Species of Cryptosporidia Causing Subclinical Infection Associated With Growth Faltering in Rural and Urban Bangladesh: A Birth Cohort Study

This birth cohort study describes the early childhood burden of cryptosporidiosis in rural and urban Bangladesh. Important findings include an association of repeated and subclinical cryptosporidiosis with growth faltering, and disparate predominant Cryptosporidium species in the urban vs rural site.

Diarrhea remains a leading global cause of morbidity and mortality in children <5 years of age [1]. Two recent multicenter studies demonstrated the importance of cryptosporidiosis, caused by an intracellular protozoan parasite, as a leading cause of childhood diarrheal disease [2,3]. Many species of Cryptosporidium can infect humans; however, Cryptosporidium hominis and Cryptosporidium parvum are typically considered the most common [4][5][6]. Clinical manifestations of cryptosporidiosis are variable as symptomatic diarrheal disease occurs in only a subset of cases while most remain subclinical [7][8][9]. Early cryptosporidiosis, whether symptomatic or subclinical, has been associated with impaired growth and cognitive development [7,10,11].
Current understanding of human cryptosporidiosis is largely based upon diagnosis with overt diarrhea; yet given the associated long-term sequelae of even subclinical cryptosporidiosis, a better epidemiologic understanding in various settings is needed. In this active surveillance study, we describe the natural history of cryptosporidiosis over the first 2 years of life in 2 sites in Bangladesh, 1 urban and 1 rural. We characterize the burden of cryptosporidiosis and, using regression modeling, estimate the impact of Cryptosporidium infection on growth faltering. Furthermore, we describe a marked and unexpected difference in causative species, with C. hominis most common in the urban site compared with Cryptosporidium meleagridis in the rural site. 13], and Mirzapur is a rural subdistrict located 60 km northwest of Dhaka, as previously described [14]. Further site descriptions are provided in Table 1 and the Supplementary Methods. Pregnant women in their second trimester were identified. In Mirpur, field research assistants performed a census. In Mirzapur, study participants were identified using a demographic surveillance system previously established by the Bangladeshi government [14]. Interested women meeting study criteria and providing informed consent underwent clinical examination, urinalysis, and gestational age confirmation by ultrasound. Exclusion criteria were age <18 years, gestational age ≥7 months, hypertension, edema, proteinuria, or intent to migrate from the study area ( Figure 1). After delivery, infants assessed by the study medical officer (SMO) within the first 7 days of life were eligible for enrollment. For logistical reasons at Mirpur, the monthly maximum enrollment was 27.
The SMO collected demographic and socioeconomic data using a structured questionnaire at enrollment. Thereafter, field research assistants performed twice-weekly, in-home visits to interview caregivers and collect information regarding child morbidity and diarrhea. The SMO assessed all children monthly. Caregivers were encouraged to bring the child to the study clinic whenever the child developed symptoms of any illness, not only diarrhea. If acutely ill during an in-home visit, the child was referred to the study clinic. If the SMO determined that further treatment was needed, care was provided free of charge at the International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) Mirpur Treatment Centre, the icddr,b Dhaka Hospital in Mirpur, or the Kumudini Hospital in Mirzapur.
Height and weight were measured for mothers at the infant enrollment visit. Infant length (to nearest 0.1 cm using length board and plastic tape) and weight (kilograms, measured with electronic scale; TANITA, HD-314) were obtained every 3 months. Weight-for-age z score (WAZ) and length-for-age z score (LAZ) were determined using World Health Organization Anthro software (version 3.2.2). The change in LAZ (Δ-LAZ) was calculated by subtracting the enrollment LAZ from that at 24 months.

