Association between Cryptosporidium Infection and Human Leukocyte Antigen Class I and Class II Alleles

Abstract

Background

Cryptosporidium species are a common cause of diarrhea, which can be severe and protracted in young children and immunocompromised individuals.

Methods

A cohort of 226 Bangladeshi children aged 2–5 years was prospectively followed for >3 years to study the role of host genetics in susceptibility to infection, as well as the community impact of cryptosporidiosis on this population.

Results

Ninety-six children (42.5%) received a diagnosis of Cryptosporidium infection. A total of 51 (22.6%) had asymptomatic infection. Fifty-eight (25.7%) had cryptosporidiosis, of whom 17 (29.3%) had recurrent disease. Children with cryptosporidiosis presented early, and most had abdominal pain and a short course of diarrhea. Infected children were more likely to carry the human leukocyte antigen (HLA) class II DQB1*0301 allele, particularly those with asymptomatic and symptomatic infection (P = .009); a strong association was found between carriage of the DQB1*0301/DRB1*1101 haplotype and development of both asymptomatic and symptomatic infection (P = .008). Infected children were also more likely to carry the B*15 HLA class I allele.

Conclusions

This is the first study to describe a possible genetic component of the immune response to Cryptosporidium infection, which includes HLA class I and II alleles. Cryptosporidiosis in Bangladeshi children aged 2–5 years is common and often recurrent, but the duration is shorter and the abdominal pain greater than that described in children aged <2 years.

Cryptosporidiosis is a common enteric infection that causes self-limiting diarrhea in healthy individuals but can cause severe, protracted, and life-threatening disease in immunocompromised adults and very young children [1–3]. The human immune response to cryptosporidiosis, particularly in children, is only partially understood [3]. Normal CD4⁺ T lymphocyte numbers and function and a T helper-1–biased cytokine pattern appear important for recovery from cryptosporidiosis, on the basis of data from susceptible animal models and individuals with advanced AIDS [1, 4, 5]. A modest pro-inflammatory cytokine response has also been reported in studies of childhood cryptosporidiosis [6, 7].

The role of host genetics, including the human leukocyte antigen (HLA) genes responsible for the major histocompatibility complex (MHC) presentation of antigen to CD4⁺ T cells, in the susceptibility or resistance to symptomatic (i.e., diarrheal) or asymptomatic Cryptosporidium infection has not been comprehensively evaluated. We prospectively studied a large cohort of preschool-aged and grade school–aged children for >3 years. Genetic information on allelic differences was used to describe the role of HLA class I and II polymorphisms in their susceptibility to Cryptosporidium infection. Detailed clinical and epidemiologic data also contributed to understanding the community impact of cryptosporidiosis in this population.

METHODS

Study protocol

The study was approved by the institutional review boards of the University of Vermont (Burlington, VT), the University of Virginia (Charlottesville, VA), the University of Maryland (Baltimore, MD), the National Institutes of Health (Bethesda, MD), and the Center for Health and Population Research, International Centre for Diarrheal Disease Research (Dhaka, Bangladesh). Human experimentation guidelines of the participating institutions and the US Department of Health and Human Services were followed in conducting the research.

Preschool-aged children (aged 2–5 years old) from an urban slum of Dhaka (Mirpur) were enrolled from 1999 through 2002, as previously described [8, 9]. Parents and children were visited and interviewed every other day by health care workers for symptoms of diarrhea. Informed consent was obtained from the parents or guardians prior to enrollment. Treatment protocols for diarrheal diseases have been described elsewhere but did not include treatment for Cryptosporidium infection during the period of the study [9].

Clinical definitions

Diarrhea was defined as ≥3 loose or watery stools in a 24-h period. Cryptosporidiosis (i.e., symptomatic Cryptosporidium infection) was defined as the presence of diarrhea and the detection of Cryptosporidium oocysts in stool by the Cryptosporidium antigen-detection kit (Techlab). Asymptomatic Cryptosporidium infection was defined as the absence of diarrhea and the detection of Cryptosporidium oocysts in stool by an antigen detection test. Control children were all other children without a diagnosis of symptomatic or asymptomatic Cryptosporidium infection during the study period. Stool specimens positive for Cryptosporidium organisms were considered indicative of a distinct episode of Cryptosporidium infection if they were collected ≥60 days after negative results of stool surveillance assays were obtained. Infections were defined as recurrent if they were <60 days apart with a negative stool examination between episodes. The severity of diarrhea was defined by a numerical scoring system known as the Ruuska score, which accounts for duration of diarrhea, maximum number of diarrheal stools/day, vomiting, fever, dehydration, and the level of clinical care required (ranging from no need for clinical care through the need for hospitalization) [10].

