A Toll-like receptor 2 genetic variant modulates occurrence of bacterial infections in patients with sickle cell disease
Summary
Despite adequate immunization and penicillin prophylaxis, bacterial infections remain a leading cause of morbidity and mortality in patients with sickle cell disease (SCD). Besides hyposplenism, inflammatory and genetic factors might modulate their susceptibility to bacterial infections. We performed a candidate gene association of single nucleotide polymorphisms (SNPs) located in Toll-like receptor (TLR) genes, encoding prominent molecules for innate immune responses, with the occurrence of bacterial infections in patients with SCD. A cohort followed in centres in Brazil, France and Senegal (n = 430) was divided in two groups: patients who presented at least one episode of bacterial infection (n = 235) and patients who never had bacterial infections (n = 195). There were no differences in gender or age distribution among the groups. The frequency of the TLR2 rs4696480 TA genotype was significantly lower in the infected group (50% vs. 67%, odds ratio [OR] = 0·50, 95% confidence interval [CI] 0·34–0·75, P < 0·001), and the TT genotype was significantly higher in the infected group (15% vs. 5%, OR = 3·18, 95% CI 1·53–6·61, P < 0·001). Previous reports demonstrated higher secretion of inflammatory factors in cells from AA individuals, lower occurrence and severity of immune diseases in T carriers. The rs4696480 TA genotype might stand between deleterious effects of over inflammatory response (AA genotype) and inefficient responses (TT genotype) to infectious agents in SCD settings.
Sickle cell disease (SCD) is the most commonly inherited haemoglobin disorder (Piel et al, 2017). A single nucleotide polymorphism (SNP) in the beta-globin gene (HBB) leads to the synthesis of the mutant haemoglobin S (HbS) (Stuart & Nagel, 2004; Yawn et al, 2014). When HbS polymerizes under deoxygenation and stress conditions, the red blood cell (RBC) acquires a typical sickle shape (Oteng-Ntim et al, 2015). SCD refers to homozygous individuals (SS) and to compound heterozygous of the HbS mutation combined with other haemoglobin traits (HbSC, HbSB and others). Although the HbS mutation is remarkably prevalent in Sub-Saharan, Arabian and Indian populations, the incidence of SCD is increasing in different regions, and it has become a public health issue worldwide (Piel et al, 2017).
Patients with SCD show a heterogeneous clinical presentation, ranging from mild clinical symptoms to death due to disease complications (Ballas et al, 2010). SCD complications are caused mainly by vaso-occlusion and haemolysis (Yawn et al, 2014; Piel et al, 2017; Kato et al, 2018). Repeated vaso-occlusive episodes in the spleen lead to loss of spleen function and atrophy in most SCD patients (Brousse et al, 2014). Impaired spleen function partially explains the susceptibility to bacterial infections observed in SCD, because the vast majority of severe infections occurs before the age of 5 years when the spleen is still, at least partially, functional; however mechanic factors, lower activity of inflammatory pathways, nutrition deficiency and genetic factors might also play a role (Booth et al, 2010). Despite adequate vaccination and prophylaxis, infection remains a leading cause of morbidity and mortality among SCD patients (Booth et al, 2010).
Patients with SCD present a steady-state inflammatory status (Pitanga et al, 2016). Although some studies addressed the role of inflammatory proteins and cells in SCD (Styles et al, 2000; Tamouza et al, 2002; Hoppe et al, 2003; Adekile et al, 2005; Wallace et al, 2009; Wallace & Linden, 2010; Vingert et al, 2015; Godefroy et al, 2016; Pitanga et al, 2016; Tatari-Calderone et al, 2016), few studies have investigated the potential associations between SCD complications and polymorphisms in genes encoding inflammatory proteins (Norris et al, 1996; Neonato et al, 1999; Styles et al, 2000; Tamouza et al, 2002; Hoppe et al, 2003; Adekile et al, 2005; Fertrin & Costa, 2010; Tatari-Calderone et al, 2016; Sippert et al, 2017; David et al, 2018). In this context, Toll-like receptors (TLRs) are a family of transmembrane and intra-cellular proteins expressed on immune cells that sense pathogen-associated molecular patterns (PAMPs) and endogenous damage-associated molecular patterns (DAMPs). TLRs bind specific related ligands and, by activating downstream signalling cascades, promote transcription of inflammatory cytokines and chemokines (Akira et al, 2006; Oliveira et al, 2015; Ntoufa et al, 2016). TLRs are part of the innate immune system and are at the forefront response against several types of pathogens, including bacteria. Previous reports showed an association between polymorphisms in genes encoding TLR proteins and susceptibility or severity of infection (Sanders et al, 2011; van Well et al, 2012; Bulat-Kardum et al, 2015; Gopalakrishnan & Salgame, 2016). In addition, TLRs might be part of the SCD steady state inflammatory pathways that lead to SCD complications. For instance, TLR4 activation by the haem group, released by intravascular haemolysis, modulates alloimmunisation in this setting (Godefroy et al, 2016; Pitanga et al, 2016).