Detection of Cryptosporidiosis
Single, fresh stool specimens were collected from children every month (monthly surveillance) and during episodes of diarrhea. A modified Qiagen stool DNA extraction protocol was performed incorporating a 95°C incubation and a 3-minute glass bead-beating step (Qiagen, Valencia, California). All stool samples were tested for Cryptosporidium by quantitative polymerase chain reaction (qPCR) assay modified from Liu et al [15] (Supplementary Methods). A cycle threshold of 40 was used. The pan-Cryptosporidium primers and probes target the 18S gene in multiple species known to infect humans. All samples positive for Cryptosporidium were subsequently assessed for species identification using previously described Lib13 qPCR for C. hominis and C. parvum and a novel C. meleagridis qPCR assay [16,17] (Supplementary Methods).

Definitions
Diarrhea was defined as ≥3 loose stools within a 24-hour period as reported by the child's caregiver. Infection with Cryptosporidium was defined as detection of Cryptosporidium DNA by qPCR from stool. Samples were grouped into a single infection if occurring within 65 days of a preceding positive sample based upon gp60 genotyping of a subset of samples (≤65 days: 93% concordance; >65 days: 33% concordance) [18]. Cryptosporidium infection phenotype (diarrheal or subclinical) was based upon symptoms at the time of initial Cryptosporidiumpositive stool sample (diarrheal stool vs monthly surveillance). Causative species was assigned based upon results of any stool sample obtained during an infection.

Statistical Analysis
Analyses were performed using R version 3.3.3, 32-bit. The P values for Table 1 were calculated using Wilcoxon or χ 2 tests. Differences in cumulative incidence of first Cryptosporidium infection, depicted by Kaplan-Meier curves, were assessed by log-rank test. Association of cryptosporidiosis during the first 24 months of life with Δ-LAZ was assessed using t test initially, and subsequently evaluated using stepwise linear regression to adjust for potentially important confounders. Variables considered in the regression with entry P value <.1: Cryptosporidium infection, enrollment LAZ, maternal body mass index (BMI), household income, water source, water treatment, and exclusive breastfeeding days. Association of recurrent Cryptosporidium infections with Δ-LAZ was evaluated among children with 0, 1, and ≥2 detected infections. Differences in Δ-LAZ among these groups were tested using 1-way analysis of variance (ANOVA) with post hoc analysis by Tukey correction for multiple comparisons, and further evaluated using stepwise linear regression adjusting for the same potential confounders stated above.

Ethical Considerations
The Ethics and Research Review Committee at icddr,b approved this study; the Institutional Review Board of the University of Virginia provided a reliance agreement. Informed written consent was obtained from a parent or guardian of each child.

Cryptosporidiosis Burden
From 1 July 2014 through 30 June 2017, 6637 and 6417 stool samples were collected in Mirpur and Mirzapur, respectively, and tested for Cryptosporidium by qPCR (Figure 1). Fewer diarrheal stools were collected in Mirzapur (n = 271) than Mirpur (n = 1243) despite high rates of stool collection during diarrheal episodes in both sites (99% and 85% in Mirzapur and Mirpur, respectively). In Mirpur, Cryptosporidium was detected in 496 separate stool samples representing 240 distinct infections among 161 children. Of these infections, 182 were subclinical (76%, initial detection in a monthly surveillance stool) and 58 were diarrheal (initial detection in a diarrheal stool sample). Of 250 children enrolled at Mirpur, 36% (n = 89) had no detectable Cryptosporidium infections, 38% (n = 96) had 1, and 26% (n = 65) had ≥2. In Mirzapur, Cryptosporidium was detected in 186 separate stool samples representing 138 distinct infections among 114 children. One hundred thirty-five infections were phenotypically subclinical (98%) and 3 were diarrheal. Of 258 children enrolled at Mirzapur, 56% (n = 144) had no detectable Cryptosporidium infections, 35% (n = 90) had 1, and 9% (n = 24) had ≥2. When stratified by site and according to presence or absence of detected Cryptosporidium infection during the first 2 years of life as well as whether Cryptosporidium infections were diarrheal or subclinical, the only statistically significant differences observed in baseline characteristics were for use of treated water in Mirpur. Compared with the no cryptosporidiosis group (n = 74 [83.2%]), use of treated water was less common among those with any cryptosporidiosis (n = 114 [70.8%], P = .044) and the subset with diarrheal cryptosporidiosis (n = 33 [64.7%], P = .023; Table 1). Boiling was the predominant form of water treatment in both sites.
Children in Mirpur experienced their first Cryptosporidium infection at a significantly younger age and had a greater overall cumulative incidence by 2 years compared with Mirzapur (

