In C. albicans, tolerance to fluconazole was described to be a sub-666 population effect, but the ratio of tolerant cells was reported to be elevated [21]. In both cases, 667 tolerance was observed in response to static drugs, so it seems that the phenomenon of tolerance in 668 fungi associates with fungistatic drugs. In C. albicans, the formation of amphotericin-B (AmB) persister 669 cells has been described [82]. Of note, this phenomenon has only been described within biofilms, and 670 cannot be found in planktonic cells [83], possibly due to the physical protection conferred by the 671 biofilm, which may be relevant for the capacity of these cells to survive in the presence of AmB. 672 Indeed, in A. fumigatus it was shown that the hypoxic microenvironment developed in biofilms led to 673 reduced metabolic activity in the basal biofilm level leading to the formation of cells that can survive 674 the antifungal challenge, and serve as a drug-resistant reservoir [84]. 675 We have observed that a small sub-population (0.1 to 5%) of certain A. fumigatus isolates can survive 676 for extended periods and even grow at slow rates in the presence of supra-MIC concentrations of the 677 fungicidal drug voriconazole. In bacteria, tolerance and persistence have been classically explained by 678 a downregulation of metabolism and cell cycle, which trigger a status of dormancy that permit survival 679 despite the action of the drug. However, recent research in pathogenic bacteria has demonstrated 680 that active metabolic responses are required for persistence [85,86], and slow-growing persisters 681 have also been detected in vitro [87] and in vivo [88]. Active metabolism may even be advantageous, 682 as in the case of Salmonella, where persisters have been shown to undermine host defences [89]. 683 Interestingly, C. albicans AmB persistence has been described to involve downregulation of primary 684 metabolism, but also an increase of stress responses and oxidative defensive mechanisms [90]. We 685 have observed that active growth is possible at two or even three-fold the MIC, and that this growth 686 entails a distinct transcriptional profile. Therefore, this phenomenon is not merely survival of dormant 687 conidia in the presence of the drug, but seems to be an active mechanism that enables a sub-688 population of certain isolates to withstand the action of voriconazole for an extended period. Indeed, 689 using resazurin (an oxidation-reduction indicator used for the measurement of metabolic activity and 690 proliferation of living cells, which use has been optimised for A. fumigatus [91]) we could detect a 691 slight metabolic activity at supra-MIC concentrations (Fig. S9C). This might explain why hypoxia is the 692 only environmental condition that reduced persistence, as in low oxygen the metabolic activity is 693 reduced [84] and the energy generating metabolism changes [92]. 694 We have observed a big difference in growth between liquid media (only microscopic growth) and 695 solid media (macroscopic colonies). We hypothesise that this may be because in agar drug 696 microdepletions around the hyphae would diffuse back much slower than in liquid, giving an 697 advantage to the hyphae in solid media. In addition, there may be a different level of contact of 698 conidia with the drug; in liquid, the conidia are completely surrounded, whilst on solid media the 699 conidia just lay on the drug and may escape from it by orientating the germling polarity as they grow. 700 It is surprising that we have not found any voriconazole tolerant isolate. It might be that the frequency 701 of this phenomenon is lower in A. fumigatus (even in filamentous fungi in general), and we simply 702 have not screened enough isolates to detect it. However, we believe that the particularities of working 703 with filamentous fungi make the definitions developed with bacteria unsuitable. Firstly, MIC 704 measurements or killing curves need to be done over a period of days, compared with hours for 705 bacteria. In addition, the inoculum used is normally conidia, a dormant cell structure that needs to be 706 activated to produce an effect. Besides, the experiments are performed with inocula of around 10 4 707 (2.5×10 4 for broth