Given the high and deleterious burden of infectious events in SCD, and the pivotal role of TLRs in anti-infectious innate immune responses, we assessed the influence of selected genetic variants of TLR loci on the susceptibility to bacterial infections in patients with SCD.
Methods
Study population
A total of 430 patients with SCD followed in the Tenon Hospital, Paris, France (n = 180, mainly from Sub-Saharan Africa and, to a lesser extent, the French West Indies), Clinics Hospital of Ribeirão Preto, Ribeirão Preto, Brazil, (n = 120), Clinics Hospital of São Paulo, São Paulo, Brazil (n = 76), and Albert Royer Hospital, Dakar, Senegal (n = 54) were enrolled in the study. All patients, previously diagnosed with SCD by haemoglobin electrophoresis or high performance liquid chromatography, had DNA samples stored and clinical data recorded, which were retrospectively collected from patients records. Bacterial infection was defined as an infectious event ascertained by identification of bacteria in blood or tissue cultures, or by clinical and laboratory findings, occurring at any time from birth to last follow-up. Only infections in the central nervous system (CNS), lower respiratory tract, upper urinary tract, abdomen, bone and joints, blood stream or at any site for tuberculosis were considered. For patients who had undergone haematopoietic stem cell transplantation (HSCT), only bacterial infections that occurred prior to HSCT were considered. Informed consent was obtained according to the declaration of Helsinki. The study was approved by all local ethical board reviews.
Genotyping
DNA was extracted from peripheral blood samples using standard methods (Miller et al, 1988). Seven SNPs in TLR1 (rs4833095), TLR2 (rs4308099, rs4696480, rs4308100), TLR6 (rs5743810) and TLR10 (rs11466653, rs11096957), all located in chromosome 4, were genotyped by mean of TaqMan 5′ nuclease assay (Applied Biosystems, Foster City, CA, USA) using pre-designed allele-specific fluorogenic oligonucleotide probes. The polymerase chain reaction (PCR) technique employed was described elsewhere (Oliveira et al, 2015). Prior to PCR reaction, DNA was pre-amplified, using the following protocol: primers 40× were diluted with Trizol to 0·2×; for each sample, we added 1·3 μl of DNA at 20 ng/μl, 1 μl of pre-amplification master mix (Fluidigm, San Francisco, CA, USA), 1·25 μl of pooling assay with primers and 1·45 μl of pure water. Plates were centrifuged at 18°C for 1 min, and the thermocycler conditions were 75°C for 2 min, 95°C for 15 s and 60°C for 4 min for 14 cycles. The final product was diluted with Trizol at 1:5 proportion and 1 μl of the final product was used for the PCR reaction.
Comparison with a non-SCD population
For some SNPs, we compared genotype distribution between the study population and the African population described on the 1000 Genomes Project database, comprising 661 healthy adult individuals, mainly from Sub-Saharan African and African-American origin (http://phase3browser.1000genomes.org/index.html).
Statistical analysis
Univariate analyses were performed with chi-square or Mann-Whitney tests for categorical and continuous variables respectively. Associations were measured by odds ratio (OR). Multiple testing was adjusted by the Bonferroni method for significant thresholds of P-value. For SNP associations, corrected significant P-value (cP) was 0·007; for genetic models, cP was 0·01. Linkage disequilibrium between SNPs and haplotype analyses were performed by logistic regression using SNPstats (Solé et al, 2006). Analyses were performed on SPSS 21.0 (SPSS Inc, Chicago, IL, USA).
