Identification of Siglec-1 null individuals infected with HIV-1

Siglec-1/CD169 is a myeloid-cell surface receptor critical for HIV-1 capture and infection of bystander target cells. To dissect the role of SIGLEC1 in natura, we scan a large population genetic database and identify a loss-of-function variant (Glu88Ter) that is found in ∼1% of healthy people. Exome analysis and direct genotyping of 4,233 HIV-1-infected individuals reveals two Glu88Ter homozygous and 97 heterozygous subjects, allowing the analysis of ex vivo and in vivo consequences of SIGLEC1 loss-of-function. Cells from these individuals are functionally null or haploinsufficient for Siglec-1 activity in HIV-1 capture and trans-infection ex vivo. However, Siglec-1 protein truncation does not have a measurable impact on HIV-1 acquisition or AIDS outcomes in vivo. This result contrasts with the known in vitro functional role of Siglec-1 in HIV-1 trans-infection. Thus, it provides evidence that the classical HIV-1 infectious routes may compensate for the lack of Siglec-1 in fuelling HIV-1 dissemination within infected individuals.

A ntigen presenting cells of the myeloid lineage, such as monocytes, macrophages and dendritic cells, initiate immune responses and are crucial to control invading viruses. In the case of HIV-1 infection, however, myeloid cells also promote viral pathogenesis through trans-infection of CD4 þ T cells 1,2 . This mechanism involves HIV-1 capture by sialic acid-binding Ig-like lectin 1 (Siglec-1/CD169), a myeloid-cell receptor that recognizes viral membrane gangliosides [3][4][5][6] . Viral capture facilitates the release of trapped viruses at a cell-to-cell contact zone promoting the trans-infection of CD4 þ T cells 7 . Immune activating signals, such as interferon alpha (IFNa) or bacterial lipopolysaccharide, are present throughout the course of HIV-1 infection 8 and induce Siglec-1 expression on myeloid cells 5,6,9 . However, under these inflammatory conditions, DC-SIGN and other C-type lectin receptors previously implicated in HIV-1 trans-infection 2,10 play a minor role in viral transmission 5,6,11 . Thus, Siglec-1 is an important inducible receptor that could accelerate HIV-1 transmission in lymphatic tissues, where many T-cells are in contact with myeloid cells. Recent studies performed in vivo in mouse models demonstrated that the robust infection of retroviruses in secondary lymphoid tissues requires Siglec-1 and does not rely on the mouse DC-SIGN homolog 12 . Siglec-1 trans-infection, cell-free virus infection and cell-to-cell viral transfer between infected and non-infected cells are all important routes of viral dissemination. Yet, the relative contribution of trans-infection to HIV-1 transmission and AIDS disease progression remains unknown.
There has been considerable interest in using human genetic diversity to dissect the role of various genes in defence against pathogens in natura 13 . Recently, protein-truncating variants have been catalogued in the human genome [14][15][16][17] . These are variants that are likely to disrupt the function of the corresponding allele 18,19 . The identification of the CCR5D32 variant was pivotal to understanding how human genetic variation contributes to differences in susceptibility to HIV-1 infection 20 , equally affecting all routes of viral spread.
Here, we aimed to identify SIGLEC1 null individuals to dissect the specific contribution of trans-infection to HIV-1 pathogenesis in vivo. We find individuals with a specific loss-of-function variant in SIGLEC1 gene that completely abrogates Siglec-1 receptor expression on primary monocytes and their capacity to trans-infect HIV-1 ex vivo. Despite this ex vivo phenotype there is a striking absence of marked differences in HIV-1 acquisition or clinical evolution of individuals carrying SIGLEC1 loss-of-function alleles.

