Quasispecies tropism and compartmentalization in gut and peripheral blood during early and chronic phases of HIV-1 infection: possible correlation with immune activation markers

Authors


Abstract

HIV quasispecies was analysed in plasma and proviral genomes hosted by duodenal mucosa and peripheral blood cells (PBMC) from patients with early or chronic infection, with respect to viral heterogeneity, tropism compartmentalization and extent of immune activation. Seventeen HIV-1-infected combined antiretroviral therapy naive patients were enrolled (11 early infection and six chronic infection). V3 and nef genomic regions were analysed by ultra-deep pyrosequencing. Sequences were used to infer co-receptor usage and to construct phylogenetic trees. As markers of immune activation, plasma sCD14 and soluble tumour necrosis factor receptor II (sTNFRII) levels were measured. Median diversity of HIV RNA was lower in patients with early infection versus chronic infection patients. Overall, direct correlation was observed between V3 diversity and X4 frequency; V3 diversity of HIV RNA was inversely correlated with CD4 T-cell count; median sCD14 and sTNFRII values were similar in early and chronic patients, but X4 frequency of HIV RNA was directly correlated with plasma sCD14. The proportion of patients harbouring X4 variants and median intra-patient X4 frequency of proviral genomes tended to be higher in chronic infection than early infection patients. More pronounced compartmentalization of proviral quasispecies in gut compared with PBMC samples was observed in patients with early infection compared with chronic patients. The loss of gut/PBMC compartmentalization in more advanced stages of HIV infection was confirmed by longitudinal observation. More studies are needed to understand the pathogenetic significance of early HIV quasispecies compartmentalization and progressive intermixing of viral variants in subsequent phases of the infection, as well as the role of immune activation in tropism switch.

Introduction

Infection with HIV-1 typically involves the interaction between the viral envelope glycoprotein gp120 and the CD4 molecule, followed by interaction with a chemokine receptor, usually CCR5 or CXCR4. CCR5-using (R5) viruses predominate in the early stages of HIV-1 infection, whereas a switch towards CXCR4 usage (X4 viruses) may occur at later stages. While the propensity to X4 switch varies according to HIV clade, the presence of X4 variants is invariably associated with a faster disease progression [1]. Several studies, based on either phenotypic tests or bulk sequencing, suggest that up to 19.2% of acutely infected patients may harbour X4 variants [2-6]. Thanks to the newly introduced ultra-deep pyrosequencing (UDPS), which allows high-resolution analysis of viral quasispecies, recently it has been shown that the proportion of patients with primary infection harbouring X4 variants is larger than previously anticipated [7]. In addition, these technologies allow us to quantify the relative frequency of even rare variants with different co-receptor usage, as compared with the dominant virus population [8-11]. In a recent study, about half of acutely infected patients were shown to harbour X4 strains by UDPS, with highly variable intra-patient frequency, and worsened clinical course in those with a high burden of X4 variants [7].

In the first phases of the infection, independently of the route of transmission, early and massive viral replication occurs in gut-associated lymphoid tissue, leading to irreversible mucosal CD4 T-cell depletion. A complex network of HIV-1-target cells is established during this phase, leading to a progressive filling of reservoirs and to viral dissemination in the body [12].

There is growing evidence that compartmentalization may be an early event during the natural history of HIV-1 infection [13, 14]. Some pathogenetic models suggest that a status of immune hyperactivation represents one of the main predictors of disease progression [15]. Recently, van Marle et al. [16] showed discordant results for compartmentalization of virus replication within different gut locations and peripheral blood mononuclear cells (PBMC) of chronically infected patients naive to antiretroviral therapy, depending on the genomic region analysed (nef and pol). In another report on gut-associated lymphoid tissue versus PBMC, compartmentalization of viral quasispecies was highlighted on the basis of different drug resistance-associated mutations [17]. In other studies, no evidence of compartmentalization was observed between gut and peripheral blood and even within different gut regions (colon versus ileum) in chronically HIV-1-infected individuals [18]. At the same time, evidence of cross-infection between gut-associated lymphoid tissue and PBMC during combined antiretroviral therapy (cART) has been suggested as a possible mechanism for HIV persistence [19].

The present study was aimed at studying, by UDPS, the tropism and compartmentalization of HIV-1 quasispecies in gut and peripheral blood in cART-naive patients with early infection compared with chronically infected patients, analysing possible correlations between the presence of X4 variants and the extent of immune activation.