dilution and 4×10 4 for disc assays) conidia, whilst bacteria persistence can be 708 assayed with inocula in the range of 10 6 -10 8 cells. Therefore, we believe it would be challenging to 709 clearly discern between the duration of killing for 99% or 99.99% of the cells, the most important 710 factor to differentiate tolerance and persistence [18,19]. Indeed, looking closely at the killing curve 711 presented in Fig. 1D, at 2 days the percentage of survivors is 99.38% for PD-9 and 99.86% for  Could this mean that PD-9 is tolerant and PD-104 a persister? This seems to be supported by the 713 experiments plating wells from the broth dilution assays (Fig. 1C and 5C), where more colonies are 714 counted for PD-9 than for PD-104. However, the number of colonies detected in the disc assays is 715 usually higher for PD-104 (2-5) than for PD-9 (1-2). Given that the mechanisms of tolerance and 716 persistence in bacteria are multiple and overlapping [93], we propose to define the phenomenon that 717 we have described in this study as persistence and utilize the term tolerance to define the effect 718 previously described with fungistatic drugs [21,26]. Therefore, we propose to adapt the current 719 definitions for pathogenic fungi as presented in Table 1. 720 The concept of persistence entails two intriguing aspects. Firstly, it is an isolate-dependent 721 phenomenon, which means that there must be a genetic basis that underlies persistence. We have 722 shown that the persister isolates do not belong to a specific lineage, suggesting that this feature has 723 not appeared as an evolutionary trait in a lineage of isolates. Recently, it has been described that each 724 A. fumigatus strain carries a particular set of accessory genes, which associates with different levels 725 of virulence and drug resistance capacities [94]. In the same vein, we speculate that the presence or 726 absence of certain accessory genes may enable certain isolates to persist in the presence of azoles. 727 Future research will evaluate if this hypothesis is true. A second intriguing aspect of persistence is that 728 it is a sub-population phenomenon, meaning that within an isogenic isolate only a few conidia are able 729 to survive and grow in the presence of the drug. In bacteria, stochastic expression of key genes has 730 often been proposed as an important underlying cause of persistence [68,69,95,96]. This includes 731 stochastic high expression of genes that directly confer resistance [97] and efflux activity [98]. 732 Similarly, in our RNA-seq analysis we have detected a higher level of expression of genes of the sterol 733 biosynthesis pathway, including cyp51A, and of the azole exporter Cdr1B in persister cells. Therefore, 734 it is plausible that stochastic high levels of these genes could generate the capacity to survive for an 735 extended period in the presence of supra-MIC concentrations of voriconazole. Additionally, we 736 detected a higher level of expression of GAG biosynthetic genes, and we have observed that externally 737 added GAG can increase the number of persisters in the isolate PD-104. Interestingly, this 738 polysaccharide did not increase the isolate's MIC, indicating that it does not provide a physical barrier 739 that shields the fungus or that it cannot somehow degrade the drug. Future investigations will aim to 740 unravel why GAG potentiates persistence in PD-104, and not in PD-9, which is interesting as it suggests 741 that there might be multiple persistence mechanisms. GAG is a very relevant A. fumigatus virulence 742 factor, with multiple described adhesion and immunosuppressive activities [99,100], and here we 743 propose that it may have one more important role in potentiating persistence to voriconazole. 