Cryptosporidiosis and Growth Faltering
Long-term impact on child growth was assessed using LAZ. Moderate or severe stunting (LAZ ≤ -2) at birth was found in 14% (34/250) and 15% (38/258) of children from Mirpur and Mirzapur, respectively. Mean LAZ consistently decreased at both sites with increased prevalence of stunting over the first 2 years of life (Supplementary Tables 2 and 3; Supplementary  Figures 2 and 3). At 2 years, 33% (69/210) of children in Mirpur and 20% (51/254) in Mirzapur were moderately or severely stunted.
Delta-LAZ was calculated as the difference between LAZ at 24 months and that at enrollment. Including all participants at both sites, mean Δ-LAZ for children with no detected Cryptosporidium infections was -0.386 (n = 205) vs -0.656 (n = 259) in children with at least 1 infection (P = .0062; Figure 3A). Adjusting for confounding variables (enrollment LAZ, maternal BMI, household size, household income, water source, water treatment, and exclusive breastfeeding days), cryptosporidiosis during the first 2 years of life was significantly associated with an absolute decline in Δ-LAZ of 0.215 (P = .0088; Table 2). Site-specific analysis demonstrated that cryptosporidiosis was associated with a statistically significant decrease in adjusted Δ-LAZ at Mirzapur (-0.253, P = .011), but not at Mirpur (-0.213, P = .13).
A marked difference in predominant Cryptosporidium species was observed between study sites. In urban Mirpur, C. hominis was identified in 93% of infections in which species could be determined (n = 155, 141 monoinfections; Figure 4A). Fortyone infections in which C. hominis was detected were diarrheal and 114 were subclinical. Cryptosporidium meleagridis was the second most frequently identified species occurring in 13% (n = 22, 9 monoinfections), and C. parvum was identified in 2% (n = 3, 2 monoinfections). In rural Mirzapur, the predominant species was C. meleagridis (90%; n = 92, 91 monoinfections; Figure 4B), while C. hominis was identified in 7% (n = 7, all monoinfections) and C. parvum in 4% (total were nearly all asymptomatic (96%; subclinical, n = 96; diarrheal, n = 4). Of 9 diarrheal infections in which C. meleagridis was identified, 5 were coinfections with C. hominis. DISCUSSION We established a longitudinal birth cohort in Bangladesh comprised of 1 rural and 1 urban location to better understand the natural history of cryptosporidiosis in disparate settings. Intensive, active surveillance revealed a high burden of cryptosporidiosis with 64% and 44% of children at the urban and rural site, respectively, experiencing at least 1 Cryptosporidium infection by age 2 years. Consistent with similar prior active surveillance cohort studies in diverse geographic locations, most Cryptosporidium infections occurred in the absence of diarrhea, which we term subclinical [2,3,7,9,19]. The predominance of subclinical infection was particularly striking in rural Mirzapur where only 2% of cryptosporidiosis episodes were diarrheal. Furthermore, a marked difference between the 2 sites in the predominant Cryptosporidium species was observed, with C. hominis most common in the urban site and C. meleagridis in the rural site. Despite these differences, when the cohort was analyzed as a whole, any cryptosporidiosis during the first 2 years of life was associated with a significant decrement in growth attainment, consistent with prior studies [7,10,11]. This adverse association was greatest in children with ≥2 episodes of cryptosporidiosis. When considered separately, the magnitude of effect on Δ-LAZ was similar at both sites, though reaching statistical significance only in the rural site; this finding is particularly striking as nearly all Cryptosporidium infections were subclinical in this location. Collectively, these results suggest that, even without overt diarrhea, cryptosporidiosis is associated with subsequent impaired growth.