Results
General population
A total of 430 patients with SCD were included in the study. Ethnical origin was known for 419 patients: 193 were Brazilian, 179 Sub-Saharan African, 41 French West Indies, 5 North African and one Indian. Median age at last follow-up was 30 years (range: 2–70) and 346 (87%) patients were older than 18 years (available n = 399); 243 (57%) patients were female and 184 (43%) were male (available n = 427). SCD genotype was SS in 344 (81%), SC in 46 (11%), SB in 31 (7%) and SD in 2 (1%) (available n = 423). Two hundred and nineteen (58%) patients received hydroxycarbamide (available n = 380), 294 (79%) received at least one red blood cell (RBC) transfusions (available n = 374), and 7 patients underwent HSCT. Eleven patients died during follow-up, mostly from acute chest syndrome and haemorrhagic stroke. In all countries, all patients had access to immunization against encapsulated bacteria and to penicillin prophylaxis at least up to 5 years of age.
Infections
Two hundred and thirty-five (55%) SCD patients had presented at least one episode of bacterial infection, and 195 (45%) SCD patients had not presented any bacterial infections. In the infected group, 135 (58%) were female, 192 (88%) were older than 18 years at last follow-up, 103 were Brazilian, 9 from Sub-Saharan Africa, 25 from French West Indies and 4 from other origin. In the non-infected group, 108 patients (56%) were female, 154 (86%) were older than 18 years at last follow-up, 90 were Brazilian, 83 from Sub-Saharan Africa, 16 from French West Indies and 2 from other origin.
Of the 235 infected patients, 44 (20%) had had more than 3 episodes of bacterial infection. Infections occurred in the lower respiratory tract in 121 cases (52%), bone and joints in 53 cases (23%), upper urinary tract in 47 cases (20%), CNS in 17 cases (7%), abdomen in 13 cases (6%), blood stream in 24 cases (10%); tuberculosis infections (n = 17, 7%) were pulmonary in 10 cases and extrapulmonary in 7 cases. In 106 cases an aetiological agent was identified. Most common agents were Escherichia coli, Streptococcus pneumoniae, Mycobacterium tuberculosis and Salmonella spp. In univariate analysis, no association was found between occurrence of infections and gender, origin or age distribution. Table 1 summarizes demography and univariate analysis.
| Variable | Non-infected (n = 235) | Infected (n = 195) | P-value |
|---|---|---|---|
| Gender (n = 427) | |||
| Female | 108 (56) | 135 (58) | NS |
| Male | 85 (44) | 99 (42) | |
| Origin (n = 419) | |||
| Brazil | 90 (47) | 103 (45) | NS |
| Subsaharan Africa | 83 (44) | 96 (42) | |
| French West Indies | 16 (8) | 25 (11) | |
| North Africa | 2 (1) | 3 (1) | |
| Other | 0 | 1 (1) | |
| Age > 18 years at last follow-up (n = 401) | |||
| No | 26 (14) | 27 (12) | NS |
| Yes | 154 (86) | 192 (88) | |
| rs3804099 (n = 426) | |||
| C/C | 41 (21) | 44 (19) | NS |
| C/T | 87 (45) | 104 (45) | |
| T/T | 65 (34) | 85 (36) | |
| rs4696480 (n = 425) | |||
| T/T | 10 (5) | 35 (15) | <0·001 |
| T/A | 127 (67) | 117 (50) | |
| A/A | 54 (28) | 82 (35) | |
| rs3804100 (n = 416) | |||
| T/T | 170 (91) | 207 (90) | NS |
| C/T | 17 (9) | 20 (9) | |
| C/C | 0 | 1 (1) | |
| rs11466653 (n = 427) | |||
| A/A | 175 (91) | 215 (92) | NS |
| G/A | 17 (8) | 19 (8) | |
| G/G | 1 (1) | 0 | |
| rs11096957 (n = 426) | |||
| G/G | 48 (25) | 57 (25) | NS |
| G/T | 103 (53) | 115 (49) | |
| T/T | 42 (22) | 61 (26) | |
| rs4383095 (n = 401) | |||
| G/G | 51 (28) | 49 (22) | NS |
| A/G | 75 (42) | 104 (47) | |
| A/A | 53 (30) | 69 (31) | |
| rs5743810 (n = 429) | |||
| G/G | 158 (81) | 205 (87) | NS |
| A/G | 31 (16) | 28 (12) | |
| A/A | 5 (3) | 2 (1) | |
- NS, not significant.
- The bold value indicates the significant P-value in univariate analysis.
