Mortality due to Cryptococcus neoformans and Cryptococcus gattii in low-income settings: an autopsy study

Cryptococcosis is a major opportunistic infection and is one of the leading causes of death in adults living with HIV in sub-Saharan Africa. Recent estimates indicate that more than 130,000 people may die annually of cryptococcal meningitis in this region. Although complete diagnostic autopsy (CDA) is considered the gold standard for determining the cause of death, it is seldom performed in low income settings. In this study, a CDA was performed in 284 deceased patients from Mozambique (n = 223) and Brazil (n = 61). In depth histopathological and microbiological analyses were carried out in all cases dying of cryptococcosis. We determined the cryptococcal species, the molecular and sero-mating types and antifungal susceptibility. We also described the organs affected and reviewed the clinical presentation and patient management. Among the 284 cases included, 17 fatal cryptococcal infections were diagnosed. Cryptococcus was responsible for 16 deaths among the 163 HIV-positive patients (10%; 95%CI: 6–15%), including four maternal deaths. One third of the cases corresponded to C. gattii (VGI and VGIV molecular types, Bα and Cα strains) and the remaining infections typed were caused by C. neoformans var. Grubii (all VNI and Aα strains). The level of pre-mortem clinical suspicion was low (7/17, 41%), and 7/17 patients (41%) died within the first 72 hours of admission. Cryptococcosis was responsible for a significant proportion of AIDS-related mortality. The clinical diagnosis and patient management were inadequate, supporting the need for cryptococcal screening for early detection of the disease. This is the first report of the presence of C. gattii infection in Mozambique.

From November 2013 to March 2015, complete diagnostic autopsies (CDAs) were performed in 284 deceased patients in the two study sites. Two hundred twenty-three cases were recruited from Mozambique: 169 were adults over 15 years of age (112 women, 57 of whom were maternal deaths, i. e., women dying during pregnancy, partum, post-partum or within 42 days of termination of pregnancy 15 ), and 54 were children from 1 month to 15 years of age (37 males and 17 females). Sixty-one cases were recruited from Brazil: 59 adults (38 males and 21 females, including one maternal death) and two children. A description of the study design and inclusion criteria for the cases has been published elsewhere [16][17][18] . All cases fulfilling the inclusion criteria were included in the study; these were (1) a CDA requested by the clinician as part of the medical evaluation of the patient and; (2) a verbal informed consent to perform the autopsy given by the relatives. Traumatic deaths were excluded. A member of the study staff was tasked with liaising with the families in cases of deaths occurring in the pediatric department, but only after the clinicians had asked for consent for postmortem examination 17 . All the clinical records available regarding admission preceding death of each patient were reviewed and the clinical data were collected in a standardized manner.
Autopsy procedures and determination of the cause of death. The autopsy procedures and a description of the pathological and microbiological methods used have been reported elsewhere 19,20 and for Briefly, samples of blood, cerebrospinal fluid (CSF), bone marrow and key organs such as the liver, lungs, CNS, spleen and kidneys, as well as uterus in all women of reproductive age were collected for histopathological and microbiological analysis. The microbiological methods included universal screening for relevant pathogens (e.g. HIV, viral hepatitis, tuberculosis, malaria) and bacterial/fungal culture of autopsy samples, targeted screening depending on the patient's condition (i.e. screening for Cryptococcus, Toxoplasma gondii and Pneumocystis jirovecii in HIV-infected cases), and additional specific testing according to the histopathological findings. Following the analysis of the CDA samples, a panel composed of a pathologist, a microbiologist, and a clinician with expertise in infectious diseases and epidemiology evaluated all the data (including the clinical information) and assigned the final cause of death.
