Combination of Monocyte Human Leukocyte Antigen-DR and β-d-Glucan for the Early Diagnosis of Invasive Candida Infection in Critically Ill Patients: A Proof of Concept Study

Background: Precision medicine risk stratication is desperately needed to both avoid systemic antifungals treatment delay and over prescription in the critically ill with risk factors. The aim of the present study was to explore the combination of host immunoparalysis biomarker (monocyte human leukocyte antigen-DR expression (mHLA-DR)) and Candida sp wall biomarker β-d-glucan in risk stratifying patients for secondary invasive Candida infection (ICI). Methods: Prospective observational study. Two intensive care units (ICU). All consecutive non-immunocompromised septic shock patients. Serial blood samples (n=286) were collected at day 0, 2 and 7 and mHLA-DR and β-d-glucan were assayed. Secondary invasive Candida sp infection occurrence were then followed. Results: Fifty patients were included, 42 (84%) had a Candida score equal or greater than 3 and 10 patients developed a secondary invasive Candida sp infection. ICU admission mHLA-DR expression and β-d-glucan (BDG) failed to predict secondary invasive Candida sp infection. Time-dependent cause-specic hazard ratio of ICI was 6.56[1.24-34.61] for mHLA-DR < 5000 Ab/c and 5.25[0.47-58.9] for BDG > 350pg/mL. Predictive negative value of mHLA-DR > 5000 Ab/c and BDG > 350 pg/mL combination at day 7 was 81% [95%CI 70-92]. Conclusion: This proof of concept study suggests that mHLA-DR predicts ICI in high risk patients with septic shock. The added value of BDG and other fungal tests should be regarded according to the host immune function markers. are expressed as number and percentage. Continuous data are expressed as median and quartiles. Comparisons were made between patients with invasive Candida infection and without invasive Candida Categorical data are expressed as number and percentage. Continuous data are expressed as median and quartiles. Comparisons were made between patients with invasive Candida infection and without invasive Candida infection.

Assessment (SOFA) score [19], the presence of comorbidities, reason for admission to the ICU and cause of septic shock. During the ICU stay, the following data were collected at days 0, 2 and 7; SOFA score, length of stay, duration of mechanical ventilation, need for vasopressor and survival at ICU discharge. Antibacterial and antifungal drugs were given according to the recommendation applied in a same way in both ICUs.
Immunological data EDTA-anti-coagulated tubes were collected on each visit day (days 0, 2 and 7) for blood cell count, lymphocyte phenotype, CD4 T-cell, CD8 T-cell count and mHLA-DR. Circulating mHLA-DR expression was assessed by ow cytometry (NAVIOS®; Beckman-Coulter) in accordance with the standardized recommendation [20,21]. Monocytes were characterized on the basis of their CD14 expression. Results were expressed as the number of anti-HLA-DR antibodies per cell (AB/c).

Mycological data
Mycological samples were collected in case of infection suspicion according to the physicians in charge of the patient. Candida colonization was screened at days 0 and 7 in at least 3 sites from urine, gastric aspiration, tracheal aspiration and cutaneous swab. Candida Score [22] was calculated for each visit day. Colonization was de ned when one nonsterile sample site was positive for Candida sp. Multifocal colonization was de ned when more than one non sterile sample site was positive for Candida sp. Standardized procedures for in vitro identi cation of microorganism were used according to the usual procedures of the local mycology laboratory.
Mycological biomarkers Β-d-glucan (BDG) was obtained on days 0, 2 and 7. Blood samples (15 mL) were centrifuged, separated into aliquots, and stored at -80 °C until analysis performed at the end of the study. The BDG assay (Fungitell®, Associates of Cap Cod Inc., Easy Falmouth, MA, USA) was performed according to the manufacturer's recommendations. The cutoff value was set according to the company recommendations at 80 pg/mL [23,24]. None of the results was available to the physician in charge.

Endpoints
The primary endpoint was the comparison of the kinetics during the rst week after septic shock of mHLA-DR alone and in combination with BDG in ICI and NICI patients.

