Carriage of Methicillin-resistant Staphylococcus aureus among people living with HIV-AIDS in inner São Paulo State, Brazil: molecular and spatial epidemiology

Background: Methicillin-resistant Staphylococcus aureus (MRSA) is increasingly recognized as a threat for people living with HIV/AIDS (PLWHA). However, the magnitude of asymptomatic MRSA colonization in that group varies among different countries and geographic regions. Methods: We conducted a study that aimed at identifying the prevalence, risk factors and spatial epidemiology of both overall S. aureus and MRSA colonization among PLWHA from small cities from inner São Paulo State, Brazil. MRSA isolates were characterized using Pulsed-Field Gel Electrophoresis (PFGE), and submitted to typing of the Staphylococcal Chromosome Cassete (SCC)mec. Spatial analysis was performed to search for geographical clusters and correlation with socioeconomic indicators. Results: In a first point prevalence survey, nasal and oropharyngeal swabs of 368 people were collected. Sixty-seven subjects from the city of Botucatu were surveyed for colonization in two other occasions, and had swabs collected from household members. The prevalence rates for S. aureus and MRSA in the first survey were 25.8% and 2.7%. The overall S. aureus colonization was negatively associated with the use of beta-lactams and of illicit drugs. On the other hand, MRSA colonized subjects were more likely to use crack and to have been admitted to a hospital during the past year. Repeated surveys found additional cases of MRSA colonization, but all subjects were positive in only one occasion. Four PFGE clusters were characterized, grouping subjects in household, city and region level. Of 19 total MRSA isolates, only one did not harbor SCCmec type IV. Spatial analysis of households of subjects living in the city of Botucatu found significant overdispersion of cases, but no association with socio-economic indicators. Conclusion: We found small but relevant prevalence of MRSA among PLWHA. Community and healthcare-associated risk factors were identified, so that predominant routes of


Introduction
The impact of colonization and infection with Methicillin-resistant Staphylococcus aureus (MRSA) among people living with HIV/AIDS (PLWHA) has been extensively studied in the late 1990s and early 2000s [1], [2]. However, the changing epidemiology of both HIV and MRSA in the past decades produced a shift in major concerns. Early studies focused on the risks of acquiring multidrug-resistant organisms during hospital admission [3], [4]. But since then the advent of highly active antiretroviral therapy (HAART) made hospital admissions less common for PLWHA in many countries [5]. On the other hand, the emergence of community-associated (CA)-MRSA -often much virulent -posed a new threat for that group [6].
In Brazil, the governmental AIDS program provides universal access to diagnosis, therapy and routine laboratory tests (including CD4 lymphocytes count, measures of viral load and genotyping of viral resistance) [7]. This approach reduced mortality and slowed the growth of the epidemics [8], [9]. On the other hand, it failed to interrupt a relevant epidemiological trend, the interiorization. Ever since the late 1990s, the proportion of PLWHA who live in small cities -often far from great urban centers -is continuously increasing [10]. In response to this picture, there was a major increase in the number of public clinics providing care to PLWHA., That number rose from 33 to 540 in 2002 and 663 in 2007 [11], [12].
On a separate -but related -topic, the global emergence of CA-MRSA still puzzles experts and public health authorities [13]. There are reports of relevant incidence of CA-MRSA infection (or prevalence of asymptomatic colonization) involving populations as diverse as USA residents [14], [15], Maltese people [16], Australian indigenous communities [17] and Pygmies in Gabon [18]. In this setting, some questions arise, concerning the burden of 4 colonization, patterns of transmission, the risk of invasive infection and the impact on vulnerable populations -including PLWHA.
Recent studies emphasize the intersection of HIV and CA-MRSA pandemics [19], [20], [21] Prevalence of colonization in this group is variable -with studies reporting rates ranging from similar to those reported for the general population [22] up to more than 16% [23].
In previous studies, we documented CA-MRSA infection and colonization in small cities from inner São Paulo State, Brazil [24], [25]. This pattern of "interiorization" has similarities with that described for AIDS, so that both epidemics intermingle in a new setting.
Our study was designed to address this intersection. We aimed to identify the prevalence of overall Staphylococcus aureus and MRSA colonization among PLWHA attending a reference outpatient clinic in inner São Paulo State, Brazil. We were particularly interested in identifying predictors for colonization, as well as clonal pattern of isolates and the spatial distribution of urban cases.

Study setting and design
The study was conducted in the "Specialized Care Center for Infectious Diseases" (SCCID) from Faculdade de Medicina de Botucatu (Botucatu Medical School). It is located in Botucatu city (130,000 inhabitants; 22º 53' 09" S, 48º 26' 42" W), and is the single referral center for surrounding municipalities -and area comprising 500,000 people.
Presently, the SCCID cares for over 500 PLWHA.
The study had a cross-sectional design, with serial prevalence surveys conducted in 2014-2016. We included all PLWHA aged 15 years or more who were attending regularly the SCCID and agreed to participate the study. There were no specific exclusion criteria. 5 However, due to ethical legislation constraints, we could not include incarcerated persons in our study.

