The authors have declared that no competing interests exist.
Conceived and designed the experiments: JD CP JC OD. Performed the experiments: FC KT NC SS. Analyzed the data: JD FC CP PR MC. Contributed reagents/materials/analysis tools: FC NC. Wrote the manuscript: FC CP JD SB.
The HLA-G molecule plays an important role in immunomodulation. In a previous study carried out on a southern French population our team showed that HLA-G haplotypes, defined by SNPs in the coding region and specific SNPs located in 5′URR and 3′UTR regulatory regions, are associated with differential soluble HLA-G expression (sHLA-G). Furthermore, the structure of these HLA-G haplotypes appears to be conserved in geographically distant populations.
The aim of our study is to confirm these expectations in a sub-Saharan African population and to explore additional factors, such as HLA-A alleles, that might influence sHLA-G expression.
DNA and plasma samples were collected from 229 Malians; HLA-G and HLA-A genotyping were respectively performed by the Snap Shot® method and by Luminex™ technology. sHLA-G dosage was performed using an ELISA kit. HLA-G and HLA-A allelic and haplotypic frequencies were estimated using an EM algorithm from the Gene[Rate] program. Associations between genetic and non genetic parameters with sHLA-G were performed using a non-parametric test with GRAPH PAD Prism 5.
Our results reveal a good conservation of the HLA-G UTR haplotype structure in populations with different origins and demographic histories. These UTR haplotypes appear to be involved in different sHLA-G expression patterns. Specifically, the UTR-2 haplotype was associated with low sHLA-G levels, displaying a dominant negative effect. Furthermore, an allelic effect of both HLA-G and HLA-A, as well as non genetic parameters, such as age and gender possibly linked to osteogenesis and sexual hormones, also seem to be involved in the modulation of sHLA-G.
These data suggest that further investigation in larger cohorts and in populations from various ethnical backgrounds is necessary not only to detect new functional polymorphism in HLA-G regulatory regions, but also to reveal the extent of biological phenomena that influence sHLA-G secretion and this might therefore have an impact on transplantation practice.
The role of the non-classical class Ib Human Leukocyte Antigen-G (HLA-G) in immune-tolerance has been well documented
Contrary to the classical HLA class I loci, HLA-G is characterized by a low polymorphism in the coding regions. To this day, 50 HLA-G alleles have been identified, which encode 16 trans-membrane proteins (HLA-G*01:01 to G*01:04, G*01:06 to G*01:12 and G*01:14 to G*01:18) and two truncated proteins (HLAG*01:05N and G*01:13N)
Several studies have suggested an association between soluble (s)HLA-G expression and specific HLA-G alleles or SNPs in the non-coding regions. Notably, HLA-G*01:04 and G*01:05N have been respectively associated with high and low HLA-G secretion
HLA-A is the closest functional gene to HLA-G. The genetic distance between these two genes is approximately 150 Kb
Castelli et al. defined 8 UTR HLA-G haplotype groups using sequenced SNPs in the 5′URR, 3′UTR and coding regions in a Brazilian population
Based on this study, our team investigated HLA-G UTR haplotype conservation and its association with the expression of sHLA-G in serum from Volunteer Bone Marrow Donors (VBMD) from South-eastern France
On this basis, we propose to investigate the following hypotheses: (1) the restricted number of UTR HLA-G haplotypes and their structure may reflect selective forces associated with differential expression of sHLA-G and its biological significance. This supposition should be confirmed in a sub-Saharan African population since these populations generally display higher genetic diversity and lower levels of linkage disequilibrium compared to populations from other continents. (2) Accordingly, the association between specific UTR haplotypes and sHLA-G levels should be reproducible at plasma level in Malian samples. (3) Finally, the HLA-A gene, due to its proximity to the HLA-G gene, may influence the expression of sHLA-G. Thus, haplotype conservation might be extended to the HLA-A gene.
