Supplement S.1 - Methods

Infant recruitment: Premature infants were recruited from nurseries at Children’s Hospital of Wisconsin (Milwaukee, WI), St. Joseph’s Hospital (Milwaukee, WI), Children’s Hospitals and Clinics of Minnesota (Minneapolis, MN), Rush University Medical Center (Chicago, IL), Kosair’s Children’s Hospital (Louisville, KY), BC and Children’s & Women’s Health Centre (Vancouver, Canada), and University of Iowa Children’s Hospital (Iowa City, IA) after institutional review board approval. After informed consent, 0.5mL of blood was collected in containers, and shipped to Children's Hospital of Wisconsin where DNA extraction and genotyping was done. For subjects from Iowa, de-identified DNA samples were sent to Children's Hospital of Wisconsin. For subjects recruited in Canada genotyping was done at the local site and the de-identified clinical and genotyping data were sent to the principal investigator’s site. The proportion of cases and controls from each center is shown in S.4. Clinical data were de-identified, assigned a study code and entered into a password-protected database.

Eligibility criteria: Infants born with a gestational age (GA) ≤ 35 weeks without major congenital anomalies of the heart, gastro-intestinal tract or lung were eligible. Case definition: Infants with clinical signs of sepsis who had a blood culture-positive bacterial infection treated with antibiotics for ≥ 5 days were diagnosed with BSI. This definition would ensure a tighter phenotype by excluding infants treated briefly for bacteria considered as contaminants, had rule-out sepsis, and had culture-negative sepsis. Urinary tract infections and pneumonias were not considered as cases. Early (<72hr of life)- and late (>72hr of life)-onset sepsis were also defined.

Selection of SNPs: Candidate NLR single nucleotide polymorphisms (SNPs) were identified based on whether: 1) associations with inflammatory bowel disease or infectious disease have been reported, 2) variants had a functional effect, and 3) variants had a mean allele frequency (MAF) > 2% in the Caucasian population.

Genotyping: DNA was extracted from blood samples using the FlexiGene DNA kit (Qiagen, Inc., Valencia, CA) and stored at 4°C. TheNOD1 (rs6958571), NOD2 (rs2066844), ATG16L1 (rs2241880), CARD8 (rs2043211), and NLRP3 (rs4353135, rs6672995, and rs35829419) variants were genotyped by performing a 5′ nuclease Taqman assay (Applied Biosystems, Foster City, CA) using custom or predesigned TaqMan® SNP Genotyping Assay probes (ABI, Foster City, CA) 1. The assay involves amplification of the genomic region of interest followed with ligation with allele-specific probes that emit a distinct fluorescent signal specific to the reference or variant allele. Samples were analyzed using an ABI HT7900 analyzer with SDS 2.3 software package. Genotyping was done by personnel blinded to clinical outcomes.Quality control:10% of the samples were re-genotyped to confirm prior results. There was >98% concordance for all samples.

Statistical analysis: Chi-square tests were used for comparisons between clinical variables and sepsis outcomes. GA and birth-weight (b.wt) were compared between groups using the Wilcoxon-Mann-Whitney rank sum test. We used genetic additive and recessive risk models to analyze relationships between SNPs and GPB and GNB BSI outcomes. Allele frequencies were compared among infants with and without BSI using Pearson’s Chi-square or Fisher’s exact tests. Cochran-Armitage trend tests were used to examine the additive risk associated with risk alleles. In pre-specified a priori analysis, relationships between SNPs and BSI would be examined in Caucasian (CAU) infants and infants with birth weight 1000gm. Power: Assuming a BSI rate of 20% in our cohort, we estimated that a sample size of 750 infants would give us 80% power with a p=0.007 (Bonferroni correction for 7 SNPs) to detect a 12 - 20 % difference in the prevalence of the variant allele between infants with and without BSI. To adjust for potential clinical confounders, we used logistic regression with backward elimination where the probability of removal was set at p≥0.05. Birth variables (GA or b.wt, antenatal steroid exposure, chorioamnionitis, race and gender) along with NLR SNPs were examined for association with BSI, and variables were removed from the model in a step-wise fashion until only those associated with BSI (p<0.05) remained. Similar modelling was done for infants with birth-weight 1000g and Caucasian infants. The Hosmer and Lemeshow Goodness-of-Fit test was used to assess the degree of fit in regression models. SPSS 21.0 (SPSS Inc., Chicago, IL) and SAS 9.4 (SAS Inc., NC) were used for data analysis.

