Candida biofilm production is associated with higher mortality in patients with candidaemia

Candidaemia is a common life‐threatening disease among hospitalised patients, but the effect of the Candida biofilm‐forming ability on the clinical outcome remains controversial.


| INTRODUC TI ON
Candidaemia is the fourth and sixth most common bloodstream infection among hospitalised patients in the United States of America and Europe, respectively, accounting for 8%-15% of all nosocomial bloodstream infections. [1][2][3][4] Candidaemia episodes are associated with unacceptably high overall mortality ranging from 38% to 75%. [5][6][7] A population-based surveillance from Italy revealed that biofilm production in Candida isolates is significantly associated with central venous or urinary catheter use and administration of total parenteral nutrition in the host patient. 8 Biofilm formation is a key virulence factor for Candida species; it serves as a focus in bloodstream infections, protects the fungal cells against the immune response; furthermore, the presence of biofilms is associated with reduced susceptibility to antimicrobial agents. 9 Until recently, data on the clinical impact of Candida biofilm on mortality were scarce and there were more contradictions than answers due to the diverse experimental settings and variable patient populations. [10][11][12] Several studies drew attention to the problems possibly arising from the variability of biofilm detection methods and the poorly standardised cut-off points to distinguish biofilm-producer isolates from strains with low biofilm production capability. [10][11][12][13] Generally, two simple rapid screening platforms are used to examine biofilm production, that is quantification of biomass production using crystal violet-based staining 14 and quantification of the metabolic activity of sessile cells using XTT (2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide) reduction assay. 15,16 Most studies addressing the association between fungal biofilms and patient parameters characterise biofilm production using either of the two approaches; however, these two measurements are not interchangeable, that is a biofilm with high biofilm mass may show low metabolic activity or vice versa.
Consequently, it is possible that some clinical parameters are differently associated with the biofilm-related variables.
Nonetheless, the association of clinical variables with biofilm mass and metabolic activity at the same time is less studied. 12 Hence, reports on the association of Candida biofilms with mortality are frequently contradictory 8,[10][11][12][13] ; several studies showed that biofilm production is linked to significantly higher mortality, 8,10 while other results do not support this observation. 11,12 In this study, we used XTT and crystal violet assays to simultaneously assess the clinical impact of biofilm development on mortality, in patients with candidaemia.

| Patients and definitions
One hundred and twenty-seven Candida isolates collected between January 2013 and December 2018 at 127 candidaemia episodes from the 1667-bed University Clinical Center in Debrecen, Hungary, were studied. An episode of candidaemia was defined as the isolation of Candida spp. from at least one blood culture in a patient. If Candida spp. was isolated more than once from the same patient, positive blood cultures at least 30 days apart were considered two different candidaemia episodes. Patients with multiple Candida spp. from the same episode were excluded. Data on patient demographics, underlying medical conditions and details of antimicrobial therapy were collected through review of the medical history. Concomitant bacteraemia was defined as isolation of potentially pathogenic bacteria from the same blood culture sample set. Patient outcomes were followed from the first positive blood culture until 30 days or death.

| Isolate identification and storage
All isolates were identified using matrix-assisted laser desorption/ ionisation/time-of-flight (MALDI/TOF) analysis. Briefly, single colonies from freshly grown isolates from Sabouraud dextrose agar were spread onto ground-steel target plate in duplicates.
Afterwards, cells were macerated by adding 1 μL formic acid solution. After drying, 1 μL of matrix solution (trans-Cinnamic acid dissolved in a 50:47.5:2.5 acetonitrile:water:trifluoroacetic acid) was added to each spot and then the plate was air-dried at room temperature. Mass spectra were generated with Microflex Biotyper (Microflex, Bruker Daltonics) using the manufacturer's standard settings. Mass fingerprints were acquired using FlexControl ver-

| Biofilm formation
Biofilms were prepared as described previously. 15 One-day-old biofilm mass was quantified with the crystal violet assay; metabolic activity was examined with the XTT assay as previously described by O'Toole (2011) 14 and Hawser et al (1996), 16 respectively. The crystal violet and XTT measures of biofilm formation do not necessarily correlate and can be considered two independent characteristics of a biofilm. 13,17 Therefore, isolates were categorised either based on their level of biomass or based on metabolic activity distribution as low, intermediate or high biofilm producers according to a study published by Rajendran et al (2016). 10 Isolates within the first quartile (Q1) were classified as low biofilm producers, isolates with a biomass higher than the third quartile (Q3) were classified as high biofilm formers and those in between were classified as intermediate biofilm producers (Q2). 10 The principle of this classification was also followed for categorisation based on metabolic activity. For better comparability with similar studies, isolates were compared as follows: low versus intermediate/high biomass and low versus intermediate/high metabolic activity.