Sample collection was conducted by the
Blood and plasma samples were collected from 229 unrelated Malians after informed consent, in the villages of: Bandiagara (n = 61), Binedama (n = 12) Madougou (n = 53), Manteourou (n = 51), N'Gono (n = 21) and Petaka (n = 31). The ethnic groups were: Dogon (n = 152, linguistic group Niger-Congo-Atlantic), Peulh (n = 68, linguistic group Niger-Congo-Atlantic), Tamashek (n = 5, linguistic group Afro-Asiatic-Berber) and Bambara (n = 4, linguistic group Niger-Congo-Mande). This cohort was composed of men and women (respectively n = 125 and n = 105,
DNA was extracted in Mali from a 200-µl whole blood sample using the QIAmp Blood DNA kit (Qiagen, Courtaboeuf, France) according to the manufacturer's instructions. Genomic analyses and serology were performed in Marseilles respectively on genomic DNA and plasma from the same cohort. 229 individuals were successfully analyzed for HLA-G coding alleles, 5′URR and 3′UTR polymorphisms and 195 individuals were analyzed for HLA-A coding alleles. sHLA-G level was determined in plasma samples from 219 individuals.
A homemade primer extension method was used to simultaneously analyse 16 SNP scattered along the HLA-G gene: 8 SNPs (codons 13C/T rs17875397; 31 A/T rs41551813; 54 A/G rs41545515; 110 C/A rs12722477; 130 C/T rs41557518; 159 T/C rs55916353; 219 C/T rs45530733 and 258 C/T rs12722482) defining HLA-G alleles (HLA-G*01:01 to G*01:09), 4 SNPs in the 5′ URR region (−725 C/G/T rs1233334, −716 G/T rs2249863, −201 G/A rs1233333 and −56 C/T rs17875397) and 4 SNPs in the 3′UTR region (
Luminex™ technology (HLA-A-One Lambda LABType® SSO) was used to determine HLA-A alleles at an intermediate resolution using the manufacturer's kit. The HLA-A allelic assignment is based on the HLA sequences listed in the official IMGT/HLA database 3.12.0 May 2013
Measurement of soluble isoforms HLA-G1 and -G5 was performed
HLA-G genotypes were automatically converted from output files (.txt) exported from GeneMapper 4.0 into coding alleles and UTR using an in-house computer program, readable by the ‘Phenotype’ application of the Gene[Rate] computer tool package (
Significant deviations from expected values at Hardy Weinberg Equilibrium (HWE) for all the 16 HLA-G SNPs were tested using a nested likelihood model
Frequencies for HLA-G alleles, SNPs in the 5′ and 3′ regions, HLA-A alleles, UTR∼HLA-G and HLA-A∼HLA-G∼UTR haplotypes were estimated using an EM algorithm from the Gene[Rate] program
Two-locus linkage disequilibrium (LD) was tested for the 16 HLA-G SNPs by a conventional goodness-of-fit test with the Arlequin v3.5.1.2 package
Gametic association between specific pairs of alleles is provided as a list of standardized residuals for each observed haplotype. The null hypothesis of independence of the loci implies a gaussian distribution of deviations and, by convention, absolute values over 2 are considered to be significant. In order to establish the relationships among the HLA-A and HLA-G haplotypes, a Median Joining (MJ) network
Associations between sHLA-G and genetic polymorphism (SNPs, allele or haplotype) or non genetic parameters (sex) were tested with non-parametric tests performed with GRAPH PAD Prism 5. Mann-Whitney t-test was used to test two modalities. Kruskal-Wallis one-way ANOVA followed by Dunn post-hoc test was used when there were more than two modalities. Statistical correlation between age and sHLA-G levels was tested using Spearman's rank test.
Frequencies of the HLA-G alleles, the 4 SNPs in the 5′URR (−725 C/G/T, −716 G/T, −201 G/A and −56 C/T) and the 4 SNPs in the 3′UTR (
HLA-G alleles, 5′URR and 3′UTR SNPs | |||||
G*01:01 | *** | ||||
G*01:04 | *** | ||||
G*01:03 | *** | ||||
G*01:06 | NS | ||||
G*01:05N | *** | ||||
−56 C | *** | ||||
−56 T | |||||
−201 G | *** | ||||
−201 A | |||||
−716 T | *** | ||||
−716 G | |||||
−725 C | NS | ||||
−725 G | *** | ||||
−725 T | *** | ||||
ex8 del | *** | ||||
ex8 ins | |||||
3142 C | *** | ||||
3142 G | |||||
3187 A | *** | ||||
3187 G | |||||
3196 C | * | ||||
3196 G |
Results are compared with previously published results on VBMD
No significant differences were observed for allele frequencies between villages or between ethnic groups, the Malian data were thus pooled together for the analysis.
