Reply to “Lack of support for an association between CLEC4M homozygosity and protection against SARS coronavirus infection”

Chan et al. reply:

Tang et al. and Zhi et al. report that in contrast to our findings¹, they were unable to find association between homozygosity or heterozygosity of the CLEC4M (L-SIGN) exon 4 tandem repeat polymorphism and SARS CoV susceptibility in the Chinese population. Their data cannot conclusively negate our findings for the reasons below.

We agree with Tang et al. that the difference between their analysis and ours is largely accounted for by the difference in the percentage of homozygosity in controls. Tang's controls included neonatal cord blood, healthy elderly individuals aged >70 years, local university students and control samples collected in Beijing. The homozygosity and heterozygosity distribution of their Beijing controls is very similar to that of our Hong Kong random controls. However, the former three control groups collected in Hong Kong are poorly matched by age for comparison with individuals with SARS and our random controls (Supplementary Table 1 online), and such an age distribution clearly excludes the population aged 25–69 years, which makes up the largest proportion (>70%) of SARS-infected individuals during the Hong Kong outbreak in 2003 (ref. 2). Age and gender are well-known confounding factors in any case-control study, and for SARS, individuals aged ≥65 years and <18 years are actually associated with a lower risk³. Thus, it is surprising that Tang et al. chose these specific control groups at extreme age ranges, given that there are no obvious obstacles for recruiting appropriate age-matched controls in Hong Kong. They also failed to perform logistic regression accounting for age differences, which should have been incorporated in their statistical analysis. The age and gender distribution of the SARS and control populations in our study, in contrast, is more properly matched, and by logistic regression, our results remain statistically significant (Table 1 and Supplementary Table 1). Our results remain consistent after accounting for comorbid conditions (Table 1).

Table 1 Statistical analysis of CLEC4M homozygosity and heterozygosity adjusted for age, gender and comorbid conditions in Chan et al.¹

Tang et al. also used control groups differing widely in age to determine if age had any effect on genotype frequencies. However, their data per se already showed a significant difference in overall genotype distribution, when comparing neonates versus elderly individuals versus university students (P = 0.029, χ² test). The difference in genotype between neonates and university students is also significant (P = 0.009 by CLUMP), suggesting that age-related selection may exist for CLEC4M genotypes. Indeed, age-dependent variation of allele and genotype frequencies has been reported for other genes^4,5.

Tang et al. also claim to show a subpopulation difference in allele and genotype frequencies between northern and southern Chinese. They note that “the seven-repeat allele was more prevalent in the Beijing sample (0.7 in Beijing versus 0.64 in Hong Kong, P = 0.05), which also largely accounted for the higher proportion of homozygotes (55.7% in Beijing versus 46.0% in Hong Kong, P = 0.02).” It seems these quoted Hong Kong figures refer to frequencies from their cord blood group alone. Such a comparison is difficult to justify, again because age is seriously mismatched. It should be noted that there is no difference in either the seven-repeat allele frequency or the proportion of homozygotes between the Hong Kong random controls of Chan et al.¹ and the Beijing controls of Tang et al. (Supplementary Table 2 online). Barreiro et al.^6,7 investigated the CLEC4M homozygote and heterozygote distribution of different ethnic groups and reported that the proportion of homozygotes in East Asians, consisting predominantly of Chinese, was 53% (Supplementary Table 3 online), a figure similar to the Hong Kong random controls of Chan et al. and the Beijing controls of Tang et al. There is also no significant difference in allele or genotype frequencies of these two control populations from the East Asian samples of Barreiro et al.⁷ (Supplementary Table 2). When these results are taken together, little subpopulation structure is observed for CLEC4M homozygote and heterozygote distribution between the northern and southern Chinese populations.

Finally, given the apparent absence of subpopulation structure for CLEC4M homozygote and heterozygote distribution in the Chinese population, we performed a meta-analysis of our data set and that of Tang et al. by the Mantel-Haenszel test using all control groups that are in Hardy-Weinberg equilibrium: our random controls and outpatient controls (n = 670), and cord blood, healthy elderly individuals and Beijing controls (n = 827) of Tang et al. compared with all SARS cases (n = 462). The combined odds ratio (OR) is significant (combined OR = 0.786, 95% confidence interval (c.i.) = 0.637–0.972, P = 0.026), indicating that a reduced risk is still associated with homozygotes, even by the approach of Tang et al. that disregards the age effect.

Tang et al. also argued against our functional studies by citing two studies in which L-SIGN oligomerization assays were performed using recombinant protein without the cytoplasmic and transmembrane domains in a cell-free model. This argument is not valid because these findings cannot be extrapolated to our cell-based functional assays, which bear more biological significance for the study of membrane-bound molecules.

As for Zhi et al., their genotyping data are a cause for concern. When the observed and expected numbers are compared, deviation from Hardy-Weinberg equilibrium of their Beijing community SARS group (P = 0.0006) reflects overcalling of homozygotes and undercalling of heterozygotes containing five and seven repeats. Moreover, a significant difference is noted in the homozygote and heterozygote distribution between their Beijing and Tianjin populations: when the overall homozygote and heterozygote genotype of the Beijing cohort (that is, combining both affected individuals and controls) is compared with that of the Tianjin cohort, significantly more homozygotes are found in the Beijing population (P = 0.0031, Table 2). Since there is no difference between our Hong Kong random controls and the Beijing controls of Tang et al., differences would hardly be expected between the populations in Beijing and Tianjin, as both cities are in northern China and are only 60–70 miles apart geographically. It is unknown if genotyping of these two cohorts by Zhi et al. was conducted independently or all in one laboratory, but such a significantly different homozygote and heterozygote distribution between these two cohorts raises the possibility of genotyping error. The authors state that cross-validation by direct sequencing was performed in some cases, and they specifically mention confirming homozygous 'five repeat/five repeat' and 'seven repeat/seven repeat' genotypes by sequencing. Surprisingly, however, their heterozygous 'five repeat/seven repeat' genotypes were confirmed only by 2% gel electrophoresis. The distinction between homozygosity and heterozygosity should depend on precise recognition of two separate bands in an agarose gel and that the ultimate verification should rely on DNA blot analysis (which had been conducted in our study), rather than confirmation by sequencing of one specific band identified in electrophoresis of PCR products.

Table 2 CLEC4M homozygote and heterozygote distribution of Beijing and Tianjin cohorts (combining affected individuals and controls) in Zhi et al.

In summary, data from Tang et al. or Zhi et al. cannot conclusively negate our finding that CLEC4M homozygosity is associated with a reduced risk for SARS susceptibility. Although there may be population differences in CLEC4M genotype at the level of major ethnic groups⁶, little evidence exists for the presence of differences between northern and southern Chinese.

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