744 In bacterial infections, there is significant evidence supporting a role for the phenomena of tolerance 745 and persistence in antibiotic treatment failure [16,101]. In contrast, the knowledge about these 746 phenomena in fungal infections is still very scarce. Probably, the best-studied phenomenon is 747 heteroresistance in Cryptococcus neoformans and C. gatii [102][103][104], which has been found to be a 748 major cause of treatment failure in cryptococcal meningitis [105-109]. As mentioned above, there is 749 already evidence indicating that fluconazole tolerance can cause treatment failure in invasive 750 candidiasis [17,21], and limited evidence suggests that caspofungin tolerance may cause treatment 751 failure in Aspergillus infections [27]. Here, using a Galleria model of infection, we have observed that 752 voriconazole treatment seems to eliminate persister isolates less efficiently than non-persister isolates 753 in various larvae. Therefore, we hypothesise that in certain individuals persistence might reduce the 754 efficacy of voriconazole treatment in A. fumigatus infections, which, if proven true, could imply that 755 persistence may cause therapeutic failure in some patients. In addition to such a potential direct effect 756 of persistence on treatment failure, resilience and relapse of persistent strains could associate with 757 the development of antifungal resistance (an effect that has already been shown for bacterial 758 infections, see [110] for references), as it is known that prolonged azole treatment can result in 759 resistance [111,112]. We appreciate that our results provide limited evidence to support the 760 hypothesis that voriconazole persistence may cause treatment failure, and acknowledge that much 761 more research is required to reach such conclusion. We will continue investigating this hypothesis and 762 we exhort the fungal research community to consider and investigate this phenomenon to clarify this 763 important matter. 764

MATERIAL and METHODS 766
A. fumigatus strains and culture conditions. 767 All isolates utilized in the course of this study are listed in Tables S1 and S8. Briefly, information about 768 the common laboratory strains used can be found in [113], the isolates from Paul Dyer collection are 769 described in [114] and the collection of isolates from the area of Manchester are described in [73]. 770 The third collection of isolates was collected at the TAU medical centre (Tel-Aviv, Israel), these strains 771 were characterized for their antifungal resistance profile and cyp51A genotype. Persister isolates are 772 labelled in all figures with an asterisk (*). 773 Isolates were routinely grown on Potato Dextrose Agar (PDA, Oxoid) for 72 hours to obtain fresh 774 spores for each experiment. All experiments were performed with fresh spores except in the test for 775 persistence with old conidia. Aspergillus Minimal Medium (AMM) was prepared following a standard 776 recipe [115]. Sabouraud (Oxoid) and yeast extract glucose (YAG, 2%, 0.5% yeast extract, 1.7% agar, 1X 777 trace elements) media were used in specific experiments. 778 To evaluate persister growth and determine the MIC, isolates were grown on RPMI-1640 (Sigma) with 779 35 g/L MOPS (Alfa Aesar) and 2% glucose, pH 7. 780 Galactosaminogalactan (GAG) was obtained from the A. fumigatus ∆ku80 strain, grown 2-day in 1.5L 781 brian fermenter at room temperature. GAG isolation and purification was carried out as previously 782 described [116]. Briefly, the medium supernatant was collected by filtration and was adjusted to pH 3 783 by addition of 100 μl 12 M HCl per 100 ml supernatant. Two volumes of precooled ethanol (4°C) were 784 added and GAG was precipitated for 3 h at 4°C. The precipitate was collected by centrifugation for 20 785 min at 5,000 g at 4°C and subsequently washed twice with 1/10 of the culture volume of 200 mM NaCl 786 for 1 h under agitation (100 rpm). GAG was dialyzed against tap water and twice against purified water 787 (24 h each) and finally lyophilized to dryness and stored at ambient temperature. 788 To calculate germination rate, 10 4 conidia of each strain were inoculated in 200 µL liquid RPMI in 96-789 well plates and imaged a Nikon TI microscope equipped with a 37 °C incubator and a 40× objective, 790 with 1 picture captured every 30 min by using NIS-Elements 4.0 (Nikon) software. Cell Counter plug-791 in of Image J platform (http://rsb.info.nih.gov/ij/index.html) was employed to differentially count 792 resting or swollen conidia versus germinated conidia. 793 To calculate growth rate on solid medium, 10 3 conidia of each strain were inoculated on RPMI plates, 794 and the colony diameter was measured in two different angles per colony at 16, 24, 40 and 48 h. To 795 calculate growth rate on liquid medium 2×10 3 of each strain were inoculated in 200 µL liquid RPMI in 796 96-well plates and optical density measurements were taken at 600 nm every 10 minutes. Growth 797 curves were analysed using the R package Bayesian Estimation of Change-points in the Slope of 798

