arising from Stravalaci et al. Nature Immunology (2022)

A recent study demonstrated that the complement recognition protein mannose-binding lectin (MBL) can bind to glycosylated SARS-CoV-2 spike protein. Binding activated complement in a spike-dependent manner and inhibited severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in in vitro models1. Furthermore, genetic variants near or in the MBL2 gene that encodes MBL in humans were suggested as being associated with COVID-19 requiring hospitalization.

In the present study, we carried out genetic and biochemical analyses of MBL activity in plasma in a multicenter cohort of critically ill COVID-19 patients and interrogated the publicly available summary statistics from the COVID-19 Human Genetics Initiative (COVID-19 HGI)2. We found no evidence that genetic variants that determine activity of the MBL pathway are associated with hospitalization or intensive care admission due to SARS-CoV-2 infection. Instead, we demonstrated that MBL2 haplotypes determine risk for thrombotic complications in critically ill COVID-19 patients. Specifically, genetically determined MBL activity confers risk for pulmonary embolism in a U-shaped manner, where haplotypes associated with intermediate MBL activity are protective. Our results demonstrate a complement-dependent mechanism for COVID-19-associated thrombosis and provide an example of how genetic variation in the innate immune system modulates thrombosis risk.

Patients, 426, admitted to the intensive care units at 2 tertiary hospitals in Uppsala and Stockholm, Sweden, were prospectively included in the SweCovid initiative3 between March 2020 and September 2021. Demographics and comorbidities of the cohort are shown in Supplementary Table 1. We analyzed MBL activity in plasma and used whole-genome genotyping data to impute genetic variants and construct haplotypes encompassing the MBL2 gene.

MBL activity in plasma displays extensive interindividual variability, which is largely determined by genetic variants located in exon 1 of the MBL2 gene or upstream of the transcription start site4. We imputed the genotypes of five genetic variants that, due to linkage disequilibrium, form six common MBL2 haplotypes (Table 1, rows 1–5). These include two high-expressing haplotypes (legacy names HYA and LYA) and one low-expressing haplotype (legacy name LXA), where ‘A’ indicates that these haplotypes express wild-type MBL protein. In addition, three commonly occurring missense variants in the collagen-like domain of MBL (legacy names D, B and C; haplotypes HYD, LYB and LYC) lead to amino-acid substitutions that impair MBL multimerization in a dominant-negative manner, resulting in monomeric or oligomeric MBL with little or no lectin activity. These variants result in complete MBL deficiency in the homozygous state or in combination with the LXA haplotype, but show moderate MBL activity in combination with the high-expressing HYA and LYA haplotypes5. In agreement with previous studies, these MBL2 haplotypes showed a strong genotype–phenotype correlation and predicted MBL activity in plasma (Fig. 1a). Allele frequencies of the individual genetic variants did not differ from the general European population in our cohort, which has predominantly European ancestry (Table 1, rows 1–5). In contrast, Stravalaci et al.1 demonstrated a moderate association between the D allele (rs5030737-A), or a combination of missense variants, and risk of COVID-19 requiring hospitalization in an Italian cohort of 332 cases and 1,668 controls. As the D allele is a loss-of-function variant, it was suggested that MBL has a protective effect against severe COVID-19. However, the B and C variants (rs1800450-T and rs1800451-T) share a common biological effect with the D variant, and would therefore be expected to give rise to similar signals in association studies if MBL loss of function were associated with COVID-19 risk.

Table 1 Candidate genetic variants in MBL2 locus and their association with COVID-19 outcome in the COVID-19 HGI (release 6)
Fig. 1: MBL2 haplotypes predict MBL activity in plasma and thromboembolic complications during intensive care.
figure 1

a, MBL activity across individual haplotype combinations. MBL activity in plasma at day 1 of intensive care was measured by a functional ELISA using a mannan matrix for MBL capture. Activity is expressed as a percentage of pooled normal human serum (NHS). The genetic variants shown in Table 1, rows 1–5 give rise to six major MBL2 haplotypes, which are denoted by their legacy names: HYA, LYA, LXA, HYD, LYB and LYC. The HYA, LYA and LXA haplotypes encode wild-type MBL protein and are referred to as ‘A’ haplotypes. The HYD, LYB and LYC haplotypes contain the missense variants referred to as the D, B or C alleles, respectively, and are shown together as ‘0’ haplotypes because they share a common deleterious effect on MBL activity. b, Haplotypes categorized according to MBL activity into high (HYA/HYA, LYA/LYA, HYA/LYA, HYA/LXA, LYA/LXA), intermediate (LXA/LXA, HYA/0, LYA/0) and deficient (LXA/0, 0/0). MBL activity differed significantly between the groups (230% (168–400, median and interquartile range) for the high group, 40% (24–70) for the intermediate group and 0.56% (0.00–1.25) for the deficient group; P < 0.0001). The horizontal bars indicate the median. c, Forest plot displaying ORs with 95% confidence intervals (CIs) for thromboembolic events, renal replacement therapy, need for mechanical ventilation and 90-day survival. Primary pulmonary thrombosis comprises events without a known concurrent distal venous thrombosis. A logistic regression model with MBL2 intermediate activity haplotypes, sex and age as covariates was used. *P < 0.05, **P < 0.01, **** P < 0.0001.

