Hematogenous candidiasis (candidemia) is the most common presentation of invasive candidiasis (IC) in nosocomial settings, and is responsible for over 5% of all bloodstream infections (Arendrup, 2013; Cantey and Milstone, 2015). Candidemia causes increase in hospital length of stay and healthcare costs, and is associated with high mortality rates, usually over 40% in developing countries (Bloos et al., 2013; Antinori et al., 2016).

Although the incidence of candidemia by C. albicans is decreasing, it remains the most frequently isolated species in several centers, closely followed by Candida glabrata, Candida tropicalis, and Candida parapsilosis, all with the potential to exhibit resistance to fluconazole (FLC) and echinocandins (da Matta et al., 2017; Lamoth et al., 2018).

In 2005, based on the genetic diversity among C. parapsilosis isolates, two new cryptic species were described, and C. parapsilosis is now considered a species complex with C. parapsilosis (sensu stricto), Candida orthopsilosis and Candida metapsilosis (Tavanti et al., 2005). Candida parapsilosis is predominantly found in clinical specimens, while C. orthopsilosis and C. metapsilosis represent around 10% of isolates (Trofa et al., 2008; Nosek et al., 2009).

The incidence of C. parapsilosis fungemia has greatly increased over the last 30 years (Guinea, 2014), being the leading cause of candidemia in some European, Asian and Latin American medical centers (Singaravelu et al., 2014; Caggiano et al., 2017; da Matta et al., 2017). Neonates with invasive devices and echinocandin exposure have been related to C. parapsilosis bloodstream infections (Pammi et al., 2013).

The presence of C. parapsilosis on healthcare workers' (HCW) hands may contribute to horizontal transmission of this organism, causing invasive disease in patients with no prior evidence of colonization (Singaravelu et al., 2014). In addition, the ability of C. parapsilosis to form biofilm has been associated with colonization of medical devices, helping it to remain viable for at least 4 weeks on plastic healthcare surfaces, facilitating the occurrence of nosocomial outbreaks (Kuhn et al., 2004; Singaravelu et al., 2014; Welsh et al., 2017). After C. albicans, C. parapsilosis is the biggest biofilm producer among Candida species (Larkin et al., 2018) and candidemia by biofilm forming isolates has been associated with higher mortality rates (Tumbarello et al., 2007; Trofa et al., 2008).

Candida parapsilosis complex outbreaks have been reported worldwide since 2004, mainly in neonatal intensive care units (NICU), however the information regarding the susceptibility pattern of isolates has only been evaluated in a few studies (Wang et al., 2016; Benedict et al., 2017). The failure of infection prevention and control programs, such as inadequate environmental disinfection and hand hygiene, may be the main reason for the emergence of these outbreaks in hospital settings (Guo et al., 2015; Qi et al., 2018).

Although echinocandins are recommended as the initial therapy of IC in critically ill patients, FLC is an alternative drug for clinically stable patients with IC, and has been frequently administered as first choice antifungal agent (Bassetti et al., 2016; Pappas et al., 2016; O'Leary et al., 2017). Indeed, in developing countries, FLC is the leading antifungal for the treatment of IC, largely because of its lower price compared to echinocandins (Nucci et al., 2013a,b). However, this practice allied to biofilm production may be contributing factors for the concerning emergence of persistent clusters of FLC-resistant C. parapsilosis, as recently reported in Brazilian (Pinhati et al., 2016) and South African (Magobo et al., 2017) centers. In addition, inadequate antifungal treatment may increase the length of hospitalization and associated costs (Arnold et al., 2010).

In this study, a microbiological investigation of C. parapsilosis complex clinical isolates collected in a Brazilian public university hospital medical center between 2012 and 2016 was conducted. During this retrospective investigation, a previously unrecognized outbreak of a persistent azole-resistant C. parapsilosis (sensu stricto) clone was identified in the intensive care unit (ICU), correlating it with the main clinical data from the patients with IC.