Point prevalence survey
In the first phase of the study, a point prevalence survey was carried out, including subjects cared for in the SCCID. All those subjects had nasal and oropharyngeal (throat) swabs collected, on the day they attended the clinic to perform routine laboratory tests (CD4 + lymphocyte counts and measure of viral load).

Serial surveys
All subjects who lived in the city of Botucatu were invited to participate the further phases of the study. Briefly, a home visit was scheduled for 30 days after inclusion in the study.
In the first visit, we performed: (a) collection of second-time samples of nasal and oropharyngeal swab; (b) collection of swabs from the same sites in all household communicants who were not infected by HIV; and (c) georreferencing of households. The third survey was performed in the patients scheduled return for routine medical appointments (in average, 3 months after the first survey).

Microbiology and molecular methods
Specimens were transported in Stuart medium and cultured in Baird Parker agar. Species identification was performed through phenotypic and genotypic techniques, as previously described [25].
Susceptibility tests followed guidelines from the Clinical Laboratory Standards Institute (CLSI), using disks for oxacillin and cefoxitin [26]. However, our definition of methicillinresistance for practical purposes was based on the detection of the mecA gene by Polymerase Chain Reactions (PCR), as described by Murakami et al [27]. The multiplex-PCR protocol described by Milheiriço et al [28] was used for the characterization of the 6 staphylococcal cassette chromosome mec (SCC mec).
Molecular strain typing of MRSA isolates was performed with Pulsed-Field Gel Electrophoresis (PFGE). We applied a protocol of DNA digestion with the enzyme smaI, modified from McDoughal et al. [29]. The analysis of similarity was performed using the Dice coefficient. Clusters were defined on the basis of similarity values over 80%.
Dendrograms were drawn based on Unweighted Pair Group Method Using Arithmetic Averages (UPGMA) in the BioNumerics 6.1 software (Applied Maths, Belgium). International MRSA clones and isolates from a general population-based survey from the city of Botucatu [25] were included as controls in the dendrogram. The first analytical step involved bivariate analysis. Dichotomous variables were analyzed using the Chi-square or Fisher Exact test. For continuous variables, we used the Mann-Whitney U test. Multivariable analysis was performed using logistic regression methods. A 7 forward selection process was applied to include variables in the models [30]. The thresholds for inclusion and removal of variables in the models were P < 0.05 and P > 0.1, respectively. The Hosmer & Lemeshow test was applied to analyze goodness-of-fit in the multivariable models [31]. Whenever colinearity of two or more variables was detected, one of them was selected on the basis of the goodness-of-fit of the resulting model. The inicial descriptive phase involved plotting subjects (overall PLWHA, S. aureus positive) and applying the Kernel density estimation for each group [32]. A second step of exploratory analysis was performed for three groups (PLWHA, S. aureus positive, S. aureus negative), who were integrated in separate to identify coincidental points. We applied the "integrate" ArcGIS tool, using a standard distance of fifteen meters between the points. Afterwards, the "collect events" tool was applied for counting overlapping points. Moran's I was calculated to estimate global spatial autocorrelation [32]. A similar approach was used to estimate the correlation of subjects coordinates with georreferenced information of families' income and average people living in households.
All the maps were generated in the coordinate system SIRGAS 2000 UTM Zone 22, with the 8 Transverse Mercator projection and datum Planimetric SIRGAS 2000.

Ethical issues
This study was conducted according to the principles expressed in the Declaration of Helsinki. It was approved by the reference Committee for Ethics in Research ("Comitê de Ética em Pesquisa" from "Faculdade de Medicina de Botucatu". City of Botucatu, São Paulo State, Brazil). A written informed consent was obtained from all study subjects or their legal guardians. 37.52, P = 0.006). In the alternative model, excluding neurocryptoocosis, recent hospital admission was the single predictor identified (OR, 4.04; 95%CI, 1.14-14.34; P = 0.04).  PFGE typing three clusters (A, B, C) grouping isolates from more than one study subject and an additional cluster (D) grouping an isolate from this sample with one from the population-based survey conducted in 2011 in Botucatu [25]. The dendrogram is shown in Fig. 1, and Table 3 describes the clusters identified. Interestingly, the only isolate for which we could not assign a SCCmec type (345) did not belong to any cluster and was the least similar to other isolates from our study. Spatial Analysis 15 We georreferenced the households of study subjects (PLWHA) who lived in the urban area of Botucatu. For the purpose of this analysis, we took in account the positivity for S. aureus or MRSA in any culture in the serial surveys. Figure 2 presents the point distribution of subjects, highlighting those who were positive for overall S. aureus or MRSA. The Kernel density is shown inf Fig. 3. The autocorrelation test did not find geographical clusters for whole subjects or for MRSA. However, we found that subjects harboring overall S. aureus were more geographically dispersed than expected at random chance (Moran's Index, -0.19; Expected Index, -0.04; P = 0.03).

