Candida auris phenotypic heterogeneity determines pathogenicity in vitro

Candida auris is an enigmatic yeast that provides substantial global risk in healthcare facilities and intensive care units. A unique phenotype exhibited by certain isolates of C. auris is their ability to form small clusters of cells known as aggregates, which have been to a limited extent described in the context of pathogenic traits. In this study, we screened several non-aggregative and aggregative C. auris isolates for biofilm formation, where we observed a level of heterogeneity amongst the different phenotypes. Next, we utilised an RNA-sequencing approach to investigate the transcriptional responses during biofilm formation of a non-aggregative and aggregative isolate of the initial pool. Observations from these analyses indicate unique transcriptional profiles in the two isolates, with several genes identified relating to proteins involved in adhesion and invasion of the host in other fungal species. From these findings we investigated for the first time the fungal recognition and inflammatory responses of a three-dimensional skin epithelial model to these isolates. In these models, a wound was induced to mimic a portal of entry for C. auris. We show both phenotypes elicited minimal response in the model minus induction of the wound, yet in the wounded tissue both phenotypes induced a greater response, with the aggregative isolate more pro-inflammatory. This capacity of aggregative C. auris biofilms to generate such responses in the wounded skin highlights how this opportunistic yeast is a high risk within the intensive care environment where susceptible patients have multiple indwelling lines. Importance Candida auris has recently emerged as an important cause of concern within healthcare environments due to its ability to persist and tolerate commonly used antiseptics and disinfectants, particularly when surface attached (biofilms). This yeast is able to colonise and subsequently infect patients, particularly those that are critically ill or immunosuppressed, which may result in death. We have undertaken analysis on two different types of this yeast, using molecular and immunological tools to determine whether either of these has a greater ability to cause serious infections. We describe that both isolates exhibit largely different transcriptional profiles during biofilm development. Finally, we show that the inability to form small aggregates (or clusters) of cells has an adverse effect on the organisms immuno-stimulatory properties, suggestive the non-aggregative phenotype may exhibit a certain level of immune evasion.

with the ability to persist in the environment before transmission to humans [5]. 7 A unique pathogenic trait exhibited by some isolates of C. auris is their ability to form 8 aggregates (Agg) [6][7][8]. Despite the well-documented prevalence of C. auris 9 worldwide, relatively little is known about the Agg phenotype of the organism. The 10 existence of four geographically and phylogenetically distinct clades of the organism 11 [2], and a fifth recently proposed [9], has restricted a definitive profiling of C. auris 12 pathogenic mechanism of these aggregates in regard to biofilm forming capabilities, 13 drug resistance pathways and interactions with the host. Of the publications that 14 exist, these have documented characteristic pathogenic traits for both phenotypes in 15 vitro and in vivo [6,10,11]. Others have shown that the Agg phenotype is inducible 16 under certain conditions [7,8], whilst histological analyses of murine models have 17 shown that aggregates can accumulate in organs following C. auris infection [7,12,18 13]. Therefore, further studies are required to investigate this characteristic Agg 19 phenomenon in C. auris isolates to fully comprehend the pathogenic pathways of the 20 organism, and to understand how such mechanisms may differ from their non-Agg 21 counterparts.

