Reduced cyclin D3 expression in erythroid cells protects against malaria

Nature News · Feb 18, 2026 · Collected from RSS

Full Article

MainA long history of studies in human genetics has investigated how the powerful evolutionary pressure of endemic malaria has resulted in the selection of genetic variants that moderate the effect of the disease. In 1948, J. B. S. Haldane first suggested that heterozygous carriers of alleles that confer β-thalassaemia, a life-threatening anaemia at that time, could reduce mortality during malaria infection, and this ‘heterozygote advantage’ might explain the increased frequency of such alleles4.Since this initial proposal—commonly referred to as the malaria hypothesis—the increased frequencies, in areas where malaria was historically endemic, of not only β-thalassaemia, but also α-thalassaemia, sickle-cell anaemia (caused by the haemoglobin S (HbS) variant), haemoglobin C (HbC) disease, haemoglobin E (HbE) disease and ovalocytosis have all been hypothesized to be the result of selection5. Malaria resistance has also been proposed to account for the high frequency of alleles causing deficiency of the enzyme glucose-6-phosphate dehydrogenase (G6PD), and genetic studies have provided evidence for positive selection of the G6PD-deficiency A allele6.Broader genome-wide association studies (GWASs) have now identified additional DNA variants that are associated with the clinical severity of malaria7,8,9,10,11,12, some of which—such as variation in the ATP2B4 gene13—are also associated with features of red blood cells (RBCs), the cells that host key stages of the malaria parasite’s life cycle in a human host. However, the mechanisms by which these variants and genes can both affect RBC parameters and ameliorate malaria remain mostly unknown.Here we focus on explaining the molecular action of one allelic variant in the CCND3 gene region, in terms of its effects on both blood-cell traits and resistance to malaria. Our first finding was an association signal with increased levels of haemoglobin A2 (HbA2), marked by the derived T allele of single-nucleotide polymorphism (SNP) rs112233623 close to CCND3, the gene encoding the cyclin D3 protein1. Notably, this allelic variant showed a much higher frequency (minor allele frequency (MAF) = 10.1%) in a population cohort from Sardinia14, where the association was first detected, than in a sample set from northern Europe1 (MAF = 1%).Within the same gene region, the CHARGE consortium study (composed mainly of northern European individuals)15 had previously found an association signal with reduced mean corpuscular volume (MCV) and increased RBC count, marked by the derived A allele of SNP rs9349205, located 160 bp upstream of rs112233623. Following up on this genetic finding, another group experimentally demonstrated the erythroid-cell enhancer activity of the core region surrounding SNP rs9349205, and a further increase in enhancer activity specifically for rs9349205-A, compared with the baseline ancestral G allele16. The researchers also showed that Ccnd3-knockout mice had a decreased RBC count and increased MCV, effects opposite to those of rs9349205-A in humans16. After further examination of human and mouse primary erythroid cells, the team proposed that cyclin D3 affects the number of cell divisions during the terminal differentiation of erythroid precursors, and thereby influences the size and number of erythrocytes, such that a lower level of CCND3 leads to fewer but larger RBCs16.A more recent study in a large UK Biobank sample used statistical fine mapping in the CCND3 gene region and stepwise conditional analysis to identify rs9349205-A as the primary association for decreased RBC count and rs112233623-T as a putative independent associated variant with opposite effects2.In addition to its effects on the cell cycle, the cyclin D3–CDK6 complex has been found in cancer cells to phosphorylate and thereby reduce the activity of the rate-limiting glycolytic pathway enzymes 6-phosphofructokinase (PFK1) and pyruvate kinase M2 (PKM2)3. Consequently, glucose-derived carbon is shunted into the pentose phosphate and serine pathways, leading to increased formation of NADPH, the source of reducing power to neutralize ROS3. To our knowledge, no investigation has so far reported whether these metabolic effects in cancer cells are recapitulated in RBCs and their precursors, and, if so, whether they are modified by rs112233623-T and rs9349205-A.We therefore began this study with a series of related questions. How are the known actions of CCND3 related to the known erythrocyte trait effects of variant rs112233623-T? Does this allelic variant affect the known CCND3 erythroid enhancer, and, in particular, does it affect the binding of specific transcription factors? What effects are seen when rs112233623-T and rs9349205-A are inherited together on the same DNA molecule (haplotype)? Is the higher frequency of rs112233623-T in Sardinian individuals, compared with other European cohorts, more likely to be explained by