Statistical analysis
Data are expressed as mean ± SD or SEM for normally distributed data, and median with 95% con dence index (95%CI) for non-normally distributed data. Comparisons and biomarkers kinetics between ICI and NICI patients were made at day 0, 2 and 7. Continuous variables were compared using Student's t test for normally distributed variables and the Mann-Whitney rank-sum test for non-normally distributed variables. The chi-square test or the Fisher exact test was used to compare categorical variables.
A cause-speci c hazard model was built to assess the association of candida-related and immunologicals variables on the probability of death in the ICU or invasive candida infection [25]. In this model, the occurrence of ICI was the variable of interest, while death was considered as a competing event. Being discharged alive from the ICU was considered as a censored variable. Considering the low occurrence of candida infection only univariate analysis was performed. Candida variables and impaired immune function variables were tted as time-dependent covariables. The direct effect on the risk of ICI was estimated by cause-speci c hazard ratio (csHR).
Analysis of sensitivity, speci city, positive predictive value and negative predictive value was calculated for mycological and immunological markers separately and in association. For each marker, value over/under the threshold was only considered if before an event (ICI, death or ICU discharge). Delay of infection was estimated by the Kaplan-Meier method and compared between high and low risk patients (according to mHLA-DR and BDG values) with the log-rank test. A Monte Carlo simulation (considering the covariance between the two slopes) was also performed in order to assess the potentially best ROC curve. Expecting a 20% incidence of Candida sp positive samples following septic shock, a drop of 50% in mHLA-DR in the ICI patients by day 7 and no signi cant difference between day 0 and day 7 in NICI patients, we estimated that 50 patients with septic shock would be needed assuming a bilateral test, an alpha risk of 0.05 and a power of 0.8. A P value of less than 0.05 was taken as the signi cance level. We used SAS 9.4 (Sas Institute, NC, USA), R 3.0.2 (Vienna, Austria) and GraphPad Prism 6 for all the statistical analyses.

Results
Fifty consecutive patients met the inclusion criteria. The baseline characteristics and outcome are shown in Table 1. Peritonitis (22/50, 44%) and pneumonia (11/50, 22%) were the main sites of initial sepsis. Intraabdominal candidiasis was the more frequent ICI (n = 5), followed by candidemia (n = 4) and pleural infection (n = 1). A Candida Score ≥ 3 was found in all ICI patients before infection occurred and in 32 (80%) NICI patients (p = 0.12). Antifungal therapy was used in 20 patients (40%). Preemptive treatment was administered in 11 (27%) NICI patients. Among the nine ICI treated patients, 5 (55%) received initial preemptive therapy that was pursued once the ICI was con rmed while 4 (45%) received initial curative treatment. In one patient, the diagnosis of ICI was made after ICU discharge and the treatment was not started.
A total of 286 immunological and mycological blood samples were collected. At day 0 and at day 2, mean mHLA-DR expression was not different between ICI and NICI patients and above the threshold of 8000 AB/c. At day 7, ICI patients had a lower mHLA-DR expression than NICI patients (3451 ± 978 AB/c vs. 12049 ± 2044 AB/c; p = 0.002) ( Fig. 2A). At day 0, mean BDG was 97.5 ± 35.6 pg/mL and 133.5 ± 24.5 pg/mL for ICI and NICI patients respectively (p = 0.38) (Fig. 2B).
In the cause-speci c univariate model (  . Patients with mHLA-DR < 5000 Ab/c and > 5000 Ab/c at day 7 had a signi cantly different cumulative incidence of ICI after day 7 with a log-rank test of 0.017 (Fig. 3A). No difference was found between patients with BDG > 80 and < 80 pg/ml (Fig. 3B).