22
Limited evidence also exists for studies investigating the interactions of C. auris with 23 components of the host, although several in vivo models have been employed to 24 document the virulence of C. auris. Of these, Galleria mellonella larvae infection 1 models and murine models of invasive candidiasis have shown varying survival rates 2 post-infection with C. auris [6,8,10,[12][13][14][15], reaffirming that genetic variability 3 amongst clades impacts the organisms virulence. However, such studies have been to the non-Agg and/or Agg phenotype. 16 In this study, we sought to investigate the level of heterogeneity amongst different 17 non-Agg and Agg isolates. We deemed this pertinent given that such traits of 18 heterogeneity amongst isolates have previously been described for other Candida 19 species, significantly impacting clinical outcomes and mortality rates [18]. To further 20 investigate this Agg versus non-Agg phenotype, transcriptome analyses were 21 performed on planktonic cells and biofilms of two selected isolates from the initial 22 pool. Upon completion of these analyses, we discovered that several genes 23 associated with cell membrane and/or cell wall proteins (e.g. cellular components) 24 were upregulated in the Agg biofilm. Such unique transcriptional profiles in respect to 25 6 the cellular components led us to investigate the host response following stimulation 1 with both C. auris phenotypes in vitro. For this, a two-and three-dimensional skin 2 wound model was employed to investigate the epithelial response to the Agg and 3 non-Agg isolates of C. auris. Both skin wound models exhibited different profiles to 4 both isolates indicating unique fungal recognition and/or host response to the Agg 5 and non-Agg phenotype. Interestingly, there was minimal response by the host to C. 6 auris without induction of the wound, suggestive the organism relies on loss of tissue 7 integrity to become invasive. For in vitro biofilm biomass assessment, a pool of aggregating (Agg; n=12) and 12 single-celled, non-aggregative (non-Agg; n=14) C. auris clinical isolates (gifted by Dr 13 Andrew Borman and Dr Elizabeth Johnson, Public Health England, UK). All C. auris 14 isolates were stored in Microbank TM beads (Pro-lab Diagnostics, UK) prior to use.

15
Each isolate was grown on Sabouraud dextrose (SAB) agar (Oxoid, UK) at 30˚C for 16 24-48 h then stored at 4˚C prior to propagation in yeast peptone dextrose (YPD;

18
Cells were pelleted by centrifugation (3,000 x g) then washed two times in phosphate 19 buffered saline (PBS). Cells were then standardised to desired concentration 20 following counting using a haemocytometer, then resuspended in selected media for 21 each assay, as described within.

22
C. auris isolate phenotypes were determined visually by suspending one colony in 1 23 mL (PBS, Sigma-Aldrich, UK). Isolates were termed as 'aggregators' if the added 24 7 colony did not disperse upon mixing in PBS. For RNA sequencing and transcriptional 1 analysis of C. auris biofilms and co-culture systems, one Agg (NCPF 8978) and non-2 Agg (NCPF 8973) isolate was used.

4
Biofilm growth and biomass assessment 5 Fungal cells were adjusted to 1 x 10 6 cells/mL in Roswell Parks Memorial Institute 6 (RPMI) media (Sigma-Aldrich, UK) and biofilms formed for 4 or 24 hours at 37˚C in 7 flat-bottom wells of 96-well plates (Corning, UK). Appropriate media controls were 8 included on each plate to test for contamination. Following incubation, biofilms were 9 washed gently once in PBS to remove any non-adhered cells. The biomass of each 10 biofilm was determined via 0.05% crystal violet (CV) staining as described previously 11 [19]. Absorbance of the CV stain was measured spectrophotometrically at 570nm in 12 a microtiter plate reader (FLUOStar Omega, BMG Labtech, UK).

14
Monitoring the growth of Candida auris biofilms in real time 15 The xCELLigence real time cell analyser (RTCA, ACEA Bioscience Inc, San Diego, 16 CA) was used to monitor the formation of C. auris biofilms in real time using electron 17 impedance measurements (presented as cell index, CI) which is directly related to 18 cell attachment and proliferation. In brief, the E-plate containing 100 µL of pre-heated 19 RPMI medium was loaded into the RTCA which had been placed in the incubator 2 h 20 prior to the experiment to test media impedance and electrode connectivity. Cultures 21 of each C. auris isolate used in this study were standardised to 2 x 10 6 CFU/mL and 22 added to the E-plate in 100 µL aliquots in triplicate. Appropriate media controls 23 minus inoculum were also included in triplicate. Biofilm formation was measured over  filter at the air-liquid interface, in a chemically defined medium grown to 17-day 24 maturity. This model is histologically similar to in vivo human epidermis. Upon arrival 1 and prior to experimental set-up, RHE was incubated with maintenance media in 24-2 well plates (Corning, UK) for 24 h, 5% CO 2 at 37 o C. Maintenance media was 3 replaced then the co-culture three-dimensional system was set up as described 4 below.

