IFNα induces CCR5 in CD4+ T-cells, causing its anti- HIV inefficiency and its subsequent pathogenic elevation, partially controlled by anti-HIV therapy

Like EC, we find that ART-treated patients control serum IFNα concentration and show few immune cell alterations enabling a healthy but fragile medical status. However, treatment interruption leads to elevated IFNα reflecting virus production indicating that like EC, ART does not achieve a virological cure. The immune system becomes overwhelmed by multiple immune cell abnormalities as found in untreated patients. These are chiefly mediated by elevated IFNα inducing signaling checkpoints abnormalities, including PD1, in cytotoxic immune cells. Importantly, during acute infection, elevated IFNα correlated with HIV load and we found that IFNα enhances CCR5, the HIV coreceptor in CD4+ T-cells, impairing its anti-viral response and accounting for the pathogenic vicious cycle: HIV → IFNα ↗ → infected CD4+ T-cells ↗ →HIV ↗. This study opens immunotherapeutic perspectives showing the need to control IFNα in order to convert ART patients into EC.


Introduction
Therapy for HIV infection was based on targeting specific steps in HIV replication and led to major clinical improvement and prognosis. However, there remain some medical issues (1,2,3,4,5,6) and the problem of residual HIV reservoirs (7,8), which lead to loss of virus control when therapy is interrupted (9,10). The chief clinical goal today is an attempt to reach a functional cure in which no further therapy is needed mimicking the elite controller (EC) status (0.5% of HIV-infected persons requiring no therapy for years or even decades) (11,12). To reach this end and in agreement with prior suggestions (13), we think a deeper understanding of pathogenesis may provide new insights for this purpose. In this context, we examined all blood immune cell types at different stages of an HIV immune reaction, comparing untreated HIV-infected EC with untreated infected non-EC (14). We found all immune cells to be altered 4 in non-EC, leading us to suggest that in addition to HIV, which directly targets only a few immune cell types, the involvement of one or more mediators. We identified the main mediator as elevated circulating pathogenic IFN, which is active in the earliest days of infection but avoided by ECs (14), possibly by their infection by a lower inoculum.
In this companion report, we further pursue these studies with a detailed analysis of all blood immune cells of HIV-infected non-EC that have received anti-retroviral therapy (ART) and compared these results with untreated patients. If therapy is interrupted in treated patients, HIV viremia returns demonstrating that anti-HIV therapy is not curative. We show here that therapy removes the bulk of these immune cell alterations while controlling IFN levels like EC but with some residual alterations, likely due to IFN effects before therapy. Importantly, we will also show the cause of the failure of IFN to contain HIV in the earliest days of infection in non-EC. This leads to a vicious cycle of IFN response to increasing HIV titers and ultimately to the pathogenic high levels of IFN In turn, this leads in non-EC to early immune cell damage, some of which persist even after IFN control by therapy. Notably, as in most infected patients, anti-HIV therapy did not begin in the earliest days of infection.

Results
The study seeks to unravel the reason that the anti-HIV effect of IFN is impaired. As detailed in Methods, we compare: 1) IFN and IFN2 serum concentrations, 2) distribution of immune cell subsets, 3) frequency of cells markers associated with immune dysfunction in treated and untreated infected non-EC patients.

Distinct blood immune cell profiles and serum IFN levels in treated patients (TP) compared to untreated patients (UP) and healthy donors (HD).
We first tested whether the major immune cell subtypes frequencies could distinguish TP from UP and HD by principal component analysis (PCA). UP and HD are clearly separated, while TP are distributed between both groups (Fig. 1A). We investigated the immune features that drive this TP immune pattern. No difference in the CD3 + T-cells frequency are observed between the studied groups (Fig. 1B1). However, like UP, TP show some decrease in CD4 + Tcells ( Fig. 1B2) with a concomitant increase of CD8 + T-cells (Fig. 1B3), compared to HD.
Furthermore, the frequency of  T-cells is reduced in TP compared to UP (Fig. 1B4). In addition, the frequency of Dendritic Cells (DC) is increased in TP, like in UP, compared to HD, though to a lesser degree (Fig. 1B5). Finally, the percentage of NK-cells is similar between TP and HD but is markedly reduced in UP (Fig. 1B6). The significant variation of these immune cell types frequencies across the three studied groups is summarized in the balloon plot (Fig.   1C).
We also compared the serum level of IFN and IFN between the three studied groups. TP and HD have highly significant lower serum IFN level than UP (Fig. 1D1). No remarkable changes in IFN concentration between the three groups is observed (Fig. 1D2). The variation of these IFNs in paired samples collected before and after cART confirmed that serum IFNα, but not IFNλ, disappears after treatment (Fig. 1D3-D4 and extended data Table 1). Interestingly, IFNα level is positively correlated with IFNλ2 level in TP as well in EC (14), but not in UP ( Fig. 1E). 6