Genetic analysis
All tested SNPs had a call rate greater than 90% and a minimum allele frequency of at least 1%. SNPs rs4696480, rs5743810 and rs4833095 were deviated from Hardy-Weinberg equilibrium (HWE) in the whole study population and in the non-infected group. Given that the whole study population was affected by SCD, and consequently there were no healthy controls in this population, these SNPs were considered for further analyses.
In univariate analysis, SNP rs4646980 in TLR2 (n = 425) was significantly associated with occurrence of infections (P < 0·001). The distribution of rs4696480 genotypes was, in the infected patients, AA = 82 (35%), TA = 117 (50%), TT = 35 (15%); in the non-infected patients, distribution was AA = 54 (28%), TA = 127 (67%), TT = 10 (5%). Next, genetic models for this association were tested. In the over-dominant model, the heterozygous genotype (TA) occurred significantly less in infected patients than in non-infected patients compared with TT + AA patients (OR = 0·50, 95% confidence interval [CI] 0·34–0·75, P < 0·001). Furthermore, in the recessive model, the TT genotype occurred significantly more in infected patients than in non-infected patients compared with TA + AA (OR = 3·18, 95% CI 1·53–6·61, P < 0·001). Among the assessed SNPs, only rs11466653 was in significant linkage disequilibrium with rs4696480. Haplotype analyses were performed, but no association between any haplotype and occurrence of infections were identified. Table 1 shows genotype distribution and univariate analysis. Table 2 summarizes the genetic models for rs4696480.
| Odds ratio | 95% CI | P-value | |
|---|---|---|---|
| Dominant | |||
| AA × TA + TT | NS | ||
| Recessive | |||
| TT × TA + AA | 3·18 | 1·53–6·61 | <0·001 |
| Overdominant | |||
| TA × AA + TT | 0·5 | 0·34–0·75 | <0·001 |
| Log-additive | NS | ||
- CI, confidence interval; NS, not significant.
Comparison with non-SCD populations
We compared the genotype distribution of SNP rs4696480 in this SCD population with a population with African origins, comprising adult healthy individuals, described in the 1000 Genomes Project database (n = 661). The TA genotype was significantly more frequent in the SCD population (n = 244, 57%) than in the 1000 Genomes Project population (n = 295, 45%) compared with TT + AA (OR 1·67, 95% CI 1·30–2·13, P < 0·001). Moreover, TT genotype was significantly less frequent than AA + TA in the SCD population (n = 45, 11%) compared to the 1000 Genomes Project population (n = 268, 41%) (OR 0·17, 95% CI 0·12–0·24, P < 0·001).
Discussion
We have shown that a polymorphism in the TLR2 gene is associated with susceptibility to/protection against infections in patients with SCD. More specifically, while the heterozygous rs4696480 TA genotype apparently confers protection against bacterial infections, the TT genotype seems to increase risk of developing such infections.
TLRs are glycoproteins that recognize a wide range of PAMPs and DAMPs, with the consequent induction of inflammatory responses via nuclear factor (NF)-κB family members (Akira et al, 2006; Oliveira et al, 2015; Ntoufa et al, 2016). Currently, 13 TLRs are known. Because they are pattern recognition receptors (PRR), each TLR binds specific related structures (Ntoufa et al, 2016). However, some TLRs can recognize several structural unrelated ligands, namely TLR2 and TLR4 (Akira et al, 2006) or act as heterodimeric complex extending the repertoire of pattern recognition. TLR2 is highly expressed in monocytes and is capable of recognizing a huge number of proinflammatory inducers, including membrane lipoproteins of gram-positive bacteria, mycobacteria, yeast, mycoplasma, spirochete and parasites (Hornung et al, 2002; Eder et al, 2004; Veltkamp et al, 2007; Oh et al, 2009; Kerkhof et al, 2010; Gopalakrishnan & Salgame, 2016; Ntoufa et al, 2016). TLR2 may function alone or as heterodimers with TLR1, TLR6 and TLR10 towards NF-κB cascade activation; TLR10 also negatively regulates TLR2 (Moresco et al, 2011). Therefore, TLR2 is a key receptor for several exogenous pathogens, playing a pivotal role in immune response against a wide range of pathogens.