6
Wound model in two-dimensional and three-dimensional co-culture systems 7 HEKa cell monolayers were scratched using a similar method to as previously 8 described to mimic a wound model [22,23]. Briefly, monolayers were grown to 9 confluence as described above then three parallel scratches were introduced across 10 the surface using a 100 μL pipette tip prior to inoculation with C. auris. For RHE, a 11 19-gauge needle was used to scratch the tissue. For all co-culture experiments, Agg 12 C. auris NCPF 8978 and non-Agg C. auris NCPF 8973 were grown as described 13 above, then standardised to 2 x 10 6 /mL (multiplicity of infection of 10 to HEKa cells;

14
MOI 10, and as previously described for Candida-tissue co-culture [24]). For the two-15 dimensional system, 2 x 10 6 /mL C. auris was prepared in 500µL supplemented 16 media 154 and added directly to the confluent HEKa cells. For the three-dimensional 17 co-culture model, 2 x 10 6 /mL C. auris was prepared in 100µL of sterile PBS and this 18 suspension was added directly to the RHE tissue. All experiments were conducted 19 for 24 hr at 5% CO 2 . Infected non-scratched HEKa cells and RHE tissue were used 20 as controls e.g., no wound, and uninoculated co-cultured cells or tissues were also 21 included for all experiments. All control and wounded models were infected in 22 triplicate with both isolates of C. auris. 23 24 Histological staining of skin epidermis 1 Following co-culture with C. auris, epithelial tissue was carefully cut from the 0.5 cm 2 2 insert using a 19-gauge needle and washed three times in sterile PBS to remove 3 non-adherent cells in a similar manner as previously described [25], as summarized 4 in the schematic in Figure 1. Tissue was then fixed in 10% neutral-buffered formalin 5 prior to embedding in paraffin. A Finnesse ME+ microtome (Thermo Scientific, UK) 6 was used to cut 2 μm sections and tissue sections stained with haematoxylin and 7 eosin, or with the fungal specific Periodic acid-Schiff (PAS) reagents and counter-8 stained with haematoxylin. To assess any cytotoxic effects of C. auris on HEKa cells and RHE tissue, a Pierce 12 lactate dehydrogenase (LDH) Cytotoxicity Assay Kit (Thermo Scientific; UK) was 13 used according to the manufacturers' instructions. Following co-culture, cell or tissue 14 spent media was assayed using the above kit to quantify the level of LDH release as 15 a measure of host cellular disruption. Green ER based-quantitative polymerase chain reaction (qPCR) or RT 2 profiler arrays 3 (Qiagen Ltd, UK). For SYBR Green ER based-qPCR analyses, the following PCR 4 thermal profiles was used; holding stage at 50˚C for 2 minutes, followed by 5 denaturation stage at 95˚C for 10 minutes and then 40 cycles of 95˚C for 3 seconds 6 and 60˚C for 15 seconds. qPCR plates were run on the StepOnePlus™ Real-Time 7 PCR System. The following primer sequences were used for SYBR Green ER based-8 qPCR analyses of host cells; GAPDH, forward primer, 5' to 3', For RNA sequencing of C. auris biofilms, RNA was extracted from 24-h C. auris 5 biofilms as described previously [29]. In brief, biofilms were grown as above on 6 Thermanox coverslips (Thermo-Fisher, UK) in 24 well plates (Corning, UK). Biofilms 7 were removed from coverslips by sonication at 35 kHz for 10 minutes in a sonic bath 8 in 1 ml of PBS and the sonicate transferred to a 2.0 ml RNase-free bead beating 9 tube (Sigma-Aldrich, UK). Cells were homogenized in TRIzol ™ (Invitrogen, UK) with 10 0.5 mm glass beads using a BeadBug microtube homogeniser for a total of 90 11 seconds (Benchmark-Scientific, USA). RNA was then extracted as described above 12 using the RNeasy Mini Kit according to manufacturers' instructions (Qiagen Ltd, UK).