Discussion
This prospective observational pilot study reports that a low value of mHLA-DR (< 5000 Ab/c) after septic shock is associated with a higher risk of invasive candida infection. Adding mycological markers such as BDG do not improve the predictive value in a high risk population.
Both innate and adaptive immune response is of importance for controlling Candida colonization and infection. Blood monocytes and tissue macrophages recognize pathogen associated molecular patterns like fungal cell wall component. Antigen-presentation and subsequent cytokine secretion lead to T-helper lymphocyte activation and secondary adaptive immune response [26]. The expression of cell surface marker such as monocyte human leukocyte antigen-DR is regulated by multiple pro and antiin ammatory mediators [4] and mHLA-DR level of expression re ects activation of monocytes and immune function. Thereby, low mHLA-DR expression has been described as a robust marker of immune dysfunction [27] and outcome following septic shock [4,28,29]. Furthermore, the expression of mHLA-DR lower than 8000 AB/c has been described to increase the risk of hospital acquired infection following septic shock or multiple trauma in the critically ill [20,21,30,31]. In this study, mHLA-DR value < 8000 Ab/c was neither associated with ICI nor death. Monocytic HLA-DR value < 5000 Ab/c was yet associated with ICI. Our population is mainly patients with sepsis related to peritonitis. Leijte et al. found median mHLA-DR over 6000 Ab/c in bacterial peritonitis [32]. Invasive candida infection commonly appears in highly immunosuppressed patients and lower value of mHLA-DR could be associated with higher risk of ICI. Continuing or declining immune de ciency is associated with an increased mortality [32]. Similarly, our results suggest an increased probability of ICI when mHLA-DR remains < 5000 Ab/c after day 7 of septic shock.
We further explored whether Candida colonization and a fungal biomarker, BDG, could be of interest.
Colonization and prediction scores have been developed to identify ICU patients at high risk of ICI. In our study, colonization and Candida Score (≥ 3) were positive in all ICI patients and 80% of NICI patients, consistent with the low positive predictive value reported elsewhere [22,[33][34][35].
Colonization driven antifungal treatment exposes to unnecessary antifungal treatment and its consequence on antifungal resistances and ecology [36][37][38]. In this regard, fungal biomarkers such as BDG have received increasing attention. Recent studies have evaluated the BDG performance to predict ICI and have reported high sensitivity and speci city [39,40]. Others have reported mitigated results with sensitivity ranging from 40 to 100% [41][42][43][44]. Because of studies reporting high negative predictive value, BDG has been used either to rule out ICI or to discontinue antifungal therapy as a part of antifungal stewardship [45][46][47]. In our study, negative predictive value and speci city of BDG by itself did not reach clinical relevance to be used to rule ICI out ( Table 2). This is in accordance with the recent study by Angebault et al. who reported a BDG sensitivity of 64% in candidemia [48]. This is also in line with studies that has shown that BDG driven therapy was not associated with outcome improvement [49,50].
When the combination of both host response (mHLA-DR) and fungal biomarker (BDG) was evaluated, the positive and negative value was not improved compared to mHLA-DR alone. Indeed, BDG predictive value may be decreased in high risk population (Candida score 3 in 84% of patients). Past history of cancer was associated with an increased risk of ICI. This result was previously described by Lortholary [51]. An increased expression of negative T-cell co-stimulatory molecule (PD1, CTLA4) was described in both cancer [52] and candida related sepsis [53] revealing possible immunological similarities.
The present study has some major limits. First, this is a single institution study with a relatively small cohort and a limited number of patients at day 7 because of deaths or discharge from the ICU. Second, 27% of patients in NICI group have received pre-emptive treatment at inclusion or during the ICU stay limiting ICI occurrence in some of them.

Conclusion
This study suggests that a sepsis induced immune dysfunction (mHLA-DR < 5000 Ab/c) is associated with a higher risk of ICU acquired invasive candida infection in patient with septic shock. Combination with fungal biomarker (BDG) do not enhance the prediction value. Flow chart At day 7 after septic shock, 24/50 patients had invasive candida infection, were discharge or dead. All invasive candida infections occurred before day 14.

Figure 2
Time course of mHLA-DR after septic shock. Mean HLA-DR in patients with invasive candida infections still decrease between day 2 and day 7 after septic shock whereas it increased in patients without ICI without statistical signi cance.

Figure 3
Time course of BDG after septic shock. Mean BDG values is not signi cantly higher in patient with or without ICI at day 2 or day 7 after septic shock.

Figure 4
Probability of invasive Candida infection after day 7 according to day 7 mHLA-DR value. In a competing risk survival model, considering death and ICU discharge as competing risks, patients with mHLA-DR < 5000 Ab/c at day 7 after septic shock have a signi cant higher probability of invasive candida infection. Figure 5