We further interrogated these variants in COVID-19 HGI release 6, the largest meta-analysis of genetic association studies on COVID-19 to date2. We found no effect of the D allele on either hospitalization or care requiring more respiratory support than supplemental oxygen (Table 1). Similar results were obtained for the B allele, whereas the C allele was nominally significant for the latter outcome, but with a modest odds ratio (OR) that did not pass multiple testing correction. These results show that genetic variants that impair MBL function do not substantially increase the risk of COVID-19 requiring hospitalization or advanced respiratory support.

Stravalaci et al.1 also conducted an extended analysis of a 1-MB region encompassing the MBL2 locus and found several candidate genetic variants for COVID-19 risk. Interrogation of COVID-19 HGI revealed that none of these variants was associated with COVID-19 requiring hospitalization (Table 1, rows 6–11). The G allele at rs74974397 was nominally significant for COVID-19 requiring advanced respiratory support, but did not pass multiple testing correction. We also analyzed these variants in the SweCovid cohort. Allele frequencies did not differ from the general population (Table 1, rows 6–11) and we did not observe a significant effect on MBL activity in plasma for any of these variants (P > 0.05).

Even though MBL does not affect the risk of hospitalization or severity on SARS-CoV-2 infection, it could still modulate the clinical phenotype. We categorized the individual MBL2 haplotypes according to MBL activity into three groups (high activity, intermediate activity and deficient; Fig. 1b) and analyzed their association with outcome in the SweCovid cohort. MBL2 haplotype groups were not associated with either the need for invasive ventilation or renal replacement therapy or 90-day survival (Supplementary Table 2). Instead, we observed a strong association with thrombotic complications and pulmonary embolism. Unexpectedly, the association was U shaped, with intermediate MBL2 activity haplotypes being protective (Fig. 1c and Supplementary Table 2). Pulmonary embolism is a major complication among hospitalized COVID-19 patients6 and may be the result of a primary pulmonary thrombosis without a coexisting distal venous thrombosis7. The protective effect of intermediate MBL2 haplotypes was even more pronounced in this group, suggesting that MBL primarily influences thrombotic reactions in the lung.

The hypercoagulable state in severe COVID-19 is well recognized and characterized by widespread fibrin deposition in lung capillaries and elevated plasma D-dimer levels8,9,10. Biomarker analysis showed a nonsignificant trend toward lower D-dimer levels in the intermediate MBL2 haplotype group, as well as significantly higher levels of the coagulation inhibitor antithrombin, which together indicate a lower degree of coagulation activation (Supplementary Table 3). Despite conferring protection from thrombosis, intermediate MBL2 haplotypes were not associated with increased survival, in line with observations that pulmonary embolisms do not contribute significantly to COVID-19 mortality if managed appropriately6,11.

Is a U-shaped association between MBL activity and thrombosis biologically plausible? Deficiency in early complement components is associated with defective clearance of cellular debris and apoptotic cells12,13. Given the data presented by Stravalaci et al.1, MBL recognition of spike protein expressed on SARS-CoV-2-infected cells could assist in their clearance, thus removing an important source of potentially thrombogenic material from the intravascular space. In contrast, in the context of haplotypes associated with high MBL activity, SARS-CoV-2-infected cells could instead provide an abundant substrate for MBL-dependent complement activation. The prothrombotic effects of excessive complement activation are well recognized and stem from complement-mediated endothelial damage and induction of tissue factor expression on monocytes14. In this scenario, intermediate MBL levels may provide sufficient clearance of thrombogenic material, and yet will not lead to bystander cell damage and cell activation due to unconstrained complement activation. In support of this concept, a U-shaped association between MBL and thrombosis risk has been observed in the case of cardiovascular disease and diabetes, where intermediate MBL levels were protective15.

A challenge in host genetic studies is the plethora of factors that can modulate genetic risk. As the COVID-19 pandemic has unfolded, changes in treatment strategies and virus variants have followed that may contribute to differences in effect size of genetic associations. In this context it should be noted that the SweCovid cohort covers the first three major pandemic waves corresponding to the original SARS-CoV-2 variant as well as the alpha and delta variants of concern. Likewise, COVID-19 HGI, release 6 represents a world-wide meta-analysis of COVID-19 studies from the start of the pandemic until 30 April 2021, thus taking into account differences in hospital systems and geographical locations.

The unique polymorphic nature of the MBL2 locus, with a high frequency of naturally occurring knockouts devoid of functional MBL, allows for an assessment of the role of MBL activity in observational studies. We show that genetically determined MBL deficiency does not influence the risk of COVID-19 requiring hospitalization or admission to intensive care. Instead, our results demonstrate that the interaction between genetic variation in the MBL2 gene and SARS-CoV-2 infection determines thrombosis risk in critically ill patients, a finding with important clinical implications given the high rate of thrombotic complications in COVID-19 patients.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.