13
Following extraction, RNA quality and quantity were determined using a Bioanalyzer 14 (Agilent, USA), where a minimum RNA integrity number and quantity of 7 and 2.5 µg, 15 respectively were obtained for each sample. Annotation of data following sample 16 submission to Edinburgh Genomics (http://genomics.ed.ac.uk/) was completed as 17 previously described [30]. Briefly, raw fastq reads were trimmed and aligned to the

13
The Agg phenotype is a unique trait of C. auris, one that can influences the To further study the pathogenic and biofilm-forming characteristics of Agg and non-

10
Agg C. auris, transcriptional profiling of 24-hour planktonic vs biofilm phenotypes 11 was performed. For these studies, two clinical isolates from the initial pool tested 12 were selected for analysis (non-Agg NCPF 8973 and Agg NCPF 8978, indicated by 13 the red points in Figure 2). Firstly, we found that a total of 701 genes were 14 upregulated in planktonic and/or biofilm form of Agg compared to the non-Agg 15 phenotype, of which 450 genes were upregulated in the biofilm state ( Figure 3A).

16
Conversely, less genes (430) Figure 2C). Only a small number of genes belonging to cellular 8 components were upregulated in the non-Agg compared to Agg biofilm ( Figure 3D).

9
The observed differences in the transcriptional profiles of genes belonging to the CC Agg C. auris was significantly more cytotoxic than non-Agg form in the wound model 23 ( Figure 4A; § § p < 0.01). It is noteworthy that wounded monolayers minus inoculum 24 were comparable to untreated monolayers (data not shown), suggestive that 25 induction of the wound did not induce cytotoxic effects on the cells. A similar trend in 1 cytotoxicity was observed between the two isolates in the three-dimensional model 2 (Episkin, SkinEthic TM reconstructed human epidermis; RHE) although this did not 3 reach statistical significance ( Figure 4B). Interestingly, cytotoxicity of Agg and non-4 Agg C. auris were comparable in both co-culture models minus wounds.

5
To further study the host response in the three-dimensional system, a transcriptional 6 response was investigated using a RT 2 profiler array containing primers specific for Agg C. auris isolate in the three-dimensional tissue model ( Figure 4D and 4E). Such 8 a response in the Agg isolate may begin to elucidate the mechanism by which this 9 phenotype generated a greater inflammatory response within the tissue. antifungals, suggestive that treatment regimens must be carefully considered to 20 combat C. auris dependant on its Agg phenotype [8]. In this study, we report a 21 previously described RNA-sequencing approach [30] in order to compare the study, the role of ALS1, which encodes for the protein Als1p, in Candida-host 23 interactions remains unclear. As such, contradictory reports state a role for this 24 adhesion in attachment to epithelial cells. Kamai and colleagues (2002)  It is apparent from these studies that the host recognizes yet fails to generate an 3 effective immune response against C. auris. From the results described here, it is 4 evident that C. auris is not cytotoxic nor pro-inflammatory to intact skin epithelial cells 5 or epidermis tissue. Only following induction of a wound in these models did C. auris Agg phenotype of C. auris to fully clarify the organisms' pathogenic mechanisms.

19
These investigations must consider interactions between C. auris and other 20 organisms that comprise the skin and/or wound microbiomes, which may function as 21 important beacons for host invasion of C. auris.

10
For this, data was presented as fold change relative to PBS control. To study the 11 host response to C. auris, a RT 2 profiler array containing genes associated with 12 inflammation and fungal recognition was utilised to assess the transcriptional profile 13 of the skin epidermis following stimulation (C). Data in the heatmap presented as 14 Log 2 fold change relative to PBS control. Finally, expression of two virulence genes,

15
ALS5 and SAP5 was determined in the isolates infected on the tissue, values 16 presented as % expression relative to the fungal specific house-keeping gene, β-17 actin (D and E). All epithelial cells or tissues were infected in triplicate and statistical 18 significance determined from raw data CT values using unpaired Student's t-tests for 19 comparison of two variables or one-way ANOVA with Tukey's multiple comparison 20 post-test for more than two variables (* p < 0.05, ** and § § p < 0.01, *** p < 0.001, 21 **** p < 0.0001).