Stool sampling and microbiological analysis

Stool specimens were collected ≤24 h after a report of a new diarrhea episode and every month for surveillance in all children. Microbiologic analysis of stool specimens has been described elsewhere but included performance of standard bacteriologic cultures, testing for pathogenic Escherichia coli; testing for parasitic ova; commercial assays for adenovirus, rotavirus, and astrovirus; and antigen testing for Entamoeba histolytica [9]. Evaluation for asymptomatic Cryptosporidium infection involved analysis of frozen and banked surveillance stools by the Cryptosporidium antigen-detection kit (Techlab).

Genetic analyses

Genomic DNA was extracted from 200 μL of peripheral blood, using the Qiagen DNA extraction kit. HLA class I and II typing was performed using polymerase chain reaction (PCR) and sequence-specific oligonucleotides provided by Dynal, as previously described [11]. For genetic comparisons, all siblings were removed from analysis.

Statistical analysis

Epidemiologic and genetic data were analyzed by SPSS software (version 7.5) or Stata software (version 8.0). The Student t test was used for comparison between means; nonparametric tests were used if the data were not normally distributed. χ² analysis and the Fisher exact test were used for categorical variables. P < .05 was considered statistically significant for all tests.

For analysis of HLA data, allele frequencies were counted. Because of the large number of HLA alleles, it was decided a priori that only cases or controls with an allele frequency of >10% would be involved in genetic association analysis. The magnitude of the association between HLA markers and occurrence of Cryptosporidium infection was measured by calculating the odds ratio (OR), using logistic regression. For multiple-case categories (no infection, Cryptosporidium infection without diarrhea, and Cryptosporidium infection with diarrhea), the OR was determined using polychotomous logistic regression. The conservative Bonferroni correction was used to correct for multiple comparisons. This method assumes that the statistical tests performed on the data set are independent. However, because of known interdependence among HLA alleles, we could not consider tests involving the 20 HLA alleles as being separate. We adjusted for this by using the following HLA class categories: A, B, C, DQB1, and DRB1 (n = 5 in the Bonferroni-adjusted analyses; the P value was multiplied by 5).

RESULTS

There were 289 children (147 boys and 142 girls) from Dhaka (Mirpur) who were enrolled during 1999–2002. A total of 226 children were observed for >3 years, with a mean follow-up duration of 2316 days (6.3 years); 8980 stool specimens collected from these patients during the first 3 years of study participation were evaluated for the presence of Cryptosporidium organisms by an antigen-detection assay. One hundred fifty-two stool specimens involving 142 distinct episodes of Cryptosporidium infection in 96 children (42.5%) tested positive for Cryptosporidium organisms. Fifty-eight (25.7%) children had Cryptosporidium diarrhea, 51 (22.6%) had asymptomatic infection, and 13 (5.8% [13.5% of 96 who tested positive for Cryptosporidium organisms]) had both. Clinical and epidemiologic characteristics of children with asymptomatic Cryptosporidium infection and those with cryptosporidiosis are presented in table 1.

Table 1.

Clinical and epidemiologic features of 226 children with symptomatic or asymptomatic Cryptosporidium infection.