The difference in predominant Cryptosporidium species observed between the sites is striking and suggests setting-specific modes of exposure. Our finding that C. meleagridis predominated in Mirzapur contrasts with prior work showing C. hominis as the most common species in this rural site; however, that study evaluated stool collected during moderate to severe diarrhea [20]. Cryptosporidium meleagridis as a cause of medically attended diarrhea has been described in diverse geographic regions including Peru, China, and Cambodia, where it comprised 23% of Cryptosporidium-positive stools [21][22][23][24]. To our knowledge, our data from rural Bangladesh is the first description of C. meleagridis as a major species infecting children in the absence of diarrhea, yet potentially associated with a shortfall in child growth. This suggests a heretofore unrecognized burden of subclinical cryptosporidiosis attributable to a species previously considered an uncommon human pathogen. This finding has significant implications for interventions aimed at reducing Cryptosporidium-attributable morbidity. Vaccines  targeting C. hominis may reduce overt diarrheal cryptosporidiosis yet be ineffective in prevention of Cryptosporidiumassociated developmental faltering if subclinical disease due to other species persists.
A limitation inherent to the observational design is an inability to fully account for all potentially confounding variables. Notably, testing for symptomatic or subclinical presence of enteric pathogens other than Cryptosporidium was not performed. We cannot determine whether and to what extent enteropathogens other than Cryptosporidium are contributing to growth impairment in this cohort, nor whether coinfections may result in additive or synergistic effect. It is possible that detection of Cryptosporidium is simply a surrogate indicator of some other exposure and therefore associated with, but not directly causative of, the observed growth impairment. This alternative explanation could be supported by the observation that water treatment was less common in the urban site among children who experienced cryptosporidiosis. Though statistically significant, the magnitude of difference in water treatment (83% vs 71%) seems insufficient to account for the observed differences in growth. Moreover, in the rural site where the association between cryptosporidiosis and impaired growth attainment was greatest, water treatment was infrequent, with no differences observed between children with or without cryptosporidiosis. Additional study-specific limitations include the definitions used to delimit infections and the characterization of each infection as diarrheal or subclinical. Cryptosporidium infections may be prolonged with oocysts shed intermittently and often in the absence of diarrheal symptoms [25,26]. This, coupled with practical limitation in frequency of surveillance stool collection and use of species-level identification, precluded exact determination of when 1 infection ceased and another began. We chose to define distinct infections when >65 days separated stools with detectable Cryptosporidium based upon concordance of gp60 genotypes performed on a subset of samples. This definition could introduce bias in either direction by over-or undercounting the number of distinct infections. However, most distinct infections using this definition were separated by many months; therefore, the potential is greater for misclassifying multiple infections as a single infection, thereby underestimating the true frequency of cryptosporidiosis. Infections were classified as diarrheal or subclinical based on presence or absence of symptoms at the time of first Cryptosporidium detection. Sole use of the first timepoint of Cryptosporidium detection may bias toward overclassification of infections as subclinical, thereby underestimating somewhat the burden of diarrheal cryptosporidiosis.
Based upon our findings, we again suggest that not only diarrhea but also child growth may be considered an important clinical outcome of cryptosporidiosis. Future studies should aim to further characterize the ecology and prevalence of Cryptosporidium species, recognizing that some may infect without overt diarrhea yet still may impair growth. Further characterization of modes of transmission, which may differ by local environment and predominant species, must inform strategies to interrupt transmission. Our findings also support renewed efforts to better understand human immune responses to Cryptosporidium and elucidation of immunologic correlates of protection. It is only through such concerted and multidisciplinary efforts that cryptosporidiosis and its associated long-lasting sequelae may be abated.