Evaluation of persistence and determination of MICs. 802
To determine persistence using the disc assay, 4×10 4 conidia of each isolate was evenly spread on a 803 solidified RPMI plate (1.5% agar) and 10 μl of 0.8 mg/mL voriconazole or 3.2 mg/mL itraconazole 804 (Acros Organics) was added to a Whatman 6 mm antibiotic assay disc which was placed in the middle 805 of the plate. For in vitro treatment with adjuvant and combinatorial drugs, the agents were added to 806 the RPMI medium at the final concentration detailed in each section. Plates were incubated for 5 days 807 at 37 °C. Persister colonies were defined as those with a degree of physical separation from the edge 808 of the loan of growth into the halo of inhibition. 48 hours at 37 °C with occasional shaking. The strains were then centrifuged for 4000 RPM for 5 819 minutes, the media discarded and the conidia resuspended in 2 mL filtered PBS. This wash was 820 repeated three times in total before the conidia were resuspended in 600 µL of drug free RPMI. 300 821 µL of each suspension, approximately 7.5 ×10 3 conidia, was added into wells of a 96 well plate. The 822 wells were then imaged for 72 hours, every 2 hours, on a Nikon Eclipse Ti microscope, using a Nikon 823 CFI Plan Fluor ELWD 20x/0.45na objective, and captured with a Hamamatsu ORCA-FLASH4.0 LT+ 824 camera (Hamamatsu Photonics) and manipulated using NIS-Elements AR 5.11.01 (Nikon). A video was 825 prepared using Fiji [119]. 826 Microscopy to observe growth at supra-MIC conditions was done in a THUNDER Imager Live Cell 827 microscope, with a HC PL FLUOTAR L 20x/0.40 DRY objective. Images were captured using a Leica-828 DFC9000GTC-VSC13067 camera and the Las X (Leica application suite) v 3.7 software. 829 To construct the killing curved, the isolates were inoculated in 10 mL of RPMI containing 4 µg/mL 830 voriconazole. Aliquots for each culture were taken at the time of inoculation (100 µL) and every 24 831 hours (1 mL). The aliquots were spun at 16,000 ×g for 5 minutes, resupended in 1 mL PBS and vortexed. 832 This was repeated twice to wash off the drug. Finally, conidia were resuspended in 1 mL PBS and a 833 fraction plated on PDA plates (50 µL at time 0h, 100 µL at 24h and 1 mL for 48, 72 and 96 h). The 834 experiment was repeated three times. 835 To calculate the percentage of germinating conidia, the wells of broth dilution assays containing 2X, 836 3X MIC and the maximum drug concentration (8 µg/mL) were stained (after reading the MIC) with 10 837 µg/mL of Calcofluor White (Sigma) for 5 minutes. The entire wells were imaged with a THUNDER 838 Imager Live Cell microscope, using a HC PL FLUOTAR L 20x/0.40 DRY objective, filter conditions: 839 EX:375-435 DCC:455 EM:450-490, a Leica-DFC9000GTC-VSC13067 camera and the Las X (Leica 840 application suite) v 3.7 software. Merged images were analysed using FIJI [119]. Briefly, the merged 841 images were converted to 8-bit, a threshold was set for the images so that the conidia and germlings 842 could be detected over the background (20 to 255). The option analyse particle was then executed, 843 setting a minimum size of 0.04 inches 2 (which was found to exclude resting conidia). control group was subjected to the same treatment, but without fungal infection. 869 To detect the fungal burden, 500-ng portions of DNA extracted from each larva were subjected to 870 qPCR using the SensiMic SybR Green kit (Bioline). Forward (5′-ACTTCCGCAATGGACGTTAC-3′) and 871 reverse (5′-GGATGTTGTTGGGAATCCAC-3') primers were used to amplify the A. fumigatus β-tubulin 872 gene (AFUA_7G00250). The primers designed to amplify the Elongation factor 1-Alpha (Ef-1a) were as 873 follows: forward (5′-AACCTCCTTACAGTGAATCC-3′) and reverse (5′-ATGTTATCTCCGTGCCAG-3′). 874 Standard curves were calculated using different concentrations of fungal and larval gDNA pure 875 template. Negative controls containing no template DNA were subjected to the same procedure to 876 exclude or detect any possible contamination. Three technical replicates were prepared for each 877 sample. qPCRs were performed using a CFX96 Real-Time System (Bio-rad) with the following thermal 878 cycling parameters: 95°C for 10 min and 40 cycles of 95°C for 15 s and 58°C for 15 s and 72°C for 15 s. 879 The fungal burden was calculated by normalizing the number of fungal genome equivalents (i.e., the 880 number of copies of the tubulin gene) to the larval genome equivalents in the sample (i.e., the number 881 of copies of the a Ef-1a gene), as we have reported before [120]. 882