Five alleles were found in the Malians: G*01:01 (45.2%)>G*01:04 (26.2%)>G*01:03 (15.2%)>G*01:05N (10.5%)>G*01:06 (2.2%). The five HLA-G alleles previously reported in the French population (VBMD) were all observed in Malians. Malian samples displayed a significantly higher frequency (p<0.001) for HLA-G*01:04, G*01:03, G*01:05N as compared to VBMD, while HLA-G*01:01 frequency was significantly higher in VBMD than in Malians (p<0.001); no statistical difference was found for G*01:06 (
Frequencies of SNPs in the 5′URR and 3′UTR were significantly different between VBMD and Malians (p<0.05), except for −725 C.
All SNPs of the HLA-G gene were in Hardy Weinberg Equilibrium (HWE), except the SNP at codon 130 that displayed a significantly higher heterozygosity than expected. A T in codon 130 specifically defines the HLA-G*01:05N allele.
Two-locus Linkage Disequilibrium (LD) for SNPs in 5′URR, 3′UTR and coding alleles is shown in
Statistical significance (p value at junction between two loci) is indicated by color (light gray: p<0.05, gray: p<0.01 and black p<0.001).
Global LD was observed between HLA-A and HLA-G alleles and between HLA-A alleles and HLA-G UTR, respectively (both p = 0 and quantile = 1000).
UTR composition, their allelic association and their frequencies in the Malian samples are shown in
UTR haplotype | −725 | −716 | −201 | −56 | ex 8 | 3142 | 3187 | 3196 | Coding allele | Ratio | Fq Malians | ||
C | T | G | C | C | G | C | G*01:01 | 1 | 0.134 | ||||
G | A | C | G | A | G | G*01:01 | 0.563 | 0.165 | NS | ||||
G*01:05N | 0.375 | 0.109 | *** | ||||||||||
G*01:06 | 0.083 | 0.025 | NS | ||||||||||
C | G | A | C | G | A | C | G*01:04 | 1 | 0.239 | *** | |||
G | T | G | C | C | A | C | G*01:01 | 1 | 0.05 | *** | |||
T | T | G | T | G | A | C | G*01:03 | 1 | 0.143 | *** | |||
C | T | G | C | C | A | C | G*01:01 | 1 | 0.121 | NS |
Comparison with previously published results on VBMD
UTR haplotypes, defined by four SNPs in the 5′URR (−725 C/G/T, −716 G/T, −201 G/A and −56 C/T) and four SNPs in the 3′UTR (
UTR-2 had the highest frequency in the Malian samples. UTR frequencies were as follows: UTR-2 (29.9%)>UTR-3 (23.9%)>UTR-5 (14.3%)>UTR-1 (13.4%)>UTR-6 (12.1%)>UTR-4 (5%). Malians displayed significantly higher frequencies (p<0.001) for UTR-2, UTR-3 and UTR-5 compared to VBMD and lower frequencies for UTR-1 and UTR-4 (p<0.001).
Concerning the association between UTR and HLA-G alleles, the same association previously reported in the VBMD population was described in the Malian samples (
Haplotype analysis was extended to HLA-A alleles in association with UTR and HLA-G alleles. These HLA-A∼UTR∼HLA-G haplotypes are in a limited number (43 HLA-A∼UTR∼HLA-G estimated out of 204 possible). HLA-A∼UTR∼HLA-G haplotypes estimated in the Malian samples and their frequencies are showed in
Association between HLA-A and HLA-G was further investigated using the Median Joining (MJ) method based on protein sequences (
The overall distribution of sHLA-G fits a Gaussian distribution. The sHLA-G mean value in the 219 plasma samples was 143.18±31.05 UI/ml.
No significant difference was found between gender and sHLA-G level even though men displayed lower values than women (139.7±29.87 UI/ml vs 147.4±32.06 UI/ml, p = 0.169).