High IFN level enhances in vitro expression of the HIV coreceptor CCR5.
Beta chemokines are the natural ligands of CCR5, and as such are potent inhibitors of HIV (15,16). We previously reported elevated IFN inhibits the production of these inhibitors of HIV (17). We hypothesized that a greater and more immediate impact would occur if IFN also directly increased the amount of HIV coreceptor CCR5. Indeed, we show here that IFN but not IFN2 enhances CCR5 mRNA expression on stimulated CD4 + T-cells isolated from HD ( Fig.2 A1), resulting in higher levels of the CCR5 protein on CD4 + T-cells (Fig.2A2). Similarly, IFN induces a dose-dependent increase in the expression of CCR5 on stimulated CD4 + T-cells isolated from UP, TP and EC (Fig. 2B).

Upregulation of CCR5 on CD4 + T-cells by IFN impairs its anti-HIV effect.
The CCR5 results suggested that the inhibitory effect of IFN on CCR5 virus replication is compromised, beginning at the onset of infection, and throughout the innate phase of the IR.
To test this hypothesis, we infected normal PBMCs with CCR5 (R5) or CXCR4 (X4)-tropic T/F HIV-1 and compared the inhibitory effect of IFN in IFN-pretreated and unpretreated cells. For both the CCR5 and CXCR4 viruses, the exponential phase of virus replication was initiated upon 24 hours after infection ( Fig. 2C1 and 2C2). For the R5-tropic virus, the level of IFN inhibition was reduced in IFN-pretreated cells compared to unpretreated cells. In the cells not pretreated, virus production was inhibited by 39.3%, 51.8% and 49.4% at 48 hours post infection at the concentration of 0.2 ng/mL, 1 ng/mL and 5 ng/mL, respectively (Fig. 2C1).
In comparison, in the pretreated cells, the same concentration of IFN only inhibited virus production by 16.3%, 22.9% and 38.5% at the same time point (Fig. 2C1). As expected for the X4-tropic virus, pretreatment with IFN had no effect in reducing the level of virus inhibition.
In contrast, pretreatment with 5 ng/mL IFN enhanced the inhibitory effect on the X4-tropic 7 virus (Fig. 2C2). A similar result was observed when PBMCs from a different donor were used (data not shown). These observations indicate that the upregulation of CCR5 by IFN compromises its inhibition of virus replication for R5 HIV-1, setting the stage for greater damage by both HIV and by a rising level of IFN.
Considering that IFN is correlated with HIV load, we assume that during the initial phase of infection elevated IFN infected CD4 + T-cells and HIV form a pathogenic vicious circle (HIV → IFN → infected CD4 + T-cells → more HIV → more IFN etc).