Some studies have shown that ANSCP isolates tend to cluster in ICUs (Pfaller et al., 2008; Raghuram et al., 2012; Govender et al., 2016), although the molecular analysis of the isolates was not performed in these reports. Magobo et al. (2017) reported previously undetected outbreaks in NICU of FLC-resistant C. parapsilosis in a South African hospital in a retrospective study employing molecular typing (Magobo et al., 2017).

The high number of cases of ANSCP (71.4%) observed in this cohort, led to a molecular analysis of these isolates to reveal the occurrence of probable outbreak in the adult ICU. This study revealed the persistence of an azole non-susceptible C. parapsilosis clone over 2 years in this unit. Indeed, although most C. parapsilosis complex isolates are usually susceptible to azoles, recent Brazilian reports have indicated the emergence of IC due to ANSCP (da Costa et al., 2014; Giacobino et al., 2016; Alencar et al., 2017). To our knowledge, this is the second reported outbreak of ANSCP in Brazil (Pinhati et al., 2016).

Low resistance rates to FLC have been reported among C. parapsilosis isolates worldwide, including Latin America (1.1%) (Nucci et al., 2013a), USA (4.0%) (Pfaller and Castanheira, 2016), and Asia-Western Pacific (5.7%) (Pfaller et al., 2015). However, undetected outbreaks of ANSCP infection could be ongoing in Brazil and other countries because at present there are very few studies aiming to continuously monitor azole resistance in this species and investigate the molecular relationship between isolates (Pinhati et al., 2016; Magobo et al., 2017).

Recent reports indicate that mutations in ERG11 gene contribute to azole resistance in C. parapsilosis (Berkow et al., 2015; Grossman et al., 2015; Asadzadeh et al., 2017). In accordance with this finding, we observed that the A395T mutation in ERG11 occurs exclusively among azole-resistant isolates. However, in one FLC-resistant isolate (58H) this mutation was not observed, indicating that azole resistance involves other molecular mechanisms, such as overexpression of ERG11 and efflux pumps (Souza et al., 2015). Further analysis of overexpression of ERG11, CDR1, and MDR1 should be performed to clarify the non-susceptibility of isolates without missense mutation in ERG11.

The high prevalence of the A395T mutation in FLC-resistant C. parapsilosis clinical isolates was also noted in previous studies by Grossman et al. (2015) and Choi et al. (2018). In addition, the latter suggested that isolates with this mutation may have a higher propensity to cause clonal transmission and to persist in nosocomial settings than FLC-resistant C. parapsilosis without A395T mutation. Regarding our ASCP isolates, the G1193T mutation was detected in only one isolate (188H). However, other studies identified this mutation in both susceptible and resistant isolates, suggesting that this mutation alone is not related to azole resistance (Berkow et al., 2015; Grossman et al., 2015; Asadzadeh et al., 2017).

Kuhn et al. correlated an outbreak of C. parapsilosis infections with virulence factors, and they observed that clonal isolates had a higher ability to form biofilms than unrelated strains, suggesting that biofilm production plays a role in C. parapsilosis outbreaks (Kuhn et al., 2004). By contrast, all clonal isolates of our study displayed a low capacity to form biofilm, indicating that this ability may not be a critical causal factor of outbreaks.

Wang et al. performed phylogenetic analyses of C. parapsilosis isolates from concomitant blood and CVC tip cultures that show identical genotypes, indicating that the catheter is an important source of infection by this species (Wang et al., 2016). Moreover, several studies have demonstrated horizontal transmission in IC outbreaks by C. parapsilosis, illustrating the genetic similarities between clinical and isolates found on HCW hands (Huang et al., 1999; Barchiesi et al., 2004; van Asbeck et al., 2007; Hernández-Castro et al., 2010). In our cohort, four clonal ANSCP were recovered from CVC tips indicating the inadequate handling of this device as a probable route of transmission of this outbreak.