However, a significant negative correlation (rS = −0.206, p = 0.002) was observed between age and sHLA-G levels. When individuals were classified according to gender, a significant negative correlation was only found for women (rS = −0.1905, p = 0.004) (
Statistical correlation between age and sHLA-G levels was tested using Spearman's rank test (rS = −0.1905, p = 0.004).
Statistical comparison was based on Kruskal-Wallis one-way ANOVA followed by Dunn post-hoc test (girls 3–25 years old: 158.4±31.6 UI/ml; boys 3–25 years old: 152.0±31.6 UI/ml; women over 26 years old: 137.5±26.6 UI/ml; men over 26 years old: 135.5±26.4 UI/ml). NS: not significant; *: p>0.05; **: p<0.01; ***: p<0.001).
Significant associations were found between 5′URR −716 G/T (p = 0.03), −201 G/A (p = 0.03) and 3′UTR +3196 C/G (p = 0.03) and sHLA-G. The Dunn post-hoc test showed significantly higher sHLA-G levels for −716 T/T; −201 G/G and +3196 C/C genotypes. No significant associations were observed for HLA-G alleles and the other 5′URR and 3′UTR SNPs.
The exon 8
sHLA-G mean values and standard deviation for each UTR genotype are shown in
UTR-2 individuals displayed a significantly lower level of sHLA compared to non-UTR-2 individuals (
Statistical comparison was based on Mann-Whitney t-test. (UTR-2 137.5±30.6 UI/ml vs. all except UTR-2 148.2±30.7 UI/ml). When six outer points were excluded (238.6 UI/ml, 217.6 UI/ml and 53.8 UI/ml for UTR-2; 242.2 UI/ml, 229.8 UI/ml and 226.5 UI/ml for non-UTR-2) UTR-2 individuals still displayed significantly lower values (p<0.05).
Statistical comparison was based on Kruskal-Wallis one-way ANOVA followed by Dunn post-hoc test.
No significant correlation was found between sHLA-G and the other UTRs.
To assess the influence of HLA-A alleles on sHLA-G expression, we tested the correlation within and among the A lineages defined by Gu X et al.
Statistical comparison was based on Mann-Whitney t-test (A*02≈UTR-5≈G*01:03 154.9±28.9 UI/ml; A*x≈UTR-5≈G*01:03 133.9±22.5 UI/ml). When outer points were excluded (212.4 UI/ml, 201.9 UI/ml, 92.2 UI/ml and 92.2 UI/ml for A*02≈UTR-5≈G*01:03; 86.4 UI/ml for A*x≈UTR-5≈G*01:03), the difference remained significant (p<0.05).
No significant differences in sHLA-G expression were found between HLA-A alleles from other HLA-A lineages.
In this study we present results based on HLA-G and HLA-A genotypes and sHLA-G serological analyses performed on 229 Malian samples. UTR haplotypes described in the Malian samples show reduced diversity (n = 6 estimated with the EM algorithm among 28 possible) and share a similar structure to those described in the French VBMD and in the Brazilian population
Stronger LD between SNPs within 5′URR and 3′UTR was found in the Malian samples than in the French samples. Previous multi-loci studies have suggested that sub-Saharan Africans display a lower LD compared to populations from other continents
Haplotype structure conservation was further confirmed when HLA-A alleles were considered since the number of HLA-A∼UTR∼HLA-G haplotypes was also greatly reduced in the Malian samples. The estimated haplotypes were limited to 43 compared to the 204 possible haplotypes according to the observed alleles. One should keep in mind that we studied 229 samples and that some rare haplotypes might not have been detected.