| Susceptibility testing of planktonic cells
Antifungal susceptibility of Candida isolates to fluconazole (Sigma), amphotericin B (Sigma), caspofungin, micafungin and anidulafungin (all from Molcan) was determined using the broth microdilution method in RPMI-1640 (with L-glutamine and without bicarbonate, pH 7.0 with MOPS; Sigma), using the CLSI standard M27-A3 protocol. 18 The final concentrations of the drugs were 0.03-32 mg/L, 0.016-8 mg/L and 0.008-4 mg/L for fluconazole, amphotericin B and the three tested echinocandins, respectively. All isolates were measured in triplicate, and medians were used for further analysis.
Candida parapsilosis ATCC 22019 and Candida krusei ATCC 6258 were used as quality control strains in each experiment. Planktonic minimum inhibitory concentrations (MICs) were read visually following a 24-hours incubation period at 35ºC. For fluconazole and the tested echinocandins, the partial inhibition criterion was used (at least 50% growth reduction as compared with the growth control), while for amphotericin B, the total inhibition criterion (100% growth reduction as compared with the growth control) was used. 18

| Susceptibility testing of biofilms
The activity of tested antifungals against biofilms was assessed using the XTT assay. [19][20][21] The concentrations tested in MIC de- The background was measured from the fungus-free well. [19][20][21] All isolates were tested in three independent experiments, and the median of the three values was used for further analysis. In each biofilm-related experiments, C albicans SC 5314 reference strain was used as the quality control.

| Statistical analysis
The biofilm mass and the metabolic activity of biofilms by different Candida species were analysed using the Kruskal-Wallis test with Dunn's post-test. The differences in MIC values against given antifungals for planktonic and sessile cells were analysed using the Wilcoxon matched-pairs tests. Fisher's exact test was used to investigate whether any of the risk factors in patients predisposed to infection with biofilm-former isolates and to evaluate whether biofilm production of the isolates had any relationship with 30-day mortality. The effect of biofilm production (logarithmised absorbance values with XTT or crystal violet) as well as the different patient characteristics and conditions on patient mortality was analysed in multivariate logistic regression models. Data were analysed using GraphPad Prism 6.05 software or in R. The results were considered significant if the P-value was <.05.

| Ethical approval
The authors confirm that the ethical policies of the journal, as