RNA sequencing (RNA-seq) 883
Total RNA was submitted to the Genomic Technologies Core Facility (GTCF) at the University of 884 Manchester, UK. Quality and integrity of the RNA samples were assessed using a 4200 TapeStation 885 (Agilent Technologies) and then libraries generated using the Illumina® Stranded mRNA Prep. Ligation 886 kit (Illumina, Inc.) according to the manufacturer's protocol. Briefly, total RNA (typically 0.025-1ug) 887 was used as input material from which polyadenylated mRNA was purified using poly-T, oligo-888 attached, magnetic beads. Next, the mRNA was fragmented under elevated temperature and then 889 reverse transcribed into first strand cDNA using random hexamer primers and in the presence of 890 Actinomycin D (thus improving strand specificity whilst mitigating spurious DNA-dependent 891 synthesis). Following removal of the template RNA, second strand cDNA was then synthesized to yield 892 blunt-ended, double-stranded cDNA fragments. Strand specificity was maintained by the 893 incorporation of deoxyuridine triphosphate (dUTP) in place of dTTP to quench the second strand 894 during subsequent amplification. Following a single adenine (A) base addition, adapters with a 895 corresponding, complementary thymine (T) overhang were ligated to the cDNA fragments. Pre-index 896 anchors were then ligated to the ends of the double-stranded cDNA fragments to prepare them for 897 dual indexing. A subsequent PCR amplification step was then used to add the index adapter sequences 898 to create the final cDNA library. The adapter indices enabled the multiplexing of the libraries, which 899 were pooled prior to cluster generation using a cBot instrument. The loaded flow-cell was then paired-900 The reads were trimmed to remove any adapter or poor quality sequence using Trimmomatic v0.39 909 [121]; reads were truncated at a sliding 4bp window, starting 5', with a mean quality <Q20, and 910 removed if the final length was less than 35bp. Additional flags included: 'ILLUMINACLIP:./Truseq3-911 PE-2_Nextera-PE.fa:2:30:10 SLIDINGWINDOW:4:20 MINLEN:35'. 912 The filtered reads were mapped either to the Aspergillus fumigatus A1163 reference sequence 913 (GCA_000150145.1/ASM15014v1) downloaded from the Ensembl Genomes Fungi v44 [122] or PD-914 104 a de novo assembled genome, using STAR v5.3a [123]. For A1163 reference the genome index 915 was created using the GTF gene annotation also from Ensembl Genomes Fungi v44. For the PD-104 916 reference GTF gene annotation was generated as described below. A suitable flag for the read length 917 (--sjdbOverhang 75) was used. '--quantMode GeneCounts' was used to generate read counts in 918 genes. 919 Subsequently, PD-104 reads were aligned to the recently generated pan-genome [65] using Salmon 920 v1.6.0 (additional parameters for salmon quant: --gcBias (corrects for any GC biases in samples) -l ISR 921 (inward / stranded / reverse read pairs)). 922 Normalisation and differential expression analysis was performed using DESeq2  We are deeply grateful to Prof Paul Dyer, who kindly gifted to us the collection of isolates that have 990 been used in this study. We would like to thank Diego Megías and Clara Martín at ISCIII for their help 991 with the microscopy. We are also grateful to Michael Bottery for relevant conversations about 992 bacterial tolerance and persistence. We acknowledge the use of the Genomic Technologies Facility 993 (Faculty of Biology Medicine and Health, University of Manchester) for the DNA and RNA sequencing. 994 Help and support from members of MFIG and LRIM is greatly appreciated. 995 JA is funded by an Atracción de Talento Modalidad 1 (020-T1/BMD-200) contract of the Madrid 996 Regional Government (CAM). JS has been funded by a BSAC Scholarship (bsac-2016-0049). CV was 997 funded by FAPESP (2108/00715-3 and 2020/01131-5). GHG has been funded by FAPESP (2016/07870-998 9 and 2021/04977-5), CNPq (301058/2019-9 and 404735/2018-5) and by the NIH/NIAID (grant no. 999 R01AI153356 all comparisons, a tight node of 6 interacting proteins, related with ergosterol production, was 1429 detected by STRING (p<1.0e-16). 1430