Results
Genetic description of SIGLEC1 variants. We used data from the Exome Aggregation Consortium (ExAc.broadinstitute.org) to identify naturally occurring knockout mutations in SIGLEC1.
In that sample of B63,000 individuals we observed 70 protein truncating variants, that is, stop-gain, frameshift or splice site (Fig. 1). With the exception of rs150358287, a stop-gain variant resulting in an early stop codon at amino acid position 88 (Glu88Ter), truncating variants are of very low frequency (o1%). The Glu88Ter variant occurs in the second exon of SIGLEC1 (C to A transversion at position 3706494 on chromosome 20, GRCh 38 build reference sequence) and is predicted to truncate both major transcripts of SIGLEC1. The stop-gain allele is found at highest frequency in individuals of European and South Asian ancestry (B1.3%) and is rare or absent in African and East Asian populations (o0.5%).
To assess the frequency distribution of this polymorphism in an HIV-1-infected population, we combined genotype data from exome sequencing (n ¼ 392), exome chip (n ¼ 2,212) and direct genotyping (n ¼ 1,129) in participants of the Swiss HIV Cohort Study (SHCS). In 3,733 individuals whose clinical characteristics are detailed in Table 1 (95% reported European ancestry), we observed 85 Glu88Ter heterozygotes and 2 homozygotes for the stop-gain variant (allele frequency ¼ 1.2%). Thus, we identified individuals in whom to assess the consequences of Siglec-1 haploinsufficiency and knockout in vivo and ex vivo.
Functional analysis of SIGLEC1 null variant. To confirm that Siglec-1 Glu88Ter homozygous individuals truly lack receptor expression, we performed functional assays with cryopreserved cells collected from individuals with all three possible genotypes. We induced Siglec-1 expression in isolated monocytes using IFNa and determined the absolute number of Siglec-1 antibody binding sites per monocyte (Fig. 2a). Compared to individuals homozygous for the common allele, heterozygous individuals expressed approximately half of the amount of protein observed, while null individuals showed only background expression levels (Fig. 2a). Next, we analyzed the ability of monocytes to capture fluorescent HIV-1 virus-like particles (VLPs) displaying specific gangliosides that are efficiently recognized by Siglec-1 (refs 3,4). IFNa-activated monocytes from individuals with the common allele showed the highest viral capture capacity followed by heterozygous and then by null individuals (Fig. 2b), which captured only residual levels of VLPs. To investigate whether this binding was specific for Siglec-1, cells were pre-treated with a monoclonal antibody (mAb) against Siglec-1. Treatment led to a significant reduction of VLP uptake in monocytes from common allele and heterozygous individuals (Fig. 2b), while it had no inhibitory effect on SIGLEC1 null individuals. The VLP uptake of monocytes from distinct SIGLEC1 genotypes strongly correlated with the mean number of Siglec-1 antibody binding sites per cell (Fig. 2c). To assess the general HIV-1 transfer capacity of Siglec-1 compared to other possible receptors on IFNa-activated monocytes from homozygous individuals, we pulsed cells with equal amounts of infectious HIV-1 NL4-3 in the presence or absence of blocking mAbs and co-cultured them with a CD4 þ reporter cell line (Fig. 2d). Monocytes from individuals with the common allele had higher capacity to trans-infect than did SIGLEC1 null cells (Fig. 2d). Trans-infection was inhibited with a mAb against Siglec-1, which had no blocking effect on SIGLEC1 null monocytes (Fig. 2d). Overall, these results indicated that SIGLEC1 null individuals lack functional Siglec-1 expression and HIV-1 trans-infection capacity, ruling out genetic compensation mechanisms or a possible stop codon read-through that could alleviate the null status 21 .
SIGLEC1 null variant in HIV-1 acquisition. We next tested for an impact of the Glu88Ter variant on HIV-1 acquisition. Given the population frequency differences at which this variant occurs, we limited these analyses to 3,558 individuals of European ancestry to prevent confounding. The working hypothesis is that if Siglec-1 were essential for infection, homozygous Glu88Ter would not be infected-at least via mucosal exposure. The observation of two HIV-1-infected Glu88Ter homozygotes ruled out a requirement for a functional Siglec-1 protein in HIV-1 acquisition (that is, it does not mirror the CCR5D32 effect regarding R5-tropic virus infection). In addition, the observed frequency of the Glu88Ter allele in the HIV-1-infected population (1.2%) is nearly identical to the frequency in Europeans from the ExAc sample (1.3%) and does not differ depending on route of infection (1.15% parenteral, 1.2% sexual, P ¼ 0.95). Taken together, these results suggest that functional SIGLEC1 is not required for HIV-1 acquisition regardless of route of exposure.
SIGLEC1 null variant in HIV-1 progression. We assessed the potential impact of the SIGLEC1 null variant on HIV-1 viral load and disease progression. As shown in Fig. 3, the presence of one or two copies of the Siglec-1 Glu88Ter allele had no impact on plasma set point viral load (n ¼ 2,243; Fig. 3a). Similarly, there was no difference in CD4 þ T-cell dynamics (n ¼ 2,302; Fig. 3b) or in CD4 þ T-cell nadir. We also investigated the disease course (viral RNA level and CD4 þ T-cell counts) for the two homozygous Glu88Ter individuals, which supported viral replication (Fig. 3c). Finally, we assessed the impact of the Glu88Ter variant on progression to AIDS (as defined in 1987) in the SHCS (Fig. 3d), where we analyzed 52 Glu88Ter heterozygous  and 1 homozygous individuals out of 2,511 individuals. We performed the same analysis in an independent cohort (the Multicenter AIDS Cohort Study, MACS) after genotyping 413 Caucasian HIV-1-infected individuals with documented seroconversion date (Fig. 3e). In the MACS cohort, we identified 12 heterozygous individuals, but no homozygous Glu88Ter individuals were found. Although there was a slow progression to AIDS observed in both the SHCS and MACS cohorts for Glu88Ter individuals, it did not reach statistical significance.

