Dear Editors,
We read with interest the bidirectional two-sample Mendelian randomization (MR) study by Guo et al.,1 which reported that genetically predicted systemic lupus erythematosus (SLE) was associated with an increased risk of vitamin B12-deficiency anemia (B12DA; inverse-variance-weighted (IVW) odds ratio (OR) = 1.08, 95% confidence interval (CI) 1.04–1.13, P = 2.3 × 10−4), while genetically predicted B12DA was associated with an increased risk of SLE (IVW OR = 2.22, 95% CI 1.93–2.55, P = 1.6 × 10−29).1 The purpose of this letter is to examine, direction by direction, whether extended major histocompatibility complex (MHC) pleiotropy can plausibly account for these associations, and to frame this explicitly as an instrument-level sensitivity consideration rather than an established mechanism.
The 8.1 ancestral haplotype (AH8.1; HLA-A*01:01–B*08:01–C4A-null–DRB1*03:01–DQA1*05:01–DQB1*02:01) has long been recognized as a major SLE risk unit characterized by dense linkage disequilibrium (LD) spanning classes I, II, and III.2,3 It is tempting to invoke AH8.1 pleiotropy to interpret SLE-related MR findings. However, Guo et al.1 selected 20 SLE exposure instruments genome-wide (LD-clumped at r² < 0.001 over windows > 10,000 kb). One of these, rs9262046, was itself later flagged as a potential outlier by the study’s own MR-PRESSO test (Supplementary Table 5)1; whether this or other instruments map to the extended MHC has not been independently confirmed from Supplementary Table 3 of Guo et al.,1 and we do not assert a specific count or percentage here. Because the original study did not report a sensitivity MR excluding MHC-region variants, we cannot determine how much the SLE→B12DA estimate depends on them. The forward MR-Egger intercept showed no evidence of directional pleiotropy (P = 0.832)1; stringent LD clumping was applied, and residual correlation should not be assumed. Establishing whether any MHC-linked variant materially influences the estimate requires a leave-one-out and MHC-excluded sensitivity reanalysis.
The study’s MR-PRESSO outlier-exclusion step also warrants careful interpretation.1 Guo et al.’s Supplementary Table 5 lists seven removed variants, including rs9262046 in the SLE→B12DA, with non-significant and partly identical P values (0.171–0.501).1 We did not classify the remaining outliers as MHC variants without a separately validated coordinate audit, and therefore do not claim an MHC-enriched outlier-removal pattern. More broadly, the original Supplementary Table 4 of Guo et al.1 reports an identical mean F-statistic of 71.14 for each of the five forward SLE-exposure analyses, despite different instrument counts and F-statistic ranges.1 This apparent discrepancy complicates interpretation of instrument strength and should be clarified by the authors. We do not replace the published values with reconstructed means from a fragmented supplementary table.
The reverse B12DA→SLE analysis relies on a distinct five-variant instrument set selected for B12DA.1 Because the forward and reverse analyses use different instruments, the specific SLE instruments used in the forward analysis cannot directly explain the reverse estimate; the reverse instrument set requires a separate pleiotropy assessment. Although the reverse IVW estimate is statistically significant (OR = 2.22, 95% CI 1.93–2.55, P = 1.6 × 10−29), statistical significance must not be conflated with evidence for causal direction. The Steiger test supported the forward direction (R²SLE = 0.105 vs. R²B12DA = 4.4 × 10−5) but did not support the reverse direction (R²B12DA = 0.00105 vs. R²SLE = 0.0101).1 With only five instruments, the nonsignificant MR-Egger estimate (P = 0.058) and intercept (P = 0.538) also have limited power.1 The reverse result should therefore be regarded as an unvalidated directional hypothesis requiring replication.