Methods

Patients

Seventeen HIV-1 infected patients, for whom there was clinical indication to perform a duodenal biopsy for diagnostic purposes, were consecutively enrolled in the study. Eleven subjects had a primary HIV infection, defined by a negative or indeterminate HIV-1 Western Blot with simultaneous positive plasma HIV viraemia or by a positive HIV-1 antibody test preceded by a documented negative test within the preceding 180 days (early infected patients); six patients had an HIV-1 infection documented for ≥3 years (chronic patients). At the moment of biopsy, all patients with primary infection were in Fiebig stage VI. At the same time-point, blood samples were also collected (PBMC and plasma). All patients had acquired HIV infection by a sexual route and were naive to cART at enrolment. One early patients who underwent cART after enrolment, was analysed at subsequent time-points (6 and 12 months from enrolment). The study was approved by the Institutional Ethics Committee, and the patients agreed to participate by signing an informed consent.

UDPS and data analysis

Total DNA extraction from PBMC and from duodenal biopsies that were first homogenized in cell lysis buffer (Qiagen, Hilden, Germany) using a pestle, was performed using the extraction kit ‘DNA blood’ (Qiagen). Plasma HIV-1 RNA was extracted using the QIAamp Viral RNA kit (Qiagen). The number of templates actually undergoing UDPS analysis was evaluated by a commercial real-time PCR for HIV-1 RNA in plasma samples (Abbott real-time HIV, Abbott Molecular Inc., Des Plaines, IL, USA), and by a quantitative real-time PCR targeting the long terminal repeat region for HIV DNA [20].

V3 amplification was performed by nested PCR. Briefly, two rounds of 30 cycles (94°C for 2 min, 94°C for 30 s, annealing at 60°C for 30 s, extension at 68°C for 30 s and final elongation at 68°C for 5 min) were carried out using a proofreading DNA polymerase (Platinum® Taq DNA Polymerase High Fidelity; Invitrogen, by Life Technologies, Monza, Italy). First- and second-round primers were described in ref. [7]. Nef amplification was performed by a nested PCR, as described in ref. [16]. For plasma samples, the first round of both env and nef included a one-step RT-PCR, using a Platinum quality proofreading reverse transcriptase (Invitrogen). Unique in-house-designed stretches of eight nucleotides (Multiplex Identifiers), were used to tag each sample. To maximize the genetic heterogeneity to be amplified and sequenced, for each sample the amplicons from at least four replicate PCR were pooled, representing, for viral RNA, the content of 1 mL of plasma, and, for DNA, the content from 2 × 106 to 6 × 106 PBMC. To minimize the occurrence of artefacts attributable to template re-sampling, a minimum of 1200 templates were analysed by UDPS, on the basis of viral and proviral load (Table 1).

Table 1. Demographic, clinical and virological features of the study patientsa
 Early infected (n = 11)Chronically infected (n = 6)pb
  1. Abbreviations: PBMC, peripheral blood mononuclear cells; sCD14, soluble CD14; sTNFRII, soluble tumour necrosis factor receptor II.

  2. a

    All data are reported as median (interquartile range, IQR), except for gender, where ratio M/F is reported.

  3. b

    Significance was calculated by Mann–Whitney non-parametric test or chi-squared test, as appropriate. p values <0.05 were considered significant.

Age (years)33.0 (24.0–37.0)39.5 (37.0–44.5)0.050
Gender (M/F)10/16/00.751
Days from diagnosis40.0 (28.0–69.0)568.0 (106.0–3402.0)0.004
CD4 T (cell/μL)563 (417–1251)383 (77–548)0.056
Nadir CD4 T (cell/μL)414 (219–584)350 (77–380)0.040
Plasma viral load (HIV-RNA Log copies/mL)5.0 (4.3–5.8)4.9 (4.5–6.1)0.920
PBMC proviral load (HIV-DNA Log cp/million cells)3.8 (3.4–4.0)3.7 (3.1–4.5)1.000
Gut proviral load (HIV-DNA Log cp/million cells)3.2 (2.8–4.2)3.2 (2.9–3.6)0.841
sCD14 (µg/mL)1.9 (1.7–2.1)2.2 (1.6–2.6)0.679
sTNFRII (ng/mL)5.1 (4.0–6.8)6.0 (4.9–6.8)0.371