Persistence of some alterations in immune cell subtypes involved in the innate phase of IR in TP.
Other than IFN, components of the innate phase of the IR include particular cells, especially NK-cells. We found differences in the distributions of the 3 major NK-cell subsets (18,19) between UP, TP and HD (Fig. 3A1). The proportion of early NK-cells in UP and TP is lower than in HD. In addition, TP, like UP, but to a lesser extent, show an increase of terminal NK compared to HD (Fig. 3A2-3). We also found that NK-cell subsets from UP and TP display distinct immune profile associated with immune dysfunction. TP mature NK-cells still exhibit lower levels of Helios and NCR and increased levels of IFNα-induced CD38 and HLA-DR ( Fig. 3A4-5). In contrast, GrzB, CD26, CD39, iKIR and PD1 expression patterns on mature NK from TP and HD are similar ( Fig. 3A4 and 6). We next investigated the distribution of CD11c + mDC and CD123 + pDC subsets in the 3 groups (Fig. 3B1). Noteworthy, the pDC frequency is reduced in both patient groups, the decrease being greater in UP, while the mDC proportion is increased in these groups, the increase being higher in TP ( Fig. 3B2 and 3B3). The γδ T-cells, another subset of cells of an innate immune response, also show functional signaling receptor 8 expression abnormalities (Fig. 3C) with IFNα-induced HLA-DR, associated in some cases with CD38. These alterations are still significatively more frequent in TP than in HD, but less than in UP.
Collectively, these results show that immune cells from TP, involved during the innate phase of IR, still present functional abnormalities compared to HD, but at a lower level than UP.

Residual T-cell phenotypic profile abnormalities in TP
To compare immune T-cell profiles of UP and TP and HD, we characterized the T-cell subtype frequency and phenotypes, as described in material and methods and in the accompanying paper (14). UP have an abnormal T-cell distribution compared to HD. The frequency of CCR7 + in CD3 + , CD4 + and CD8 + T-cells ( Fig. 4A1-3) is decreased in UP compared to HD. This loss of T-cells expressing CCR7 is also present in TP, albeit at a lesser degree. UP also exhibit altered proportions of Naïve, CM, EM and TEMRA among CD4 + (Fig. 4B1-2) and CD8 + (Fig. 4B3-4) T-cell subsets compared to HD. TP still show an abnormal differentiation profile but with a normalization of their CM and EM CD8 + frequency. In addition, UP exhibit an altered phenotypic profile in T-cell subsets, compared to HD (Fig. 4C). This can be partly attributed to high IFNα levels (14). These phenotypic alterations are characterized by an abnormal level of activation/differentiation markers (CD38, HLA-DR, CD25, CD26, CD28) and inhibitory receptors (CD39, PD1, CTLA4) ( Fig. 4C and extended data Fig. 2). By and large, in contrast to UP, TP display similar immune profiles to HD. However, in TP, we still identified 9 phenotypes in CD8 + T-cells, such as HLA-DR + CM, CD38 + HLA-DR + EM, which remain significantly different to HD. The phenotypic analysis shows that after treatment partial reversion of T cell phenotype anomalies in all CD4 + and CD8 + T-cell subsets is observed, as evaluated by the phenotypic alterations scores (see Methods) ( Fig. 4C3-4). Interestingly, in UP, the expression levels of various markers associated with immune dysfunction in CD4 + and CD8 + CM population are directly correlated (extended data Fig. 2A4 and 2B4). This is in keeping with the interpretation that these are alterations induced by one major mediator such as elevated IFN. No such correlation is found after therapy.
The Treg phenotypic profile analysis revealed an increase frequency of Treg in UP compared to HD (Fig. 5A1-2). This increase in Treg frequency is, however, not seen in TP. In contrast, we found that after treatment, patients still have a high frequency of Treg lacking CD25 expression (20), and the frequency of this CD25 neg Treg variant is higher in UP ( Fig. 5A3-4).
However, the altered immune profile of memory Treg observed in UP is partially restored in TP (Fig. 5A5).
As to the CD8 + T-cell subsets, quantitative and qualitative defects of differentiated CTL (KIR − ) and CD8 + supp (KIR + ) among the TEMRA subset were found in HIV-infected patients (Fig.   5B). We found an increase of CTL with a concomitant reduction of CD8 + supp in TP compared to HD ( Fig. 5B1 and 5B3). These populations in UP and HD are similar. In addition, UP show a higher frequency of CTL and CD8 + supp cells expressing various markers of activation, differentiation and exhaustion compared to HD, while TP display almost similar phenotypic pattern to HD ( Fig. 5B2 and 5B4). We also found a correlation between IFNα-induced CD38 and expression of inhibitory checkpoints (CTLA-4 and PD1) in both CTL and CD8 + supp in UP, but not after therapy (Fig. 5B5).
In summary, the phenotypic anomalies associated with loss of immune cell function of Tconv and Treg cells is markedly reduced in TP compared to UP, though they are of similar nature as in UP, and in large part these are proteins induced by IFN. The CD8 + functionally-linked phenotypic alterations are also of the same nature as found in CD4 + T-cells, but far more, and are markedly reduced after therapy. These immune alterations in both Tconv and CD8 + cytotoxic T-cells observed before therapy and still present after cART may account for the fragile health indicated by the comorbidities that still occur in the TP.