| RE SULTS
The majority of patients were male (59%), with a mean age of 61 years.
Seventy-nine per cent of examined patients was treated in the intensive care unit. Candida albicans was the most prevalent species; it accounted for 51% (65/127) of examined candidaemia episodes, followed by C parapsilosis (23/127; 18%), Candida tropicalis (19/127; 15%), C krusei (10/127; 8%), Candida glabrata (4/127; 3%) and other less common species (6/127; 5%) Candida lypolitica, Candida catenulata, Candida guilliermondii, Candida dubliniensis, Candida inconspicua and Candida orthopsilosis each causing one episode. Figure 1A,B show that biofilm mass and the metabolic activity of sessile cells by different Candida species were heterogeneous, irrespective of the species examined. C tropicalis isolates had significantly higher biofilm mass compared with other Candida species (P < .001-.05); furthermore, their metabolic activity was significantly higher compared with C glabrata and C krusei (P < .01-.05; Figure 1A Three out of these nineteen C albicans isolates showed intermediate/high biofilm mass and intermediate/high metabolic activity. This ratio was 83%, 20% and 67% for C tropicalis, C krusei and C glabrata, respectively. Sessile MIC values were determined in cases of biofilms produced by tested C albicans and non-albicans isolates with intermediate and high metabolic activity as determined according to the above described quartile-based classification (Figure 2A We assessed the relationship between biofilm mass or metabolic activity and 30-day mortality, clinical characteristics and underlying conditions (Table 1A and 1). The highest 30-day mortality was observed in C tropicalis candidaemia (68%), followed by C albicans (62%), C parapsilosis (30%), C krusei (30%) and C glabrata episodes (25%). Significantly higher 30-day mortality was observed in cases of intermediate and high biofilm formers (61%) (P = .023; Table 1A).
Notably, all C tropicalis, C parapsilosis and C glabrata isolates related to a fatal outcome belonged to intermediate/high category in terms of both biofilm mass and metabolic activity, whereas this ratio was F I G U R E 1 Biofilm formation of different Candida species by crystal violet (A) and XTT (B) assays. Biomass and metabolic activity were quantified spectrophotometrically by reading absorbance at 540 nm and 492 nm, respectively. Three replicates were used for each isolate with the mean of each represented. C tropicalis isolates showed significantly higher biofilm mass compared with other Candida species (P < .001-0.05) (A); furthermore, their metabolic activity was significantly higher compared with C glabrata and C krusei (P < .01-0.05) (B)  ; Table 1B). In contrast, Candida cells with low metabolic activity were associated with a 33% mortality rate (Table 1B). Interestingly, the ratio of bacteraemia was markedly higher (P = .015) for Candida species with low metabolic activity (53% and 28% for low metabolic activity and intermediate/ high metabolic activity, respectively; Table 1B). In multivariate analysis, infection with an isolate producing biofilm with intermediate/ high metabolic activity is an independent predictor of mortality (Table 2B), while this association was not demonstrated using crystal violet-based biofilm mass determination (Table 2A).  26 According to a number of studies, patients infected by biofilm-producing isolates have a worse prognosis 8,10,27 ; however, contradicting studies also exist. 11,12 This controversy is probably due, at least partly, to variability in the cut-off values used to distinguish low from intermediate and high biofilm producers 12,13 The other possible reason may be the high variability of the examined patient populations. 12,13 Currently, two in vitro methods (crystal violet and XTT reduction assays) are used to assess the biofilm-forming ability. 14,16 These colorimetric assays are non-invasive and non-destructive compared with alternative methods (eg viable cell counting by flow cytometry). 14,16 However, there is no consensus on the criteria for stratification of biofilm production (as discussed earlier). In this study, we used both of these standardised methods separately to assess simultaneously the impact of biofilm mass as well as metabolic activity on mortality. We applied the method suggested by Rajendran et al (2016) for categorisation of biofilm production, who found a notable difference in 30-day survival between patients infected with low and high biofilm-producing isolates (35% and 41% mortality for low and high biofilm formers, respectively) based on the applied assays, which included XTT assay, crystal violet assay and SYTO-9 assay; in this study, the results of all three assays correlated and their effect was not analysed separately. 10   safranin-related assay, to be associated with increased mortality. 12,28 However, the latter study involved a large proportion of C parapsilosis isolates, which are generally associated with low mortality. 28 Rajendran et al (2016) considered biofilm mass as a predictor of mortality, but in their analysis the different markers of biofilm production were not analysed separately. 10 In the present study, mortality was significantly less frequent among patients infected with isolates producing low than with those producing intermediate/high biofilm mass, but biofilm mass was not a predictor of mortality in multivariate analysis.

| D ISCUSS I ON
In our study, we isolated a significantly higher ratio of bio- which may be associated with the increased frequency of fungal mucosal infections. 34,35 We observed candidaemia with bacteraemia in 34% of all cases consistent with the previously described data ranging from 6% to 34.5%. 36 Interestingly, the ratio of candidaemia with concomitant bacteraemia was significantly higher for Candida isolates that produced biofilm with low metabolic activity (this was not seen with biofilm mass). Bacteria may release certain compounds that interfere with the metabolism of planktonic and sessile Candida cells, interfere with biofilm formation or with fungal quorum sensing. Morales et al (2013) showed that phenazines derived from P aeruginosa impair C albicans metabolic activity and influence the cellular morphology, cell-cell interactions and biofilm formation. 37  In conclusion, our analysis confirmed that the ability to form biofilms especially with intermediate/high metabolic activity is related to higher mortality in candidaemia. Furthermore, it highlights that considering biofilm mass and metabolic activity equivalent may lead to conflicting results and these should be analysed and interpreted separately.

CO N FLI C T O F I NTE R E S T
László Majoros received conference travel grants from Cidara, MSD, Astellas and Pfizer. All other authors report no conflicts of interest.

AUTH O R CO NTR I B UTI O N
RK conceived the ideas; EV, RK and LM collected the data; FN, ZT, LF and AB performed the biofilm-forming-related tests and susceptibility tests; EV, RK and G. K. analysed the data; and EV, RK, LM and G.
K. wrote the manuscript.