random drift or positive selection during the evolutionary history of this population? And if selective mechanisms were involved, what was the selective pressure? In the present work, we address all of these queries, and we outline a role for reduced CCND3 function in malaria resistance and identify its probable mechanism. Our results suggest that inhibiting CCND3 could constitute a therapeutic modality for malaria.Genetic associations in the CCND3 enhancerExtending our previous work on the genetic architecture of haemoglobin levels in 6,305 individuals1, we examined genetic associations in the CCND3 gene region for a larger set of haematological traits in 6,824 individuals from the SardiNIA general population cohort (Methods). In contrast to findings in cohorts from northern Europe, where the association plots for the haematological traits RBC and MCV in this gene region are led by rs9349205-A (ref. 2), in Sardinia, rs112233623-T dominates the association profile, alongside another variant, rs113267280-G, that is in strong linkage disequilibrium (LD) with it (r2 = 0.96). These two variants form the credible set for HbA2, RBC MCV and mean corpuscular haemoglobin (MCH) associations (Methods, Extended Data Fig. 1 and Supplementary Table 1), and it is not possible to determine from the genetic association data which variant is primarily associated with these traits and which is secondarily associated owing to high LD. Still, there is evidence—such as a higher index of predicted deleterious effect for rs112233623-T than for rs113267280-G, using the combined annotation-dependent depletion (CADD) score (https://cadd.gs.washington.edu/score) (16.21 versus 2.27, respectively)—suggesting that rs112233623-T is the primarily associated variant, as was originally proposed for the HbA2 association1.Focusing on this SNP and integrating previously reported associations1,2, the derived T allele of rs112233623 is associated with increased HbA2 (P = 4.03 × 10−22, β = +0.31), MCV (P = 3.77 × 10−29, β = +0.36) and MCH (P = 2.48 × 10−24, β = +0.32), and decreased RBC count (P = 1.18 × 10−16, β = −0.27), as well as with increased fetal haemoglobin (HbF) (P = 4.32 × 10−03, β = +0.09) (Table 1 and Supplementary Information). The phenotypic effects on MCV and RBC count are similar to the haematological profile seen in Ccnd3-knockout mice16, suggesting that the association signal marked by rs112233623-T results from reduced CCND3 activity, probably owing to decreased expression.Table 1 Association results for the rs112233623-T variantFull size tableOf note, in the SardiNIA cohort, the derived allele A of rs9349205 associates with increased MCV (P = 8.8 × 10−5, β = +0.09) and MCH (P = 1.4 × 10−4, β = +0.09), and decreased RBC count (P = 1.6 × 10−2, β = −0.06). This is seemingly the same direction of effects as rs112233623-T, and opposite to the decrease in MCV and increase in RBC count previously reported for rs9349205-A in other populations2,15. However, the apparent contradiction can be resolved with conditional association analysis. After controlling for the effect of rs112233623-T, rs9349205-A was indeed associated with a reduction in MCV (P = 2.3 × 10−2, β = −0.05) and MCH (P = 3.3 × 10−2, β = −0.05), and an increased RBC count (P = 1.6 × 10−2, β = +0.06), consistent with findings in other populations and with increased CCND3 expression driven by rs9349205-A (refs. 16,17).Similarly, in single-variant analyses, we find that rs9349205-A is associated with increased HbA2 (P = 1.4 × 10−9, β = +0.14), again with the same direction of effect as rs112233623-T. However, after conditional association analyses accounting for the effect of rs112233623-T, the association of rs9349205-A with HbA2 completely disappeared (P = 0.64), in contrast to the opposite effects of rs112233623-T on RBC count, MCV and MCH traits. This suggests that, unlike the putative decrease of CCND3 expression due to rs112233623-T, which has a strong positive effect on HbA2 levels, the increase of CCND3 expression due to rs9349205-A alone might not have any primary effect on HbA2 levels.To better understand the contrasting association results for rs9349205-A and haematological traits in the SardiNIA cohort, compared with other European cohorts, and why conditional analysis adjusting for the effect of rs112233623-T renders the associations consistent, it is useful to examine the allele and haplotype frequencies of the two SNPs in various populations (for example, Sardinian individuals from the SardiNIA study cohort and European (EUR) individuals from the 1000 Genomes Project (1KG)18). The derived rs9349205-A allele is common both in SardiNIA and in 1KG EUR (frequency 21.8% and 24.4%, respectively), whereas the derived rs112233623-T allele is common in SardiNIA (frequency 10.1%) but much rarer in 1KG EUR, especially in the northern Europe subset (frequency 1%; Fig. 1). In Sardinian individuals, the two variants show considerable LD (r2 = 0.4, D′ = 1); the rs112233623-T allele is always found on an rs9349205-A background (Extended Data Fig.