The UDPS was carried out with the 454 Life Sciences platform (GS-FLX, Roche Applied Science, Monza, Italy) as described previously [7], using Titanium chemistry. For V3 amplicons the correction pipeline and the evaluation of experimental error necessary to establish the sensitivity threshold have already been described [7]. Considering the number of viral templates actually undergoing UDPS (at least 1200 copies) and the corrected error rate that we had previously calculated for V3 (i.e. 0.058%, [7, 8]), a sensitivity threshold of 0.3% was adopted for this region—i.e. a value exceeding by at least five-fold the error rate; all nucleotide changes detected with a frequency ≤0.3% after the correction pipeline were ignored. For nef amplicons, consensus reads were directly extracted using GS Amplicon Variant Analyzer software (v.2.5.3; Roche); each consensus read corresponds to sets of individual reads that were collapsed into a single representative read. In this case, the threshold of sensitivity was conservatively set to 0.5%.

The nucleotide sequences resulting from the correction pipeline (from nucleotide (nt) 7010 to nt 7332 of reference HXB2 genome positions for V3 and from nt 8758 to nt 9454 of reference HXB2 genome positions for nef amplicons), were also analysed to establish genetic heterogeneity (diversity, i.e. mean number of nucleotide substitutions/site) of the viral quasispecies using DNADIST (F84 algorithm from Phylip package). To construct individual phylogenetic trees, maximum likelihood (ML) phylogenetic trees were inferred with the MEGA program, using the GTR+G nucleotide model. Bootstrap values >80% were considered significant.

For co-receptor usage prediction, V3 sequences obtained by UDPS were submitted to the genotypic predictor algorithm geno2pheno [454] (http://454.geno2pheno.org/index.php). The false-positive rate was set to 5.75%, corresponding to the false-positive rate suggested for population sequencing of RNA by retrospective re-analysis of the Maraviroc licensing studies [21, 22], and intra-patient X4 frequency was calculated for each sample type.

Immune activation markers

Commercially available ELISA kits were used to quantify plasma levels of sCD14 and soluble tumour necrosis factor receptor II (sTNFRII; R&D Systems Europe, Abingdon, UK). Plasma samples from 14 patients were available for this analysis. The samples were diluted 1 : 200 and 1 : 20 for sCD14 and sTNFRII, respectively, using the Calibrator Diluent provided in the kit, including Triton X (final concentration 0.1%), to produce samples with values within the dynamic range of the assay.

Histological evaluation

Duodenal biopsy specimens obtained endoscopically were fixed in a 4% buffered formalin for 24 h, dehydrated and embedded in paraffin. The tissues were analysed on the basis of morphological and immunohistochemical analysis. The villous/crypt ratio and the numbers of lymphocytes in the intraepithelial region were established. Staining for CD3 (Clone 2GV6) and CD4 (Clone SP 35) was carried out using a Ventana ES automatic immunostainer (Ventana Medical System Inc., Tucson, AZ, USA). The sections were counterstained with haematoxylin & eosin, and ten randomly selected areas per slide were scored using the 20× objective with 100 enterocytes counted in each field in the surface epithelium on the top of the villi.

Statistical analysis

Mann–Whitney and Spearman rank correlation non-parametric tests were used as appropriate. For statistical calculation, an X4 frequency equal to the sensitivity threshold (i.e. 0.3%) was assigned to the X4-negative samples.

Results

Clinical features, genetic heterogeneity and X4 variants in the viral quasispecies associated with gut and peripheral blood

Demographic, clinical and virological characteristics of study patients are reported in Table 1. Nadir CD4 T-cell counts were significantly lower (p 0.039), while median values of plasma viral load, proviral load in PBMC and gut biopsy, sCD14 and sTNFRII were similar in patients with chronic infection compared with early infection patients.

To evaluate genetic heterogeneity, genetic distances were calculated from env V3 and nef sequences. UDPS analysis of the V3 region provided a total of 69 080 edited sequences for plasma HIV RNA, 82 210 for PBMC-associated proviral DNA and 113 162 for gut biopsy proviral DNA. The UDPS analysis of nef was performed on a subgroup of patients (in four early patients the analysis was performed on all the three sample types, whereas in eight early and in five chronic patients the analysis was conducted only on plasma samples). The total edited nef sequences were 5639 for plasma HIV, 3945 for PBMC proviral DNA and 2014 for gut proviral DNA.