Discussion
The identification of two homozygous HIV-1-infected individuals with a confirmed loss-of-function variant in SIGLEC1, both of whom were infected via mucosal exposure, indicates that Siglec-1 trans-infection is not indispensable to establish HIV-1 infection through sexual contact. Thus, Siglec-1 independent mechanisms of infection are sufficient to support mucosal HIV-1 spread. In the absence of trans-infection, the classical HIV-1 infectious routes, including cell-free virus infection or cell-to-cell HIV-1 transmission, compensate for the lack of Siglec-1 and are able to fuel HIV-1 dissemination within infected individuals. This explains why we did not identify marked differences in clinical evolution of individuals with one or two copies of the Siglec-1 Glu88Ter allele. Given the available sample size of this study and the low Glu88Ter frequency, we can only rule out a large effect of this allele on disease progression. Power simulations indicate that we would need 410,000 samples to detect a relative risk of 5 (similar to the effect of B*57:01 on HIV-1 control) at Po0.05 under a recessive model. This sample size far exceeds even the largest genome-wide studies of HIV-1 progression that comprises B6,000 patients 22 , which unfortunately does not genotype the Glu88Ter variant and cannot be used to accurately impute the presence of this rare allele. In addition, given that the proposed effect (if any) requires long-term follow-up off therapy (410 years) it is extremely unlikely that a sufficient sample size could be reached to assess the long-term consequences of this variant on HIV-1 disease. However, the identification of the functional consequences of the Glu88Ter allele might encourage future genetic studies to include this variant and complement our analyses.
These results in humans are in contrast to those observed in the BLT humanized mouse model, where Siglec-1 blockade significantly lowered HIV-1 infection of splenocytes 12 . Discrepancies could be attributed to the fact that our study mainly focused on heterozygous Siglec-1 Glu88Ter individuals, which maintain partial receptor expression and function. Yet, the clinical course of the two homozygous individuals did not show any obvious benefit of absence of Siglec-1. Alternatively, differences with the murine model could be due to the distinct timeframes analyzed in each case. The murine study was limited to the early dynamics of HIV-1 infection 12 , a phase that is missing from the clinical records of most patients, including the two null-homozygous individuals identified here. Hence, it will be important to determine if the lack of Siglec-1 in humans impacts early viral kinetics. If that is the case, current antiretroviral treatments implemented early after infection could include Siglec-1 blocking agents to tackle multiple infectious routes simultaneously and limit the settlement of viral reservoirs. Importantly, the identification of SIGLEC1 null individuals in natura indicates that Siglec-1 could represent a safe therapeutic target.
The analysis of loss-of-function variants serves to illuminate in vivo mechanisms of disease. Recent whole-genome and -exome sequencing initiatives indicate that protein truncating alleles can be identified in a large proportion of human genes. We have estimated that up to 85% of all human genes can be observed carrying heterozygous variants 17 . There is strong evidence that heterozygous loss-of-function in innate immunity genes is not compensated in the majority of cases 16 . Thus, the discovery and characterization of human protein truncating variants serves as a general model for the analysis of other human genes involved in pathogen containment.

Methods
Patients. Siglec-1 surface expression analysis by FACS. IFNa-activated monocytes were blocked with 1 mg ml À 1 of human IgG (Baxter, Hyland Immuno) and stained with 1/10 dilution of a-Siglec-1-PE 7-239 mAb (AbD Serotec) following manufactures instructions at 4°C for 20 min. Samples were analyzed with FACSCalibur (Becton-Dickinson) using CellQuest software to evaluate collected data. The mean number of Siglec-1 Ab binding sites per cell was obtained with a Quantibrite kit (Becton-Dickinson) as previously described 5 .
VLP capture and HIV-1 trans-infection assays. VLP HIV-Gag-eGFP and HIV-1 NL4-3 were obtained by calcium phosphate transfection of previously described plasmids 5 . IFNa-activated monocytes (2 Â 10 5 ) were pre-incubated at 16°C for 30 min with 10 mg ml À 1 of a-Siglec-1 mAb (7- Statistical analysis. Statistics of VLP functional assays were performed using unpaired and paired t test (considered significant at Pr0.05) or the Spearman correlation with GraphPad Prism v.5 software. Set point viral load was calculated as the average of at least three measurements obtained during the chronic phase of infection (minimum 6 months after infection and before the initiation of antiretroviral therapy (ART)). A comparison of variance in set point viral load between genotype groups showed no significant differences (F test P ¼ 0.18). Association between viral RNA level and genotype was tested using linear regression. Rate of disease progression was measured using all available CD4 þ T-cell counts beginning at enrolment or estimated date of infection (if known) before the initiation of anti-HIV therapy. The impact of genotype on disease progression was tested using the Cox proportional hazards models.
Data availability. The clinical and genetic data that support the findings of this study are available on request from the corresponding authors; these data are not publicly available due to information that could compromise research participant privacy. The authors declare that all other data supporting the findings of this study are available within the article or from the corresponding authors upon request.