The SNP rs4696480 is located in the promoter region of the TLR2 gene. This SNP was previously studied in the context of inflammatory diseases. Some reports in asthma, atopic dermatitis, inflammatory bowel disease and B hepatitis settings suggested that the presence of a T allele or TT genotype was associated with a lower inflammatory response, lower response to treatment with anti-tumour necrosis factor, and that the presence of an A allele led to a more severe inflammatory response (Eder et al, 2004; Oh et al, 2009; Kerkhof et al, 2010; Bank et al, 2014; Lin et al, 2018; Loft et al, 2018). Nevertheless, other studies did not show any effect of this SNP on the incidence of asthma (Kormann et al, 2008; Miedema et al, 2012). In addition, a previous case-control study assessed the effect of the S180L SNP in the Toll-interleukin 1 receptor (TIR) domain-containing adaptor protein (TIRAP), part of the downstream signalling of TLR2, in severe bacterial infections in a non-SCD population (Khor et al, 2007). Similar to our results, the heterozygous genotype showed a protective effect against infection. Heterozygous transfected cells in vitro failed to bind TLR2, reducing activation of the NF-κB pathway (Khor et al, 2007). Based on these findings, we postulate that the rs4696480 TA genotype enables an adequate inflammatory response to be mounted against bacterial infections in SCD patients. The AA genotype might lead to an exacerbated inflammatory reaction, whereas the response determined by the TT genotype is weaker than necessary to defeat bacteria.
In our study, the SNP rs4696480 was in HWE deviation in the whole SCD population and in the non-infected SCD patients. In addition, TA was more frequent, and TT was significantly less frequent in the SCD population compared with the African population described on the 1000 Genomes Project database. Our study population mostly consisted of adult patients. Knowing that bacterial infections are the leading cause of mortality among SCD children, the HWE and the differences in genotype frequencies might be explained by a positive selective pressure due to the higher occurrence of infections in the SCD population. In other words, we cannot exclude that some patients with SCD carrying the TT genotype might have died in early childhood and could not have been considered in a study such as ours. However, this hypothesis requires further assessment.
Conflicting results were reported by David et al (2018). In a cohort of 95 paediatric infected and non-infected homozygous SS patients, the TA genotype was associated with more severe infections, and the AA genotype was protective against respiratory infections. Differences in sample size and age distribution might explain the discrepancy in the results, and could be in line with our observations concerning HWE expectations. Also, in that study, presence of alpha thalassaemia 3·7 kb deletion was associated with less infection occurrence. Previous studies reported a protective effect of alpha thalassaemia against hospitalisations due to infections and respiratory infections in non-SCD settings (Allen et al, 1997; Wambua et al, 2006). Alpha thalassaemia status was not available for most patients in our SCD cohort; therefore, we were unable to investigate its possible association with infection in our study.
Our study has several limitations: the retrospective nature of collected data and the low sensitivity of bacterial culture methods prevented further analyses on the type of bacteria involved. Similarly, because the study population is comprised mostly of adults, we cannot perform any comparisons on genotype distribution among different age groups. Nevertheless, the strength of this study relies on the large sample size and on the power of the associations found.
Loss of spleen function results in defective opsonization and impairs destruction of encapsulated bacteria. However, hyposplenism does not completely explain the higher occurrence of infections in SCD. Patients with SCD are also at greater risk of bacterial infections caused by non-encapsulated bacteria (Morrissey et al, 2015; Sobota et al, 2015). In addition, even among patients carrying the same SCD genotype and presenting hyposplenism, occurrence of bacterial infections is highly variable. Our results might contribute to better understand the mechanisms involved in the greater risk of bacterial infections in SCD settings, to ameliorate strategies for prevention and treatment of infections. Further studies, including candidate gene and genome-wide association studies, are needed to evaluate the impact of genetic factors in patients with SCD, with the aim of explaining the heterogeneity of clinical outcomes in this setting and, also, to confirm our findings.
Acknowledgments
This work was supported by the Centre Scientifique de Monaco and the Brazilian Coordination for the Improvement of Higher Education Personnel (CAPES). The authors thank the Centre Scientifique de Monaco, CAPES, all patients and their families.
Conflict of interest
There are no competing interests to report.
Author contributions
KTM, BPS, EG, RG and RT designed the study. KTM, IDL and CK collected clinical data. KTM performed experiments. CM and ILA advised experiments. KTM performed statistical analysis. RG, YB, ID, IDL, ACSP, SK, ESR, SFMG, GHHF, VR, CLD and LJ provided cases and samples for the study. KTM, FV, AR, EG, BPS, RG, HR, BC and RT wrote the manuscript. All authors edited and approved the manuscript.