Variable	Cryptosporidium infection type
	Symptomatic (n = 58)	Asymptomatic (n = 51)
Percentage of overall study group	25.7	22.6
Male sex	29 (50)	28 (54.9)
Age at diagnosis, mean ± SD, months	76 ± 26	73.3 ± 16
Episodes >1 month apart, no.	84	58
Cryptosporidium-positive stool specimens, no.	91	60
Fever	45 (49.5)	NA
Abdominal pain	34 (58.6)	NA
Vomiting	17 (18.7)	NA
Watery stool	88 (96.7)	NA
3–6 stools/day	78 (85.7)	NA
≥7 stools/day	13 (14.3)	NA
Mild-to-moderate dehydration	55 (60.4)	NA
ORS use before diagnosis	17 (18.7)	NA
Duration of diarrhea <3 days before admission	88 (96.7)	NA
Ruuska score ≥6	36 (39.6)	NA
Recurrent episode(s)	17 (7.5)	6 (2.6)
Recurrent episodes, no. (range)	26 (2–5)	7 (2–3)
Copathogens recovered, no. (%)a
Overall	18 (19.7)	ND
Giardia lamblia	9 (9.9)	ND
Shigella species	5 (5.5)	ND
Aeromonas species	4 (4.4)	ND

NOTE. Data are no. (%) of patients, unless otherwise indicated. NA, not applicable; ND, not done; ORS, oral rehydration solution.

^aData are no. (%) recovered from 91 Cryptosporidium-positive stool specimens.

Thirty-two episodes of recurrent symptomatic and asymptomatic cryptosporidiosis were diagnosed, not including the initial episode. Twenty-six episodes of recurrent Cryptosporidium diarrhea were detected in 17 children (7.5% [29.3% of 58 with prior Cryptosporidium diarrhea]); 12 had 1 recurrent episode, 2 had 2 recurrent episodes, 2 had 3 recurrent episodes, and 1 had 4 recurrent episodes. Six children had recurrent asymptomatic infection, for a total of 13 episodes. Children with symptomatic infection did not differ from children with asymptomatic infection with respect to age at enrollment (47.9 vs. 50.2 months) or age at the time of Cryptosporidium diagnosis (76 vs. 73.3 months).

Among the children with symptomatic infection, 57 (75%) presented within the first day of illness, which is a possible indication of the severity of diarrhea; almost all (56 [96.6%]) presented by the third day of diarrhea. More than half (34 [58.6%]) had abdominal pain, and 18.6% had vomiting at presentation. Using the Ruuska score of disease severity, 40% of children with cryptosporidiosis had a score ≥6, indicating that most children had a disease severity similar to that of other watery, nonrotavirus causes of diarrhea.

For the genetic analysis, 78 unrelated children with Cryptosporidium infection (i.e., cases) were compared to 104 children without infection (i.e., controls). For the HLA class I alleles, 2 of these 182 individuals did not have sufficient DNA for complete analysis and were not included. There were 20 class I or II alleles with an allele frequency >10% in either the cases or the controls (table 2). For the HLA class I region, 1 allele (HLA B*15) showed statistically significant associations with Cryptosporidium infection (OR, 2.16; P = .04). For HLA class II, children with Cryptosporidium infection were nearly 3 times as likely than controls to carry the DQB1*0301 allele (OR, 2.75; P = .005 [P_corrected = .025]).

Table 2.

Association between human leukocyte antigen class I and class II alleles with Cryptosporidium infection among children with (cases) and children without (controls) infection.

Allele	Controls	Cases	OR (95% CI)	P
A*11	36 (17)	26 (16)	1.13 (0.60–2.12)	.70
A*2402	35 (17)	22 (14)	0.81 (0.42–1.58)	.54
A*3301	26 (12)	20 (13)	1.06 (0.53–2.14)	.87
A*6801	23 (11)	10 (6)	0.46 (0.20–1.06)	.07
B*15	16 (8)	23 (15)	2.2 (1.05–4.62)	.04
B*35	24 (12)	17 (11)	0.94 (0.45–1.98)	.88
B*40	28 (14)	24 (16)	1.15 (0.58–2.29)	.68
B*44	48 (23)	34 (22)	0.99 (0.50–1.93)	.97
B*52	23 (11)	16 (10)	0.89 (0.43–1.86)	.76
C*401	21 (10)	20 (13)	1.44 (0.71–2.97)	.31
C*701	40 (19)	36 (23)	1.17 (0.63–2.17)	.61
C*1202	25 (12)	18 (12)	0.86 (0.42–1.76)	.69
C*1502	18 (9)	21 (14)	1.77 (0.86–3.68)	.12
DQB1*0202	25 (12)	24 (15)	1.24 (0.64–2.42)	.52
DQB1*0301	17 (8)	26 (17)	2.75 (1.35–5.59)	.005a
DQB1*0501	22 (11)	10 (6)	0.62 (0.27–1.41)	.25
DQB1*0503	22 (11)	11 (8)	0.72 (0.33–1.57)	.41
DQB1*0601	54 (26)	35 (22)	0.89 (0.49–1.62)	.69
DRB1*0701	38 (18)	30 (19)	1.07 (0.58–1.97)	.82
DRB1*1501	69 (33)	47 (31)	1.18 (0.66–2.13)	.58