Fig. S9. 1431
A) Representative disc diffusion plates carried out with the collection of isolates from the TAU medical 1432 centre. Six isolates were able to form CoHs, which did upon reinoculation form halos of the same size 1433 (MIC did not increase) and create a similar number of CoHs. B) Survival curves of larvae infected with 1434 10 4 or 5×10 4 conidia of the isolates ATCC, PD-60 (non-persisters), PD-9 or PD-104 (persisters). All 1435 strains killed larvae at a similar rate, demonstrating that they are equally virulent. For 10 4 , three 1436 independent experiments were done with 15-20 larvae/isolate in each. For 5×10 4 , two independent 1437 experiments were done with 10-15 larvae/isolate in each. C) Resazurin was added to the broth dilution 1438 RPMI plate at a 0.002% (w/v) final concentration. The plate was incubated for 24 hours in a Tecan 1439 Infinity M-Plex plate reader and fluorescence was measured every 30 minutes with an excitation 1440 wavelength of 544 nm and reading emission at 590 nm using the i-control 2.0 software. The read for 1441 each well at each time point was normalize as follows: (read well time X-read 8 µg/mL voriconazole 1442 time X)/read well at time 0. The time point with the best dynamic range of values in the 1X, 2X and 3X 1443 MIC was found to be 20 hours, which was selected to determine the metabolic activity. A value of 0.1 1444 was assigned as background and values above 1 were detected in sub-MIC (macroscopic growth). At 1445 1X MIC, a slight metabolic activity could be detected for both the non-persister ATCC and the persister 1446 isolates PD-9 and PD-104. At 2X and 3X the MIC, slight metabolic activity could only be detected for 1447 the persister isolates. The experiment was performed once with two biological replicates, the graph 1448 represent the means and SEM. 1449 Supplementary Files (in https://doi.org/10.5281/zenodo.7021623) 1450 Table S1: Collection of isolates used in this study. 1451 Table S2: Synonymous polymorphisms detected in the cyp51B gene of the isolates ATCC46645, PD-9, 1452 PD-60 and PD-104. 1453 Table S3: GO enrichment of genes regulated exclusively in Persistence (A1163 genome as reference). 1454 Table S4: GO enrichment of genes regulated in Persiter VS Low Drug (Pangenome as reference). 1455 Table S5: GO enrichment of genes regulated exclusively in Persistence (Pangenome as reference). 1456 Table S6: Proteins included in the STRING network (genes upregulated only in Persistence). 1457 Table S7: GO enrichment of genes regulated more in Persistence (Pangenome as reference). 1458 Table S8: Proteins included in the STRING network (genes upregulated more in Persistence). 1459 Table S9: List of environmental isolates in the UK collection. Identity of the sequenced genomes. List 1460 of the clinical isolates of the TAU collection. 1461 Table S10: Example of the Quast obtained to determine the quality of the assembly, using the PD-1462 104 genome. 1463 DataSet1: RNA-seq results and analysis of the A1160 isolate grown in NoDrug and Low Drug. 1464 DataSet2: RNA-seq results and analysis of the PD-104 isolate grown in NoDrug, Low Drug and 1465 Persister, using the A1163 genome as reference for the alignment of reads. 1466 DataSet3: RNA-seq results and analysis of the PD-104 isolate grown in NoDrug, Low Drug and 1467 Persister, using the pangenome genome as reference for the alignment of reads. 1468 Video S1: The persister isolate PD-9 pre-incubated in voriconazole resumes growth when the drug is 1469 withdrawn. 1470

Video S2
The persister isolate PD-104 pre-incubated in voriconazole resumes growth when the drug 1471 is withdrawn. 1472

Video S3
The non-persister isolate ATCC46645 pre-incubated in voriconazole resumes growth when 1473 the drug is withdrawn. 1474