Considering all sample types, a direct correlation was observed between V3 diversity and X4 frequency (r = 0.665, p <0.001). An inverse correlation was observed between V3 diversity of plasma viral RNA and CD4 T-cell counts (r = −0.661, p 0.005).

The distributions of intra-patient values of V3 genetic diversity and of X4 frequencies in gut, PBMC and plasma from early and chronic patients are shown in Fig. 1. Median values of V3 diversity for all three sample types were lower in early patients compared with chronic patients (Fig. 1a–c), reaching statistical significance for plasma HIV RNA values (median values 0.003 versus 0.024 mean substitutions/site, p 0.011). Consistent with V3 diversity, median values of nef diversity in plasma HIV RNA were also significantly lower in early patients compared with chronic patients (0.010 versus 0.021, p 0.019, not shown in the figure).

Figure 1.

HIV diversity and intra-patient frequencies of X4 variants of V3 region in gut and peripheral blood from patients with early and chronic infections. The genetic diversity of V3 (a–c) and intra-patient frequencies of X4 variants (d–f) in the three sample types (gut, peripheral blood mononuclear cells and plasma) from early and chronic patients are reported. Bars indicate median values. In panels d–f, dotted line represents the sensitivity threshold for X4 frequency (0.3%). Statistically significant difference between the two groups is indicated. [This figure was corrected on 25/11/2013 after original online publication 18/10/2013. The title of panel C previously read “Plasma proviral RNA”. This has been amended.]

The proportion of patients harbouring X4 variants (as either minor or predominant variants) in at least one compartment tended to be lower in early versus chronic patients (45% versus 67%); consistently, intrapatient frequencies of X4 variants in the proviral DNA tended to be lower in early compared with chronic patients (Fig. 1d–f), although not reaching statistical significance.

Considering immune activation markers, plasma levels of sCD14 and sTNFRII were directly correlated (r = 0.729, p 0.002). Of note, considering all 14 patients where the activation markers could be evaluated, plasma levels of sCD14 were directly correlated with the X4 frequencies present in circulating plasma RNA quasispecies (r = 0.602, p 0.023). No correlation was observed between proviral load in gut biopsies and plasma levels of sCD14.

Compartmentalization of viral replication in early and chronic patients

To analyse viral compartmentalization between gut and peripheral blood, phylogenetic trees were constructed for those patients who showed a genetic diversity >0.02 substitutions/site in V3 region in at least one compartment (six early and four chronic patients).

V3 compartmentalization of proviral quasispecies in gut and PBMC proviral DNA was more pronounced in early than in chronic patients (evident compartmentalization in five of six early patients versus one of four chronic patients). In Fig. 2 the phylogenetic trees built on V3 and nef UDPS sequences from two representative early patients (#1 and #2) are shown; in Fig. 3 the V3 sequences from a representative chronic patient is shown. As can be seen in Fig. 2(a), in one early patient the vast majority of viral quasispecies variants in all three compartments are R5, and are grouped into two well-defined clusters, including either proviral PBMC DNA or proviral gut DNA + plasma RNA. So, in this patient the replicating viral components are originated presumably by the proviral population from gut and not from PBMC. In this patient only gut-associated proviral DNA harbours X4 variants, as minority quasispecies (0.58%). In Fig. 2(c), a strong compartmentalization between gut and PBMC is shown for another early patient, where the proportion between R5 and X4 variants is reversed compared with the patient from Fig. 2(a). In fact, replicating virions are essentially X4, and are closely related to proviral genomes harboured by gut tissue, with only a minimal contribution by the X4 genomes harboured by PBMC. On the other hand, the vast majority (97%) of PBMC variants displaying R5 characteristics do not seem to contribute to replicating plasma viral RNA. The phylogenetic trees built with nef sequences from the same patients (Fig. 2b,d) strongly support the compartmentalization highlighted by V3 analysis. The phylogenetic tree built with the env sequences from one typical chronic patient (Fig. 3) clearly shows intermixed viral quasispecies components from the three compartments.

Figure 2.