S.2 - Table: Bacteria associated with early- and late-onset BSI in our study cohort.

Early-onset BSI (n=19) / Late-onset BSI (n=119)
Group B streptococcus (n=3) / Staphylococcus Epidermidis (n=78)
Escherichia coli (n=3) / Escherichia coli (n=8)
Staphylococcus Epidermidis (n=5) / Klebsiella spp (n=6)
Haemophilus influenza (n=2) / Candida albicans (n=4)
Enterobacter spp (n=1) / Staphylococcus aureus# (n=9)
Enterococcus spp (n=1) / Methicillin-resistant S. aureus* (n=4)
Pseudomonas spp (n=1) / Group B streptococcus (n=2)
Streptococcus mitis (n=1) / Enterococcus spp (n=6)
Acinetobacter lwoffi (n=1) / Serratia spp (n=2)
Corynebacterium spp (n=1) / Pseudomonas spp, Citrobacter freundi, Enterobacter cloacae, Gram-negative bacilli, Micrococcus, Neisseria, Multiple organisms, Staphylococcus simulans, Streptococcus spp, Streptococcus pneumoniae, Candida parapsilosis (n=1)

# - One infant had CSF infection and BSI; * - One infant had MRSA meningitis only.

S.3 - Distribution of NLR genetic variants in relation to BSI in the entire cohort (N=764)

Variant / Genotype Frequency – Number (%)
No sepsis (n=626) / BSI (n=138)
ATG16L1 rs2241880 AA
AG
GG / 202 (32.5)
300 (48.3)
119 (19.2) / 52 (38.0)
56 (40.9)
29 (21.1)#
CARD8 rs2043211 AA
AT
TT / 312 (50.0)
262 (41.9)
50 (8.0) / 65 (47.1)
57 (41.3)
16 (11.6)
NLRP3 rs4353135 TT
GT
GG / 314 (50.4)
257 (41.3)
52 (8.3) / 68 (49.6)
55 (40.2)
14 (10.2)
NLRP3 rs6672995 GG
AG
AA / 451 (72.0)
161 (25.7)
14 (2.2) / 97 (70.3)
38 (27.5)
3 (2.2)
NLRP3 rs35829419 CC
AC
AA / 578 (92.5)
46 (7.4)
1 (0.2) / 134 (97.1)
4 (2.9)
0 (0.0)
NOD1 rs6958571 AA
AC
CC / 343 (55.1)
238 (38.3)
41 (6.6) / 76 (55.1)
49 (35.5)
13 (9.4)
NOD2 rs2066844 CC
CT
TT / 587 (94.2)
36 (5.8)
0 (0.0) / 130 (94.2)
8 (5.8)
0 (0.0)

Genotype frequencies of NLR variants classified by phenotype is presented. rs number; reference SNP accession ID number. Genotyping results were unavailable for the following variants in our cohort; CARD8 (n=2), NOD2 (n=3), ATG16L1 (n=6), NLRP3 rs4353135 and NOD1 (n=4 each), and NLRP3 rs35829419 (n=1). Chi-square tests were used to determine associations between variants and BSI. We did not find significant associations between SNPs and GPB or GNB BSI (data not shown). # -p=0.08

S.4 – Data on cases and controls recruited from each center (N=764)

Center / Infants recruited (n) / Controls (n) / Cases (n)
Iowa Children’s Hospital / 176 / 154 / 22
Children’s Hospital of Wisconsin system / 171 / 141 / 30
St. Joseph’s Hospital, Wisconsin / 62 / 54 / 8
Children’s Hospital of Vancouver / 157 / 111 / 46
Rush University Hospital, Chicago / 33 / 30 / 3
Kosair Children’s Hospital, Louisville / 45 / 40 / 5
Children’s Minnesota Hospital & Clinics / 120 / 96 / 24