Our molecular epidemiological investigation revealed that all clonal ANSCP isolates were detected in the adult ICU (Figure 2). One of the related isolates (29H) displayed an exclusive heterozygosis in Locus B, which may indicate a possible genetic microevolution that may be related to stress conditions, such as previous antifungal exposure (Sabino et al., 2010; Pulcrano et al., 2012). Although FLC-resistant, the 58H isolate presented low genetic similarity (42.9%) to the ANSCP clone (Figure 2). We believe that the geographic distance of the nephrology unit from the adult ICU, as well as the time that elapsed between the isolation of the 58H and the clonal ANSCP isolates (over 1 year) may explain the lack of relation between them.

Pinhati et al. (2016) found that diabetes was an independent risk factor for infection by ANSCP. They concluded that diabetic patients are more prone to be colonized by Candida, as this medical condition is associated with increased contact between HCW and patients (Pinhati et al., 2016). A similar pattern was found in this study, with 44 and 0% of the patients with ANSCP and ASCP isolates, respectively, having diabetes. However, due to the low number of patients analyzed, the difference between the groups was not statistically significant. There was a trend (p = 0.07) for previous antifungal exposure and positive ANSCP cultures. Three of the patients with ANSCP did not receive FLC for over 48 h, which may explain the probable horizontal transmission between the patients (data not shown). We found that having received corticosteroids was associated with positive culture for the clonal ANSCP (Table 3). Indeed, most of the patients that were admitted to the adult ICU had pulmonary diseases requiring corticosteroid prescription. We believe that the persistence of the clonal ANSCP isolates may have been due to contaminated instruments, environment and HCW's hands. Unfortunately, as this study was performed retrospectively, samples from neither HCW hands nor the environment could be collected.

Echinocandins have been recommended as first-line treatment for candidemia by the main clinical guidelines (Cornely et al., 2012; Pappas et al., 2016), including fungemia by C. parapsilosis. Indeed, similar 30-days mortality rates have been observed between echinocandin- and fluconazole-treated patients with fungemia by this species (Fernández-Ruiz et al., 2014; Chiotos et al., 2016). On the other hand, C. parapsilosis isolates demonstrate innately high MICs for echinocandins (Garcia-Effron et al., 2008), which led to EUCAST to classify isolates with MICs below 2 mg/L as intermediate, instead of susceptible as recommended by the CLSI (Meletiadis et al., 2016). Most of our patients with candidemia were treated with echinocandins and 87.5% showed poor outcome. However, at least 60% of these patients had severe comorbidities (Table 4) that could have contributed more importantly to lethality (Lin et al., 2015).

In conclusion, our results suggest the possibility of persistence of clonally related azole-resistant C. parapsilosis (sensu stricto) isolates in hospital settings, irrespective of their capacity of biofilm formation. Furthermore, azole resistance in C. parapsilosis isolates is emerging in Brazilian medical centers, and the molecular mechanisms involved should be monitored in conjunction with azole susceptibility profiles. Finally, the screening of additional samples from HCW hands, indwelling devices and environmental surfaces will improve understanding of the nosocomial transmission process, and will help to control the spread of ANSCP clones more effectively.

This study was carried out in accordance with the recommendations of the research ethics committee of the Medical School of the University of São Paulo, São Paulo, Brazil. The protocol number 1.912.028 was approved by the research ethics committee of the Medical School of the University of São Paulo for analysis of microorganisms and clinical data records, without the need for written informed consent of the subjects.

DT and GD designed the study, helped with acquisition and data analysis, drafted and revised the work, approved the final work, and agreed with all the aspects of the work. JA helped with acquisition and data analysis, drafted and revised the work, approved the final work, and agreed with all the aspects of the work. GL and MN helped with acquisition of the data, approved the final work, and agreed with all the aspects of the work. CC and RG helped with data analysis, approved the final work, and agreed with all the aspects of the work. GB helped with data analysis, drafted and revised the work, approved the final work, and agreed with all the aspects of the work.