Discussion
Our findings can be interpreted in many ways. First, we documented colonization with MRSA among PLWHA from inner São Paulo State. The point prevalence rate (2.4%) is not particularly high when compared to the international literature -where prevalence of up to 16.8% are reported [23]. In fact, our rate is lower than the pooled prevalence of 6.9% reported in a recent meta-analysis [33]. On the other hand, a survey performed in the early 2000s including HIV-positive outpatients from inner São Paulo did not find any subject harboring MRSA [34]. Even though that study was conducted in another city, it is reasonable to infer that prevalence of MRSA colonization may be growing. It is also worth noting that, among our subjects who were studied in three surveys, the cumulative prevalence (i.e., MRSA positive in any of the surveys) was higher than 10%.
The emergence of CA-MRSA was a turning point both in the epidemiology and in the clinical relevance of staphylococcal infections [35]. However, authors have recently reported a blurring of the definitions of "community-associated" and "healthcareassociated" infections [36]. This is especially the case for special groups. PLWHA are a heterogeneous population that includes both seemingly "healthy" persons (asymptomatic, most achieving viral control with proper therapy) and others with poor compliance to therapy and a history of several opportunistic infections. Often, this latter group presents variable amounts of social vulnerability, including poverty, alcoholism and addiction to illicit drugs [38]. In countries -such as Brazil -where the poorest people have access to public health, this group is more often admitted to acute care hospitals. Therefore the same population may be exposed to risk factors associated with CA-MRSA (illicit drugs, poor hygiene practices) and HA-MRSA (frequent admissions), making it difficult to ascertain the origin of isolates on epidemiological grounds [37], [38].
Our results are exemplary. Eighteen out of 19 isolates tested positive for SCCmec type IV, usually found in CA-MRSA. However, recent hospital admission was epidemiologically associated with MRSA carriage in one of the logistic regression models. An alternative model associated MRSA with previous neurocryptococosis (an infection that invariably requires hospital admission) and the use of crack. Neurocryptococosis may be a proxy that suggests an association of MRSA to admissions that took place more than a year before the survey (and therefore did not meet the "recent admission" criteria). On the other hand, the use of crack has become epidemic in Brazil, and is part of a common milieu that combines poverty, violence and sexually transmitted diseases [39].
It is worth noting that, despite the small number of subjects colonized with MRSA in the first survey, the factors associated with this outcome were more meaningful than those associated with overall S. aureus. Indeed, subjects colonized with S. aureus were less likely to have received beta-lactams antimicrobials and more likely to use illicit drugs. This latter finding is puzzling, but it may reflect changes in microbial ecology of nares and throat that favor overgrowth of competing microorganisms. One should notice that the use of intravenous drugs -a reported risk factor for MRSA -is rare in Brazil, a pattern reflected in our sample. [40] There is now sufficient evidence for international spread of specific MRSA clones [41].
However -and contrary to the case for HIV -the routes for this dissemination are far less clear [42]. Therefore, studies that approach networks of transmission -such as households and neighborhoods -are required [43], [44]. Miller et al [45] carried out a survey selecting subjects in a household level, and found that, in dwellings with more than one member colonized with S. aureus, 50% carried that same strains. Sexual transmission has been demonstrated [46], but it obviously does not explain all events of spread among household members. We addressed this issue by performing a survey among HIV-negative household contactants of study subjects. As presented above, a positive index subject was associated with greater risk of colonization of household members in analysis for both overall S. aureus and MRSA. However, only one patient had a household contact colonized with the same strain as his. We did not type methicillin-susceptible S. aureus, and therefore could not detect possible transmission of those strains among family members.
Interesing insights arise from comparison of this study with the population-based survey of nasal S. aureus colonization, conducted in Botucatu in 2011. In that study, we identified two instances of MRSA transmission among family members [25]. Also, two subjects living in the same street harbored similar isolates. It is worth noting that 3 out of 6 isolates from that study grouped with strains from the present investigation. This finding suggests the long-term persistence of specific clones in the population. The fact that clusters grouped isolates from people living in Botucatu and in other neighboring cities points out either to regional spread or to cross-transmission during outpatients appointments. 18 The clinical significance of MRSA colonization among PLWHA is not completely clear. While for some authors colonization is a major risk factor for invasive infections [47], others believe this association does not apply to the dinamics of CA-MRSA [48]. Our study was not designed to address those issues, and we found no association between MRSA (or S. PLWHA -was addressed previously, but in that study no subject carried MRSA. [34] In that study, advanced HIV disease was associated with persistent colonization.
In order to assess both the distribution of PLWHA and of colonization, we georreferenced dwellings of subjects who lived in Botucatu. This allowed us to identify "hot spots" concentrating cases of HIV/Aids and colonized subjects. However, we failed to demonstrate special correlations with two important measures of social vulnerability -the average number of people in dwellings and the average family monthly income. The number of MRSA-positive subjects was too low to warrant a specific analysis.
Our study has some limits, which regard the relatively small sample (especially in serial surveys), not collecting swabs from other body sites (e.g., groin) and the fact that typing of methicillin-susceptible strains was not performed. However, it also has strengths, including an effort to address extensively the subjects vulnerability and the combined use of classical, molecular and spatial epidemiologic methods. In

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.