Laboratory methods. All cases in which the cause of death was assigned to a cryptococcal infection were further characterized. The histological evaluation included periodic acid-Schiff staining (PAS) of the samples of both lungs, CNS, bone marrow, spleen, liver, kidney and uterus, as well as Grocott-Gomorimethenamine silver staining in all CNS samples. Microbiological evaluation included a specific real time PCR for Cryptococcus spp. 21 , which was performed in the samples of both lungs, CNS, bone marrow, spleen, liver and uterus. PCR cycle www.nature.com/scientificreports www.nature.com/scientificreports/ threshold values >38 were considered negative. Plasma and CSF samples were tested by both real time PCR and the Cryptococcus antigen (CrAg) lateral flow assay (LFA) (IMMY Inc., Norman, Oklahoma). Discrimination between C. neoformans var. grubii, C. neoformans var. neoformans, and C. gattii was achieved by amplification of the rRNA intergenic spacer (IGS) region, followed by forward and reverse Sanger sequencing of the amplicons. Identities of the cryptococcal species were assigned based on a >98% match to the IGS sequence of a Cryptococcus reference strain (C. gattii: CBS 6289, ATCC MYA-4561, CBS 6955, ATCC MYA-4563; C. neoformans var. grubii: ATCC MYA-4564, ATCC MYA-4565 and C. neoformans var. neoformans: ATCC MYA-4567) as described previously 22 .
Cryptococcal strains isolated from the cultures of the autopsy samples underwent a consensus multi-locus sequence typing scheme for C. neoformans and C. gattii 23 . In addition, the sero-mating type of these strains was determined by multiplex PCR as described previously 24,25 . An anti-fungogram was performed for each strain using the Sensititre ® YeastOne ® susceptibility plate (TREK Diagnostic Systems, Thermo Fisher Scientific, Oakwood Village, USA). statistical methods. Descriptive analysis was performed using univariate statistics with means and standard deviations for continuous variables and frequency distributions for categorical variables. All analyses, data manipulation, and implementations were done using Stata MP version 15 (Stata, College Station, TX, USA).  Table 1. The median age was 34 years (range 6-44 years); 11 cases (65%) were men. In 13 out of the 16 (81%) HIV-infected patients, a positive HIV test result was reported in the clinical records and was apparently unknown by the clinician in the other 3 cases. Four out of the 16 (25%) HIV-infected patients were on ART, but the duration of ART was only reported in one case.

Results
Headache was the most common symptom (13 patients, 76%), followed by fever and vomiting (eight cases each, 47%). Upon admission to hospital, eight patients (47%) were confused and/or agitated, two patients (12%) were lethargic, and another two (12%) were fully comatose. Meningeal signs were detected in seven patients (41%). The mean time from admission to death was 9.3 days (95%CI: 2.4-16.2). Cryptococcal infection was considered the first clinical diagnostic option in only 4/17 (23%) of the confirmed cases, whereas it was included in the differential diagnosis in eight patients (47%). Antifungal treatment (fluconazole or Amphotericin B) had been prescribed to seven patients but to only five of the clinically suspected cases. Seven of the patients (41%) died within 72 hours of admission, and 12 out of the 16 HIV-positive patients (75%) died within one week of admission.
The clinical records of the remaining 267 cases included in this study were reviewed for clinical diagnosis of cryptococcal infection. Among these, three HIV-infected patients were clinically diagnosed with cryptococcal meningitis, but no evidence of cryptococcal infection was found in the autopsy (the cause of death was identified as toxoplasmosis in two cases and tuberculosis in one case). Table 2 shows the final cause of death, the cryptococcal species, other associated diagnoses identified in the CDA, as well as the results of the PAS staining, Cryptococcus real time PCR and CrAg test in the different samples analysed. Identification of the cryptococcal species was successful in 15 out of the 17 cases. In two cases, the tissue samples provided insufficient DNA to perform IGS amplification and sequencing. Interestingly, among the 15 cases identified, five (33%) were C. gattii whereas the remaining cases (66.6%) were C. neoformans var. grubii.

Histopathological and microbiological analysis of the fatal cryptococcal infections.