Phylogenetic trees built with sequences from two representative patients with early infection(#1 and #2). The phylogenetic tree constructed with V3 and nef sequences are reported in (a) and (b) for early patient #1, and in (c) and (d) for early patient #2, respectively. Only bootstrap values >80% are indicated. Patient #1. (a) R5 proviral DNA sequences from peripheral blood mononuclear cells (PBMC) are shown in blue (sequence name); R5 proviral DNA sequences from gut biopsy are shown in red (sequence name) whereas X4 sequences in orange (sequence name); R5 sequences from plasma viral RNA are shown in green (sequence name). (b) Proviral DNA sequences from PBMC are shown in blue (sequence name); proviral DNA sequences from gut biopsy are shown in red (sequence name); sequences from plasma viral RNA are shown in green (sequence name). Patient #2. (c) R5 proviral DNA sequences from PBMC are shown in blue (sequence name) whereas X4 proviral DNA sequences in turquoise (sequence name); R5 proviral DNA sequences from gut biopsy are shown in red (sequence name) whereas X4 proviral DNA sequences in orange (sequence name); X4 sequences from plasma viral RNA are shown in light green (sequence name). (d) Proviral DNA sequences from PBMC are shown in blue (sequence name); proviral DNA sequences from gut biopsy are shown in red (sequence name); sequences from plasma viral RNA are shown in green (sequence name).

Figure 3.

Phylogenetic tree built with V3 sequences from a representative patient with chronic infection. R5 proviral DNA sequences from peripheral blood mononuclear cells are shown in blue (sequence name) whereas X4 proviral DNA sequences in turquoise (sequence name); R5 proviral DNA sequences from gut biopsy are shown in red (sequence name) whereas X4 proviral DNA sequences in orange (sequence name); R5 sequences from plasma viral RNA are shown in green (sequence name). Only bootstrap values >80% are indicated.

In one early patient, who started successful cART soon after the first sampling, it was possible to perform longitudinal analysis over a period of 1 year. As shown in Fig. 4, HIV proviral variants segregated in different clusters between PBMC and gut at the time of early infection. At 6 months, PBMC quasispecies included additional variants, corresponding to time 0 (T0) gut clusters. After 1 year, gut/PBMC compartmentalization was no longer observed, with common clusters present in both compartments, although with different frequencies. The progressive loss of gut/PBMC compartmentalization was paralleled by the enrichment of X4 variants in proviral DNA from both PBMC (from 4% to ≥90%) and gut (from 1% to 23%), despite complete cART-driven suppression of plasma viral load. It is noteworthy that at T0 the X4 variants present in circulating RNA corresponded mainly to T0 PBMC clusters, whereas R5 variants corresponded mainly to T0 gut clusters.

Figure 4.

Longitudinal evaluation of V3 quasispecies compartmentalization during 1 year in one representative patient with early infection. (a–c) Phylogenetic trees built with V3 sequences from patient #3 obtained at different time-points (T0, a; T6, b; T12, c) from different compartments. Proviral DNA sequences from peripheral blood mononuclear cells (PBMC) are shown in turquoise (sequence name) at T0; in light blue (sequence name) at T6; in dark blue (sequence name) at T12. Proviral DNA sequences from gut biopsy are shown in red (sequence name) at T0; in bordeaux (sequence name) at T12. Sequences from plasma viral RNA are shown in light green (sequence name) at T0. X4 sequences are marked with an asterisk. Individual clusters of variants are identified with a number. Only bootstrap values >80% are indicated. (d) Distribution of variants according to individual clusters identified in (a–c). To highlight low frequency variants, the vertical axis scale has been censored to 12; individual frequency of variants is indicated on each column. The colour code is the same used in (a–c); X4 frequencies are marked with an asterisk.

Histological evaluation, conducted on a subgroup of patients (two early and one chronic), suggested a trend towards a more preserved anatomical structure of gut epithelium in early versus chronic patients. Fig. 5(a,b) shows a nearly normal mucosa from an early patient, with subtle reactive epithelial changes: the epithelium is infiltrated by neutrophils with inflammation in the lamina propria, without significant loss of villous architecture, and normal villous height to crypt depth ratio of 3 : 1 (Fig. 5a). The total number of CD3+ intraepithelial lymphocytes (IEL) was normal, with a ratio of about 20 lymphocytes per 100 epithelial cells and two CD4+ lymphocytes per 20 CD3+ IEL (Fig. 5b). In Fig. 5(c,d), the duodenal mucosa from another early patient is shown, with normal villous architecture (Fig. 5c), with 25 IEL per 100 epithelial cells and three CD4+ lymphocytes per 20 CD3+ IEL (Fig. 5d). In contrast, the duodenal mucosa sample from a chronic patient (Fig. 5e,f) shows mild blunting of villi and elongated hypertrophic crypts (Fig. 5e). More than 29 CD3+ IEL per 100 epithelial cells were counted, and CD4+ IELs were not detected (Fig. 5f).

