Genome x Environment analysis of Sudden Unexpected Infant Death unveils etiologic heterogeneity and strong cannabis and genetic disease risks

Abstract

Sudden Unexpected Infant Death (SUID), the third leading cause of infant death, has increasing incidence and multifactorial etiology. Identification of preventative interventions has hitherto been hindered by etiologic studies limited to genetic or environmental effects in isolation. Here we report a multifactorial genome x environment analysis of SUID risk. Births in San Diego County California from 2005-2018 were linked to hospital discharge summaries and death files, yielding 212 SUID cases and 620,392 infants alive at age 1 year. Whole genome sequencing (WGS) identified probable and possible genetic etiologies in 16% and 48% of SUID cases, respectively. Genetic risks were extremely heterogeneous with 144 loci contributing 173 risks in 57% of SUID cases. Genetic risk was very strong (Prevalence Risk Ratio, PRR >99) or strong (PRR 3.7 - 99) in 12% and 34% of SUID cases, respectively. Six of sixteen significant environmental risks lost significance when SUID cases without strong or very strong genetic risk were compared with infants alive at age 1 year, while SUID risk associated with prenatal cannabis increased from adjusted hazard ratio (aHR) 3.7 to 6.0, other substance abuse from aHR 2.6 to 3.5, and black race from aHR 1.9 to 2.5. Thus, genome x environment analysis of a large cohort unveiled etiologic heterogeneity and hidden SUID risks, highlighting cannabis and genetic diseases as strong risk factors. Since preventative or therapeutic interventions were available for 83% of genetic risks, newborn screening by WGS has potential for substantial SUID reduction. Educational campaigns for SUID should emphasize perinatal cannabis avoidance.

Fig. 1:. Graphs of integrated G * E models of causal association of SUID.

Arrows with solid lines represent direct associations and those with dotted lines represent associations subsumed by a direct association. Birth prevalences are indicated. Risk magnitudes are indicated by prevalence risk ratios (PRR) for G factors and cohort adjusted hazard ratios (aHR) for E factors (from Table 1). a Interactions of SUID risk genetic (G, red) and non-genetic factors (E, blue). E_A and G_A indicate factors with independent, direct causal association with SUID (not confounded by other factors). E_B, E_C, E_D, G_B, G_C, and G_D represent factors with direct but dependent causal association with SUID (potentially confounded or amplified by other factors). Examples shown are new prone sleep (b), TSPYL1–SIDS with dysgenesis of the testes (SIDSTD, c), CACNA1C–Timothy syndrome (TS, d), CACNA1C–Long QT syndrome 8 (LTQS8, e), SERPINH1–susceptibility to preterm premature rupture of the membranes (SPROM, f), SLC26A3–Congenital secretory chloride diarrhea 1 (CSCD1, g), SFTPB–pulmonary surfactant metabolism dysfunction disorder 1 (PSMD1, h), GATA4–Tetralogy of Fallot (TOF, i), Tetralogy of Fallot of all causes (j), and Miller-Dieker Lissencephaly (k). These are discussed in Supplementary Results. Fever, hypoglycemia, and various drugs exacerbate cardiac dysrhythmias in TS and LTQS8. Other than nicotine, alcohol, and substance abuse disorder these have not been shown to have an association with SUID. Prematurity can have either a direct association with SUID (as in SPROM) or indirect (as in CSCD1). Newborn respiratory distress syndrome and major congenital malformations can have either direct or indirect associations with SUID (as in PSMD1 and TOF, respectively).

Fig. 2:. Pie charts showing categorization of 212 SUID cases by genetic etiology.

a. Categorization of SUID etiology as possibly, probably, or not genetic based on WGS findings. b. Number of genetic risk factors identified by WGS in SUID cases with probable or possible genetic etiology. c. Categorization of SUID genetic risk by principal organ affected.

Fig. 3:. Graphs of integrated G x E models of causal association of SUID.

Arrows with solid lines represent direct associations and those with dotted lines represent associations subsumed by a direct association. Birth prevalence (%) is indicated. Risk magnitudes are indicated by prevalence risk ratios (PRR) for G factors and cohort adjusted hazard ratios (aHR) for E factors (from Table 1). a Infant 1235114 with Jacobsen syndrome, a very strong SUID risk factor, considered sufficient for SUID alone. It is causally associated with congenital anomalies, respiratory distress syndrome, and premature birth. E SUID risk factors were considered subsumed to Jacobsen syndrome for SUID causality. b Infant 1253290 with two very strong SUID risk factors (CHARGE syndrome, Heterotaxy 5). Both are associated with congenital anomalies and respiratory distress syndrome, and Heterotaxy with premature birth. These and all other E SUID risk factors were considered subsumed to the G factors for SUID causality. c Infant 1229840 with a strong SUID risk factor (a VUS in CACNA1C, the causal gene in Timothy syndrome). The E factors fever, infection, hypoglycemia and various drugs, modify Timothy syndrome-associated SUID risk. d Infant 1253265 with Spherocytosis 3, a weak SUID risk factor that is associated with premature birth and respiratory distress. However, given the relatively weak G risk, all E factors were retained. e-g Causal G x E models for SUID cases with very strong (e), strong (f), and weak (g) G risk.

Figure 4:. Categorization of 212 SUID cases by the magnitude of genetic (G) and environmental (E) risks.

a Categorization by G risk Prevalence Risk Ratio (PRR). b Pie chart showing integrated G*E models for SUID cases with weak/no G risk, strong G risk, and very strong G risk. The magnitude of individual E risks is shown as adjusted hazard ratios. c. Pie chart showing integrated G*E models for SUID cases without genetic etiology, possible genetic etiology and probable genetic etiology.