Twelve of the 17 cases (70%) were diagnosed as disseminated infections and the remaining 5(30%) as meningoencephalitis.The CNS was affected in all 17 cases. The following were the most affected organs (PAS and/ or PCR positive): the lungs (88%), spleen (76%), liver (71%), bone marrow (59%) and kidney (47%). In addition, in five out of six (83%) deceased women, the pathogen was detected in uterus samples, and in one case it was detected in the placenta. Cryptococcus was found in more than 5 different samples in 11 patients (65% of the cases). More organs were affected in C. neoformans var. grubii than in C. gattii infections (mean of 6.2 vs. 2.8 organs positive by PAS staining), but no specific histopathological differences were observed between the two Cryptococcus species. Molecular testing expanded the detection of Cryptococcus in 14 additional tissues that were negative for PAS staining. In contrast, two tissue samples were positive for PAS staining but negative by real time PCR. Periodic acid-Schiff staining and Grocott-Gomorimethenamine silver staining showed identical results. Representative histopathological images of different affected tissues are shown in Fig. 1 Cryptococcus strain characterization. Eight strains were isolated by culture, seven from the CSF and one from the blood. The molecular characterization and antifungal susceptibility of the cryptococcal strains isolated are shown in Table 3. All C. neoformans var. grubii corresponded to the VNI molecular type, whereas two different genotypes, VGI and VGIV, were identified in the two C. gattii isolates. Sero-mating type Aα was found in all the C. neoformans isolates. Bα and Cα strains were found among the C. gattii strains.
Minimum inhibitory concentrations (MICs) of a variety of antifungal agents were determined (Table 3). All triazole antifungal agents (fluconazole, voriconazole, itraconazole and posaconazole) had MICs less than or equal to the defined epidemiological cut-off values (ECVs) 26 . One isolate showed a MIC to flucytosine (16 µg/mL), one dilution higher than the ECV (8 µg/mL). Although all the isolates presented a MIC ≤ 1 µg/mL to amphotericin, the MICs of three C. neoformans molecular type VNI (1 µg/mL) and two C. gattii (1 µg/mL) were one dilution higher than the defined ECVs (0.5 µg/mL) 27 .

Discussion
In a large series of nearly 300 autopsies, we performed a thorough histopathological and microbiological analysis of 17 fatal cryptococcal infections, 12 from Mozambique and five from Brazil. Several studies on cryptococcal infection have been carried out in Brazil 14,28 , but there are no data from Mozambique. Indeed, despite being a highly endemic area for HIV, a literature search was unable to find reports describing the burden of this pathogen in this country. Thus, to our knowledge, this is the largest autopsy-based description of fatal cryptococcal infections in Mozambique. Interestingly, albeit being Maputo and Manaus two sites from different countries in terms of climate and income, some of the results obtained were very similar, such as the HIV prevalence in deceased    www.nature.com/scientificreports www.nature.com/scientificreports/ patients (over 60%) and the Cryptococcus associated mortality in HIV positive patients. In this study Cryptococcus was responsible for 10% of the deaths among HIV-infected patients. This figure is in agreement with current estimates of 15% of AIDS-related deaths due to CM 4 . Although treatment with ART has led to an important reduction in the mortality of HIV-infected patients in sub-Saharan Africa, at present, cryptococcal-related death seems to remain similar to 2008 estimates. In agreement with other autopsy series, HIV-associated cryptococcosis is frequently presented as a disseminated infection 14,28 . Other studies have reported that up to one half of cryptococcal  www.nature.com/scientificreports www.nature.com/scientificreports/ infections are detected in HIV-infected patients receiving ART 29 . In our series, only a quarter of the HIV-positive cases (4/16) were under ART. Unfortunately, the duration of ART was only available in one case. This case had a 6-year history of ART and a high viral load, and therefore, likely represents a treatment failure or defaulted. Other explanations for cryptococcal infection in patients receiving adequate ART include recent treatment initiation in patients with very low CD4 counts (late HIV diagnosis) and cases of immune reconstitution inflammatory syndrome 30 , which might also have been present in our series. Despite progress in ART