Figure 5.

Immunostaining of duodenal biopsy from early and chronic patients with anti-CD3 and anti-CD4. (a, c, e) Immunostaining with anti-CD3 of duodenal biopsy from two patients with early infection (a, c) and one chronically HIV-infected patient (e) (200 ×, 100 × and 100 ×, respectively). (b, d, f) Immunostaining with anti-CD4 of duodenal biopsy from two patients with early infection (b, b) and one chronically HIV-infected patient (f) (400 ×, 200 × and 400 ×, respectively).

Discussion

From this proof of concept study, based on ultrasensitive evaluation of HIV quasispecies in two small groups of patients (early versus chronic infection), the presence of X4 strains appears to be common in all compartments at both stages of infection. The presence of X4 variants in the gut is not unexpected, as CXCR4 expression has been demonstrated on both naive (CD45RO) and memory (CD45RO) CD4 T cells from gut mucosa. In addition, X4 HIV isolates have been reported to be able to actively replicate in these cells, although to a lesser extent than R5 isolates [23, 24]. Our findings also suggest the accumulation of viral diversity during chronic phases of the infection, in line with previous studies [25]. Interestingly, compartmentalization of viral quasispecies between gut and peripheral blood was typically observed in early patients. This finding suggests that gut compartmentalization of HIV quasispecies is an early event during the natural history of the infection, as already reported for other body compartments, such as central nervous system, possibly as a consequence of divergent evolution of HIV lineages associated with trafficking of lymphomonocyte clones [26].

In contrast, an intermixed pattern of HIV quasispecies distribution between peripheral blood and gut tissue was the most frequent condition observed in chronic patients, suggesting that during disease, progression compartmentalization tends to be lost, in parallel with increased viral diversity and X4 frequency. In fact, the longitudinal observation of one patient during 1 year after primary infection strongly supports this interpretation, showing progressive loss of gut/PBMC compartmentalization, paralleled by increased X4 variant frequency, despite effective suppression of plasma viraemia.

Lack of compartmentalization has been observed in chronic patients under long-lasting effective cART [19]. In contrast with these results, in another study based on patients on prolonged suppressive cART, a clear segregation of plasma and CD4 T-cell-derived viral sequences has been shown, with low-level plasma HIV produced mostly by cells other than CD4 T cells [27]. Interestingly, in this study the plasma sequences obtained from two patients before and after therapy interruption were apparently produced by the same compartment, which was different from the circulating CD4 T cells. The authors suggest that plasma viruses in patients on long-term suppressive cART may be persistently released from a cellular source yet to be identified, capable of spreading quickly in vivo, accounting for the rapid rebound of plasma viral loads after therapy interruption.

From the present study it is not possible to infer whether the loss of quasispecies compartmentalization observed in chronic patients was connected with the loss of integrity of gut architecture, because histological evaluation of intestinal biopsy was performed in only one chronic patient.

Another interesting finding from the present study is the positive correlation between X4 variant frequency and extent of immune activation assessed by indirect soluble markers (sCD14 levels). In a recent study, a direct correlation between proviral load in rectosigmoid biopsies and plasma levels of a marker of immune activation, i.e. LPS, has been observed in virologically suppressed patients [28]; however, in the present study no correlation between proviral load in gut biopsies and plasma levels of sCD14 was observed. This apparent discrepancy may be accounted for by the fact that in our study the immune activation markers were evaluated in untreated patients.

More studies are needed to better understand the pathogenetic significance of early HIV quasispecies compartmentalization in the gut, plasma and peripheral mononuclear cells, the role of immune activation on tropism switch and, finally, the clinical significance of the progressive intermixing of viral variants in subsequent phases of the infection.

Acknowledgements

This work has been partially supported by Grants to MRC and IA from Istituto Superiore di Sanita' (National AIDS Project), and by the Italian Ministry of Health (Fondi Ricerca Corrente and Ricerca Finalizzata).

Authors’ Contributions

GR, IA and MRC designed the study, wrote and drafted the manuscript. GR, PZ and BB performed molecular experiments. EG and MS collected and assembled the data. CV, AC, GD, RL were the clinical referents of the study. AB and FD performed histopathological evaluation. All authors read and approved the final manuscript.

Transparency Declaration

The authors declare that they have no competing interests.

Ancillary