Iron is considered an essential nutrient for both humans and pathogenic microbes. Iron limitation defends pathogenic microbes as a key form of innate immune, termed “nutritional immunity” (1, 2). Interestingly, iron overload-associated diseases, such as hereditary hemochromatosis and β-thalassemia, were also found increased infection susceptibility (1, 3). Recently, several MR studies also gave evidence of a significant effect of genetically high iron status on the susceptibility of skin and soft tissue infections.

Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection (4). Although knowledge about sepsis is growing all the time, it is a depressing fact that current efforts to develop new therapies have not been successful (5). Early antibiotic therapy was a widely accepted treatment for sepsis, which benefitted patients by improving survival prognosis (6–8). The above references suggest that early intervention is of great significance. The general consensus is that early identification of risk factors is essential for the early intervention of sepsis.

Several observational studies showed that patients with sepsis had significantly elevated levels of serum iron, ferritin, and transferrin saturation, but transferrin level was in the opposite direction (9–11). Nevertheless, the evidence from conventional observational studies always remains vulnerable to the threat of confounding and reverse causation. Thus, the direction of causality between iron status and sepsis remains unclear. As an epidemiological approach, mendelian randomization (MR) uses genetic variants as instrumental variables (IVs) to infer expose-outcome causation from observational data, which is closely analogous to the clinical randomized controlled trials (RCTs). This also presents an opportunity to promote drug repurposing (12). Based on Mendel's laws of independent assortment and segregation, each allele is randomly transmitted to the offspring with an equal probability, which is exploited as a way of randomizing participants into different levels of the iron status traits (13, 14). Besides, the unidirectional flow of biological information from the parental generation to progeny can limit the effects of reverse causation (15). Herein, we designed a two-sample bidirectional MR study to identify the direction of causal effects between seven iron status traits and sepsis, using available public data from a large-scale genome-wide association study (GWAS) for sepsis and two GWAS meta-analyses for iron status traits.

Independent SNPs were selected as genetic instrumental variables based on independent and LD analyses, including two serum iron-associated SNPs, three independent ferritin-associated SNPs, three transferrin saturation-associated SNPs, and eight transferrin-associated SNPs (Supplementary Tables 2, 3).

The primary MR analyses by fixed-effects IVW showed a significant causal effect of serum iron, ferritin, and transferrin saturation on the risk of sepsis (Table 1 and Figure 1). The results showed that the genetically predicted high levels of serum iron (OR = 1.21, 95%CI = 1.13–1.29, p = 3.16 × 10⁻⁴), ferritin (OR = 1.32, 95%CI = 1.07–1.62, p = 0.009), and transferrin saturation (OR = 1.14, 95%CI = 1.06–1.23, p = 5.43 × 10⁻⁴) were associated with an increased risk of sepsis. Results were consistent across random effect IVW and WM (Table 1). The above results were consistent and robust in all sensitivity analyses, and no heterogeneity and pleiotropy were observed (Table 1).

Conversely, low transferrin levels were significantly associated with increased risk of sepsis (OR = 0.94, 95%CI = 0.86–1.02, p = 0.023). Nevertheless, substantial heterogeneity among estimates of individual SNPs was detected in the analysis of Transferrin levels (IVW, Cochran's Q Statistic = 16.85, I² = 58.45%, p = 0.018; MR-Egger, Cochran's Q Statistic = 14.94, I² = 59.83%, p = 0.021). Thus, the random effect IVW model was employed, of which the results showed no significant association (OR = 0.94, 95%CI = 0.89–0.99, p = 0.142).

Single-SNP MR analyses showed that SNPs rs1800562 (HFE) and rs855791 (TMPRSS6) were associated with a significant effect of serum iron or transferrin saturation on sepsis, while SNPs rs1800562 (HFE) and rs2413450 (TMPRSS6) were significantly related to the causal effect of ferritin on sepsis (Figure 2). Using the leave-one-out analyses, we found that one independent SNP (rs1800562) could drive a significant effect of ferritin on sepsis. The significant associations of ferritin and sepsis did not remain after removing rs1800562 (Supplementary Figure 1).

The second set of MR analyses aimed to assess potential causal effects of sepsis on iron status (Supplementary Tables 4, 5). Results showed no significant causal relationship between sepsis and iron status (Table 2 and Supplementary Figure 2). As described in Table 2, we observed high statistical heterogeneity in the fixed-effects IVW MR analysis of transferrin saturation and transferrin, therefore, the random-effect IVW model was adopted as a sensitivity analysis, of which the results still showed no significant association.