deployment in sub-Saharan Africa, close to 50% of HIV-infected patients continue presenting to health facilities with advanced HIV disease 31 , thereby being at high risk of cryptococcal infection and death. In this regard, the CrAg can be detected up to three weeks before the onset of CM symptoms 32 and therefore, screening of asymptomatic HIV patients followed by preemptive antifungal treatment might identify patients at risk of developing the disease and contribute to reducing CM-related mortality 33,34 . In the present study, most cases died within the first week of hospital admission, which is in contrast with other series reporting a mean of two weeks between admission and death 35 . Several factors might have contributed to the rapid fatal outcome in our series. Firstly, the low frequency of clinical suspicion of cryptococcosis may explain the absence of prescribed antifungal treatment in more than half of the patients. On the other hand, some patients may have presented to the hospital with advanced severe stages of cryptococcal infection, considering that seven died within three days of admission (four within 24 hours). The longest hospital admission involved a child who died of C. gattii meningitis about two months after hospital admission. C. gattii infections are rare in children, and as in the child in this series, the infection often occurs in previously healthy children and presents with CNS involvement 9 . The little knowledge available about C. gattii in sub-Saharan Africa has been obtained from studies performed in South Africa. Interestingly, in our study four out of the 12 cryptococcal infections (25%) from Mozambique were due to C. gattii infections. Although the numbers are small and should be interpreted with caution, these data contrast with previous findings from South Africa, where C. gattii represented only 2.4% of the cryptococcal strains isolated over a two-year period 36 . C. gattii infections have mainly been reported in immunocompetent hosts, until studies from South Africa revealed that it affects immunocompromised patients as well 37,38 . It has been reported that the VGI C. gattii molecular type is much more likely to affect immunocompetent patients than VGIV 9 . However, both genotypes were detected in HIV-infected patients in our study. All strains of C. neoformans var. grubii were molecular type VNI and possessed the sero-mating type Aα, similar to most of the strains described in South Africa. To our knowledge, this is the first report of C. gattii in Mozambique, and the Bα and Cα types detected have been previously reported in South Africa 39,40 . Although no clinical breakpoints are available for Cryptococcus, antifungal susceptibility testing did not suggest clear resistance patterns, with only MICs one dilution above the ECV being found.
A few isolated autopsy reports of fatal C. gattii infections in immunocompetent patients have been reported 41 , which may be explained by the fact that the distinction between C. gattii and C. neoformans was not performed in many studies. Although the numbers are small, dissemination of the infection seems to be less intense in HIV-infected patients with C. gattii compared to C. neoformans. Larger studies are needed to better assess the histopathological features of C. gattii in HIV-infected patients. Moreover, studies based on post-mortem examinations should be performed to better assess the mortality attributable to specific pathogens in low-and middle-income countries. Accurate mortality data can then impact public health policies, for example, guiding prophylactic and treatment schemes for infectious diseases.
In conclusion, our study highlights the substantial mortality associated with cryptococcal infections among HIV-infected patients, supporting current recommendations 42 of CrAg screening and preemptive therapy, which is even more relevant in settings in which insufficient pre-mortem clinical suspicion is given to this highly prevalent and life-threatening opportunistic infection. ethics approval and consent to participate. The study protocol received approval of the National Mozambican Ethics Committee (ref.342/CNBS/13) and the Ethics Committee of the Hospital Clinic of Barcelona (Spain; approved, File 2013/8677). MIA and CDA procedures were only conducted after verbal informed consent was provided by the relatives.

Data Availability
All relevant data are within the paper. Any additional data use and transfer is monitored by ISGlobal's Data Management and Biostatistics Unit (contact e-mail: ubioesdm@isglobal.org).