The second objective of this study was to confirm the correlation between sHLA-G expression level and UTR haplotypes. Highly variable sHLA-G values have been reported with the Elisa kit using the MEM-G/9 antibody according to the biological fluid analyzed or the calibration standard used; our results are in accordance with published data based on plasma samples and expressing their results in U/ml according to the calibration standard displayed by the supplier
We confirmed an association between the UTR-2 haplotype and lower sHLA-G levels in the Malian samples. UTR-2 individuals showed, both in homozygous and heterozygous state, a significant association with lower values. This result tends to show a Dominant Negative Effect (DNE) of UTR-2. Moreover, even though no significant association was observed between HLA-G alleles and sHLA-G levels, an allelic effect was still observed for UTR-2. Indeed, HLA-G*01:05N/G*01:05N showed higher (but not significantly) sHLA-G levels compared to other UTR-2 allelic combinations. This putative allelic effect of HLA-G*01:05N/G*01:05N on sHLA-G does not exclude a haplotype effect of HLA-A on UTR-2; indeed HLA-G*01:05N∼UTR-2 was only coupled with HLA-A*30:01. However, HLA-G*01:05N codes for a truncated protein since HLA-G*01:05N presents a stop codon in position +189, therefore the mRNA of G*01:05N translates only HLA-G5 and -G6 soluble isoforms
No significant correlation has been found between the other UTRs and sHLA-G levels. However, UTR-1 and UTR-5 homozygous individuals displayed higher sHLA-G levels than the overall sample mean. Interestingly HLA-A*02 subtypes bearing UTR-5 displayed significantly higher sHLA-G levels compared to other HLA-A alleles bearing UTR-5. This differential sHLA-G expression may be due to an HLA-A*02 effect or to supplementary polymorphisms in regulatory regions of UTR-5. Indeed, 3 SNPs in the 5′URR, −646, −540 and −509 described by Castelli et al. (2011) can separate UTR-5 in 2 sub haplotypes
Our analyses on HLA-A showed that HLA-A alleles display LD with both HLA-G alleles and UTR. HLA-A*23:01:01∼HLA-G*01:04∼UTR3 displayed the highest frequency, confirming the reported LD between HLA-A*23 (and HLA-A*24) and HLA-G*01:04 in different populations
HLA-G*01:04 has been previously associated with a high sHLA-G production
This study showed a negative correlation between age and sHLA-G levels in Malian women which remained significant when UTR-2 individuals were excluded. None of the previously published studies have shown correlation between age and sHLA-G levels, even related with gender. Moreover, we also found that boys and girls between 3–25 years old showed statistically higher sHLA-G values compared to men and women over 26 years old. It has recently been reported that mesenchymal progenitors and osteoblastic cells specifically express HLA-G5 during osteogenesis, with a key role in bone homeostasis
Taken together, these data support the theory of a conservation of UTR haplotype structure in populations with different origins and demographic history, such as Malians, French and Brazilians. These UTR haplotypes appear to be implicated in different sHLA-G expression patterns. Particularly, the association between the UTR-2 haplotype and low sHLA-G levels seems to be further confirmed and preserved in different populations, displaying a dominant negative effect of UTR-2. However, the allelic effect of HLA-G and HLA-A genes, independent from UTRs, seems to be implicated in sHLA-G modulation. Moreover, results on age and gender indicate that both of these parameters should be further investigated in studies involving sHLA-G expression. Finally these data suggest that sHLA-G production is not only regulated by UTR but also potentially by specific microenvironments. For exemple, UTR-1 and UTR-3 have been associated with different levels of malaria infection under a recessive model
These results may constitute essential elements on the one hand to optimize the selection of donors for organ transplantation and on the other hand the diagnosis and treatment of infectious and parasitic diseases. Further investigations in larger cohorts and in different populations are necessary not only to detect new functional polymorphisms in HLA-G regulatory regions, but also to reveal the extent of biological phenomena that influence sHLA-G secretion.
Median Joining (MJ) network on HLA-A∼HLA-G haplotypes constructed using the Network program(
(TIF)
Comparison between homozygous HLA-A*02 (associated to UTR-1/UTR-1, UTR-1/UTR-5 or UTR-5/UTR-5; mean = 183.5±25.83 UI/ml) and HLA-A*68 (associated to UTR-2/UTR-2 and UTR-2/UTR-5; mean 117.2±33.51 UI/ml) (p = 0.002).
(TIF)
HLA-A∼UTR∼HLA-G haplotype frequencies (Fq) estimated with the Gene[Rate] program in the Malian samples.
(DOCX)
Observed (Obs) and expected (Exp) frequencies, differences (Diff) and standardized residuals (StdRes) for HLA-A and HLA-G and for HLA-A and UTR.
(DOCX)
Soluble HLA-G mean and standard deviation (SD) are shown for each HLA-G UTR genotype.
(DOCX)
We are very grateful to all the donors who contributed to DNA and plasma samples.