NOTE. Data are no. of alleles detected/no. of alleles analyzed (%), unless otherwise indicated. Only HLA alleles with an allele frequency of >10 were evaluated. CI, confidence interval; OR, odds ratio.

^aBonferroni-adjusted P = .025.

To determine whether these alleles were associated with symptomatic infection, we stratified our analysis into children with no infection, children with asymptomatic infection, children with symptomatic infection, and children with both asymptomatic and symptomatic infection (table 3). For both the HLA class I B*15 allele and the DQB1*0301 allele, the association persisted for children with at least 1 episode of asymptomatic Cryptosporidium infection. However, children with only symptomatic infection were not more likely to carry the B*15 and/or DQB1*0301 alleles.

Table 3.

Association between human leukocyte antigen (HLA) class I B*15 and class II DQB1*0301 alleles and symptomatic and/or asymptomatic Cryptosporidium infection.

HLA allele, infection type	Allele absent	Allele present	OR (95% CI)	P	Pcorrected
B*15
None	88	15	1.0
Symptomatic only	27	7	1.52 (0.56–4.11)	.41	NA
Asymptomatic only	21	10	2.79 (1.10–7.09)	.03	.15
Symptomatic and asymptomatic	8	4	2.93 (0.78–11.0)	.11	NA
DQB1*0301
None	88	16	1.0
Symptomatic only	25	9	1.98 (0.78–5.02)	.13	NA
Asymptomatic only	20	12	3.3 (1.35–5.02)	.009	.045
Symptomatic and asymptomatic	7	5	3.92 (1.11–13.9)	.034	.15

NOTE. CI, confidence interval; OR, odds ratio.

In addition, children who had both asymptomatic and symptomatic infection appeared to share a haplotype with DRB1*1101 (DQB1*0301/DRB1*1101; OR, 7; P = .008 [P_corrected = .04]). Associations with other HLA class II alleles were not found, including DQB1*0601 and DRBq*1501, which have been previously described in this cohort to be associated with E. histolytica infection [11].

DISCUSSION

The evaluation of a large, prospectively followed cohort of children offered a unique opportunity to evaluate the role of host genetics in susceptibility to both symptomatic and asymptomatic Cryptosporidium infection. We hypothesized that the HLA class I or II alleles could be an important component of the human immune response to Cryptosporidium infection. The most important findings of our work is that children with Cryptosporidium infection are more likely to carry the HLA class II DQB1*0301 allele (OR, 2.75) and the class I B*15 allele (OR, 2.16). These associations were strongest in children with asymptomatic infection (DQB1*0301: OR, 3.3; B*15: OR, 2.78) and in children with both symptomatic and asymptomatic infection (DQB1*0301: OR, 3.92; B*15: OR, 3.2). However, children with only symptomatic infection associated with diarrhea were not associated with these alleles. This suggests a role for these alleles in the susceptibility to asymptomatic infection. Our cohort also demonstrated an association between Cryptosporidium infection (both symptomatic and asymptomatic) and both alleles of the HLA II *0301/DRB1*1101 haplotype (OR, 7; P = .008 [P_corrected = .04]). However, the number of children with this haplotype is small in our cohort (n = 20), and further work will be needed to determine its clinical significance in Cryptosporidium infection.