Iron is needed mainly for hemoglobin and erythrocytes production, therefore, we wondered whether the levels of hemoglobin, erythrocytes, and reticulocytes could affect the risk of sepsis and whether sepsis could influence the level of this biomarker. The results of MR analysis suggested no statistically significant effects on whether the sepsis served as exposure or outcome, which agreed with the results of sensitivity analyses (Supplementary Figures 3, 7 and Supplementary Table 6). A random-effects model was applied to estimate the causal effects of sepsis on the above three biomarkers due to the high heterogeneity level. Moreover, no pleiotropy was found in all analyses.

In this bidirectional MR study, we investigated the association of iron status biomarkers with the risk of sepsis, of which genetic variants that proxy the effect were identified from publicly available large-scale GWAS data. Our findings showed evidence that genetic predisposition to higher levels of serum iron, ferritin, and transferrin saturation was causally associated with a higher risk of sepsis. Genetic predisposition to sepsis did not appear to influence any of the studied iron status markers concentration (serum iron, ferritin, transferrin saturation, and transferrin).

In contrast with the previous MR studies of iron status (29, 30), the number of IVs in our study was more conservative. Previous research confirmed that rs1800562 and rs855791 were the primary determinants of variation associated with HFE and TMPRSS6 in serum iron status (31, 32). Searching the GWAS Catalog and PubMed, we found that rs1800562 and rs1495741 were associated with low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC), or triglyceride (30, 33), but there is currently no evidence to confirm the above lipid metabolism biomarkers as risk factors for sepsis susceptibility (34, 35).

The results from this MR study were fundamentally consistent with recent observational studies (9–11), except for transferrin. Increased ferritin and serum iron concentrations were also associated with greater severity (36) and a poor prognosis (37, 38) for sepsis. As an important form of host defense, nutritional immunity presents with limiting the availability of iron to pathogens (1, 39, 40). One such example is anemia of inflammation (41), but it does not contradict our idea that iron status parameters were causally positively associated with the risk of sepsis. High iron status may enhance susceptibility for sepsis via dysregulated host-pathogens interaction (39), which is a disorder of iron homeostasis. An important fact from the perspective of pathogens is that iron accumulation favors the growth and survival of the pathogens on the host (42–44). Besides, from the host perspective, high iron status could affect the innate immune functions and promotes the synthesis and release of inflammatory factors (2, 3, 45). The evidence of pre-clinical studies supported iron chelators as a potential therapeutic agent of sepsis (46–49). Therefore, targeting high iron status was a potential therapeutics and promising direction for the prevention and treatment of sepsis.

There are several limitations that we could not overlook in this study. First, population limited to European ancestry hampered the generalization of findings to individuals of other ancestries. Second, we could not determine whether there were non-linear between iron status biomarkers levels and the risk of sepsis, such as threshold effect or U-shaped relationship. Third, this study did not adequately investigate all indicators for the assessment of human iron status traits, such as hemoglobin and transferrin receptors.

In summary, this present bidirectional MR analysis suggested the causal association of high iron status with an increased risk of sepsis, while the reverse causality hypothesis did not hold. The levels of transferrin, hemoglobin, erythrocytes, and reticulocytes were not significantly associated with sepsis. Future studies are needed to explore the exact mechanism and confirm the potential clinical value of such a prevention and treatment strategy.

This data can be found at: Publicly available datasets were analyzed in this MR study. GWAS summary-level datasets were downloaded from IEU OpenGWAS Project (https://gwas.mrcieu.ac.uk/), including serum iron (GWAS ID: ieu-a-1049), transferrin (GWAS ID: ieu-a-1052), transferrin saturation (GWAS ID: ieu-a-1051), ferritin (GWAS ID: ieu-a-1050), hemoglobin (GWAS ID: ebi-a-GCST004615), erythrocytes count (GWAS ID: ebi-a-GCST004601), reticulocytes count (GWAS ID: ebi-a-GCST004622), and sepsis (GWAS ID: ieu-b-69).

All authors contributed to the design, interpretation, and writing of the paper.

This work was supported by Shandong Provincial Natural Science Foundation, China (ZR2020MH392), Shandong Medical and Health Technology Development Project (2018WS193), and Qilu Wellness and Health Leading Talent Training Project in 2020.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.