Lack of support for an association between CLEC4M homozygosity and protection against SARS coronavirus infection

To the Editor:

In the January 2006 issue, Chan et al.¹ reported a significant association between severe acute respiratory syndrome (SARS) and a variable number of tandem repeats (VNTR) polymorphism in exon 4 of CLEC4M in a collection of individuals from Hong Kong. CLEC4M encodes L-SIGN ('liver/lymph node-specific ICAM-3 grabbing nonintegrin'), which serves as a receptor for many viruses, including SARS coronavirus (CoV)². Individuals homozygous for CLEC4M tandem repeats were reported to be less susceptible to SARS CoV infection. The authors also showed that cells homozygous for CLEC4M repeats had a higher binding capacity for SARS CoV, higher proteasome-dependent viral degradation and a lower capacity for trans infection. Thus, both genetic and functional studies suggested that homozygosity for CLEC4M was associated with protection against SARS CoV infection.

It is important to bear in mind that association studies require replication in independent populations³. We therefore attempted to replicate the findings of Chan et al. by genotyping the VNTR polymorphism in three additional collections of case-control samples from northern China: (i) the 'Beijing community population', consisting of 339 individuals with SARS and 227 random controls recruited from the community⁴; (ii) the 'Beijing health care worker (HCW) population', consisting of 42 health care workers infected with SARS during the course of hospital duty and 40 health care workers who had worked in SARS wards but remained free of disease and were confirmed to be seronegative for SARS⁵ and (iii) the 'Tianjin population', consisting of 60 individuals with SARS and 129 disease-free controls (including 85 random controls and 44 health care workers)⁶. The three collections of case-control samples and their ascertainment criteria have been described in detail previously (Supplementary Methods online)^4,5,6. All groups except the individuals with SARS from the Beijing community were in Hardy-Weinberg equilibrium. We found no significant differences in allele, genotype and homozygote or heterozygote frequencies between affected individuals and controls in the three populations (Table 1 and Supplementary Table 1 online). Early reports have shown that some nongenetic factors, such as comorbid conditions (including diabetes mellitus, hypertension, heart disease, tuberculosis, asthma and malignancy), are risk factors for the development of SARS^7,8 and may confound the contribution of genetic factors to this disorder. However, after stratification by comorbid conditions, the association remained nonsignificant in our Beijing community population (Table 1). Of the 287 affected individuals without comorbid conditions, 19 were individuals with severe SARS who were identified by their admission to intensive care units or by their death, and the remaining 268 were individuals with mildly symptomatic SARS. To account for this, we assessed whether there was an association between homozygosity for CLEC4M tandem repeats and severity of SARS, but we found none (Table 1). This result may be due to the limited number of individuals with severe SARS in the current study and will require confirmation in additional studies.

Table 1 Homozygote or heterozygote frequencies of the CLEC4M polymorphism in samples from northern China

Collectively, the results in our three collections of case-control samples from northern China are not supportive of the findings of significant association between the VNTR polymorphism and SARS risk reported by Chan et al.¹. There are several possible reasons for the inconsistent results. First, inadequate power may be an explanation of our negative results. However, the sample size in our Beijing community population (with 339 affected individuals and 227 controls) is approximately similar to that used by Chan et al. (with 285 affected individuals and 380 random controls), and this sample size had power >0.94 to replicate the effects by Chan et al. (calculated by the genetic power calculator⁹). Additionally, just by using our Beijing community sample set, we have successfully confirmed the positive associations between the mannose-binding lectin polymorphisms and SARS risk⁴ that were observed previously in a case-control sample from Hong Kong¹⁰. Furthermore, the consistency of the negative associations in our Beijing HCW population and our Tianjin population strengthens our results.

Second, there may be a small, population-specific difference in the contribution of CLEC4M polymorphism to SARS susceptibility. This might occur if there were population differences in linkage disequilibrium pattern or allele frequencies of CLEC4M. Indeed, the homozygote or heterozygote frequencies in individuals with SARS in our Beijing community or Beijing HCW populations were significantly different from those reported for affected individuals in Hong Kong community or Hong Kong HCW populations (P = 0.0066 and P = 0.017, respectively; Supplementary Table 1). Additionally, the difference in allele frequency between Hong Kong outpatient controls and Tianjin random controls and the differences in homozygote or heterozygote frequencies between Hong Kong HCW controls and Tianjin HCW controls were also significant (P = 0.0064 and P = 0.017, respectively; Supplementary Table 1). Furthermore, there was also a significant difference in the genotype and homozygote or heterozygote frequency between the Chinese population and Europeans¹.

Another possibility is that Chan et al. unwittingly neglected to account for the potential confounding factors that may distort the contribution of CLEC4M VNTR polymorphism to SARS susceptibility. Although Chan et al. took stringent precautions to stratify their samples by 'health care worker', population samples may still differ by many other factors that depend on setting and context of recruitment, such as age of the subjects at SARS onset, sex of the subjects and any comorbid conditions. Unfortunately, Chan et al. do not provide data with regard to such important information on confounding factors. Last, the initial findings of Chan et al. may not represent real associations and might be false positives. In genetic association studies of common diseases, there is a very low prior probability of detecting a true association result when accounting for statistical adjustment for multiple comparisons. Indeed, the genetic associations presented by Chan et al. were marginally statistically significant (P = 0.027, 0.045 and 0.031, when comparing all SARS samples to random controls, community SARS to outpatient controls and HCW SARS to HCW controls, respectively).

Thus, we did not find any significant differences in allele, genotype and homozygote or heterozygote frequencies between cases and controls in our three independent populations of northern Chinese. Although the biological plausibility of L-SIGN and the functional evidence of the VNTR polymorphism in the original report remain interesting, we urge that the association between CLEC4M polymorphism and SARS be investigated in other subpopulations of ethnic Chinese origin (for example, Taiwanese or Guangdong Chinese) or in those of different ancestry, such as Europeans.

Note: Supplementary information is available on the Nature Genetics website.