Several prior studies illustrate the internal complexity of MHC associations across ancestries. Table 1 summarizes the evidence layers relevant to this SLE–B12DA causal inference.1-11
| Evidence layer | Key finding | Status and relevance |
|---|
| Guo et al.’s forward instruments | 20 LD-clumped SLE instruments; rs9262046 is documented in the SLE→B12DA outlier table1 | Instrument-level MHC membership should be confirmed from the complete Supplementary Table 3 list |
| MHC/AH8.1 biology | AH8.1 is a high-LD SLE-associated haplotype spanning MHC classes I–III2,3 | Established background biology; not proof that AH8.1 drives this MR estimate |
| Ancestry-specific MHC mapping | European/African American LD architecture differs from Korean HLA/C4 fine-mapping findings4,5 | Established independent findings; transferability requires testing |
| Cross-disease MHC loci | SKIV2L and NOTCH4 have been implicated in shared autoimmune architecture6,7 | Plausible pleiotropy; relevance to the retained instruments remains untested |
| Outcome-side genetics | Pernicious-anemia genome-wide association study signal maps to the DR15 haplotype; DR3/DR4 associations are context-specific8,9,11 | Does not establish a shared SLE–B12DA DR3/DR4 pathway |
| Cellular observation | GZMK⁺GZMH⁺HLA-DR⁺ effector-memory CD8⁺ T-cell expansion reported in active SLE10 | Preliminary abstract-level observation; no demonstrated AH8.1 link |
| Summary-statistic quality control |
Supplementary Table 4 reports the same mean F-statistic (71.14) for five distinct forward analyses1 | Requires clarification before using these values for comparative power claims |
Hanscombe et al.4 reported ancestry-specific differences in SLE-associated MHC linkage disequilibrium structure in European and African American cohorts. Separately, Yu et al.5 reported independent HLA amino-acid and C4 copy-number signals in Korean cohorts. These findings should not be merged into a single ancestry-invariant model, and their transferability to Guo et al.’s European-ancestry instruments remains uncertain.
The specific bridge to B12DA requires outcome-side qualification. Autoimmune gastritis and pernicious anemia are possible autoimmune causes of B12 deficiency, but B12DA is not synonymous with pernicious anemia.8,9 Earlier literature has reported DR3/DR4 associations in selected autoimmune contexts,8,9 whereas the pernicious-anemia genome-wide association study by Laisk et al.11 identified its strongest MHC signal at the DR15 haplotype (HLA-DRB1*15:01–DQA1*01:02–DQB1*06:02). These observations do not establish shared DR3/DR4 horizontal pleiotropy between SLE and B12DA.
A conference abstract reported an expanded GZMK⁺GZMH⁺HLA-DR⁺ effector-memory CD8⁺ T-cell population in active SLE.10 Because the report did not genotype HLA alleles and is not a full-length peer-reviewed article, it does not establish that AH8.1 or HLA-DRB1*03:01 drives this cellular state; it only motivates future genotype-resolved studies.
Future analyses would benefit from the following: (1) a validated audit of the 20 SLE instruments, including chromosomal coordinates and HLA proxies, followed by leave-one-out and MHC-excluded MR analyses; (2) clarification of the identical mean F-statistics(71.14) reported for all five forward analyses despite different numbers of SNPs and R² values; (3) replication of the reverse B12DA→SLE direction with expanded instruments and robust directionality assessment; and (4) prospective evaluation of autoimmune and non-autoimmune causes of B12 deficiency in patients with SLE. These are research priorities, not recommendations for routine autoimmune-gastritis screening.
In conclusion, Guo et al.1 provide useful hypothesis-generating evidence, but the two MR directions involve distinct instrument architectures and require different interpretations. One SLE→B12DA instrument (rs9262046) is documented via the study’s own outlier table, whereas the complete retained -instrument composition should be confirmed from Guo et al.’s Supplementary Table 3 before broader MHC conclusions are drawn. Treating MHC pleiotropy as a sensitivity-analysis consideration, rather than an assumed explanation, will improve causal inference in future autoimmune MR studies.
Declarations
Acknowledgments
Perplexity AI was used for language support and text rearrangement.
Funding
The authors received no grants or other financial support for this work.
Conflict of interest
The authors declare no conflicts of interest.
Author contributions
Conceptualization (HR), writing – original draft (HR) revision in response to peer review (HR), and writing – review & editing (NB, HB, SI). All authors have approved the final version and publication of the manuscript.