The human immune response to Cryptosporidium infection is incompletely understood, and the severity of clinical disease varies widely by age and immune status [12]. Nevertheless, the role of cellular immunity, particularly the CD4⁺ T cell, in protection from and resolution of disease has been demonstrated by the presence of cryptosporidiosis in individuals with advanced AIDS and findings of CD4⁺ adoptive-transfer experiments in susceptible animal models [4, 5, 13]. The HLA-encoded MHC II molecules are equally essential to present Cryptosporidium antigens to naive CD4⁺ T cells. As anticipated, the severity of Cryptosporidium disease is enhanced in MHC class II–deficient mice, which are incapable of presenting antigen and activating CD4⁺ T cells [13].

Increased allelic frequencies of HLA II DQB1*0301 have been associated with other intracellular pathogens, particularly the hepatitis viruses. In hepatitis C virus (HCV) infection involving both ethnically homogenous and mixed populations, DQB1*0301 is associated with susceptibility to infection and with clearance of circulating virus [14, 15]. In contrast, DQB1*0301 is associated with persistent infection with hepatitis B virus in a population of injection drug users [16]. Of interest, clearance of HCV infection is also associated with DRB1*11 in linkage disequilibrium with DQB1*0301; the protective roles of these alleles in HCV has been confirmed by a recent meta-analysis [14, 15].

The additional finding of an association between Cryptosporidium infection and the HLA class I B*15 allele suggests that the human immune response involves the interaction of additional lymphocyte populations in addition to the role of CD4⁺ T cells. Further work on human populations is needed to understand the significance of the HLA class IB association, including the additional role of CD8⁺ T cells, natural killer (NK) cells, and other atypical lymphocyte populations (e.g., NK T cells and γδ T cells). In addition, genetic analysis of other components of the immune response (e.g., single nucleotide polymorphisms of the cytokine genes thought to be involved in susceptibility to or control of Cryptosporidium infection) is needed. Further advances in high-throughput analyses of these genes, as well as gene chip assays to assess actual gene expression, will be essential to more completely understand the role of these genes in Cryptosporidium infection. Larger samples sizes will also be needed to increase the power of our observations.

By virtue of the size and follow-up duration of this cohort, our work also adds to data on the community impact of Cryptosporidium infection. In particular, it highlights the differences in disease expression between preschool-aged and grade school–aged children and the more severe and often persistent disease found in infants and toddlers [17, 18]. Although the children were all beyond the age of highest incidence of infection (≤2 years of age) at enrollment, infection was still extremely common. During the first 3 years of the study, 43% of the children had an episode of Cryptosporidium infection. More than 25% of children had cryptosporidiosis (i.e., Cryptosporidium diarrhea), and ~23% had asymptomatic disease, which was diagnosed by detection of antigen in surveillance stool specimens. Recurrent episodes of cryptosporidiosis were found in 29.3% of children (7.5% of the total population). Unlike studies involving younger children, the incidence of infection was similar across age groups in this cohort [19, 20]. Abdominal pain in more than half of children with cryptosporidiosis and a short duration (i.e., 3 days) of watery diarrhea characterizes cryptosporidiosis in this population. Persistent and chronic diarrhea, a high severity of disease (evaluated by the Ruuska score), and severe dehydration were rarely seen.

Although we were fortunate to work with an exceptionally well-followed cohort (children have been observed every other day for >6 years at the time of writing), our work has several important limitations. First, children with Cryptosporidium were compared to control children, all of whom had known infection or exposure to another protozoan, E. histolytica, and some of whom had coinfections with other enteric pathogens. Comparison to a control group of children with no enteric infections or coinfections, although difficult in this population, would help confirm our data. Second, our diagnostic work on asymptomatic infections was limited to antigen testing of frozen surveillance stool aliquots, which may have underestimated the prevalence of asymptomatic infection in our cohort. Future work will benefit from recently developed PCR-based diagnostic techniques and more-frequent surveillance. Finally, our study evaluated only children >2 years of age and cannot address the impact of Cryptosporidium infection in younger members of this population. This potential confounding factor can only be eliminated by further work involving close observation of a birth cohort of children to detect their first Cryptosporidium infection. These data are necessary to clarify whether infection at a very young age has a protective effect against future infections and to confirm the association between HLA alleles and infection. Nevertheless, this work is the first human data to identify HLA genetic differences in Cryptosporidium infection that may explain clinical disease variability, as well as a new avenue for identifying the immunodominant antigens of Cryptosporidium organisms.