TP and EC exhibit similarities in immune profiles
We next analyzed the cell blood immune profiles in TP compared to EC (Fig. 6). VISNE and PCA analysis indicate that TP and EC display similar cell distribution distinct from UP and HD ( Fig. 6A1-2). Both TP and EC also have normal background levels of serum IFNα. One difference between them, occurs with  T-cells ( Fig 6A3). No difference was found in the proportion of the NK-cell subsets among TP and EC, whereas a decrease of early NK-cells is observed in TP compared to HD (Fig. 6B1). Analysis of phenotypic abnormalities in mature NK-cells showed a higher frequency of CD38 + HLA-DR + and PD1 + cells in TP and EC than in HD (Fig. 6B2). Interestingly, TP and EC also share a similar percentage of DCs ( Fig. 6C1-2).
One difference observed in the DC compartment is a significant increase of mDC subset in TP.

11
Dysregulation of T-cells homeostasis, observed in untreated HIV-1 infected patients (21), is maintained in TP, albeit to a lesser degree, than in EC. TP have less CCR7 + CD4 + T-cells than HD (Fig. 6D1). In CD8 + T-cells, TP and EC have decreased frequencies of CCR7 + T-cells compared to HD (Fig. 6D2). Furthermore, the few T-cell phenotypic alterations of TP and EC are similar (Fig. 6E1 and 6F1 and extended data Fig. 3). Finally, analysis of Treg cells ( Fig.   6E2-3) revealed that their frequencies are similar in the three groups (Fig. 6E2). Noteworthy, TP and EC have an increase in the CD25 neg Treg variant compared to HD (Fig. 6E3). Within Collectively these data show that as opposed to UP, TP and EC do not express abnormal high serum IFNα level, avoiding the pathogenic IFNα effects on immune cells. Particularly notable were the marked improvement of the early effective NK-cells and the later developing effective antigen specific HLA-1 restricted CD8 + CTL and HLA-E restricted CD8 + supp in TP and EC. However, they do not completely return to the status of HD.

Discussion
The advances in therapy have derived from understanding the molecular events of HIV replication and targeting one or more stages with specific drugs. However, therapy is required lifelong except for EC. Except for developing long-lasting therapies and perhaps targeting the integrated HIV proviral DNA of HIV-reservoir cells, we need alternative approaches. In agreement with the EC consortium (13) we think a deeper penetration of HIV pathogenesis and learning how EC avoids these pathogenic mechanisms may be key to future advances in which 12 the >99% of infected non-EC patients can be converted to an EC status without treatment and possibly functionally cured.
There is an abundance of evidence showing that the key events leading to HIV progression and AIDS or a favorable clinical outcome are predicated on the events of early infection (22,23).
What that early damage is and its prognostic underpinning have been unclear. Its first signs are the HIV peak viremia and subsequent virus set point. These occur during the very initial stage of the immune response, the time of innate immunity. If not kept under tight control, a high HIV level leaves its stamp on future developments. We have known that HIV kills infected CD4 + T-cells, chiefly occurring after antigen activation (24). Though a small percentage of macrophages are also infected, most occur without cytopathic effects (25), and other immune cells, are not infected. It was evident, however, that there is much more to the story, and many reports show some other uninfected immune cells are not functioning normally. This suggests there is one or more unidentified critical mediator(s) of these effects, which is very early, thereby excluding specific adaptive immune responses, and not from the later inflammatory cytokines of the chronic stages. Instead, this mediator must be active at the time of the innate stage of the immune response (extended data Table 2).
Our results and the literature lead us to conclude that the main mediator is elevated levels of IFNα (14). Negative aspects of IFNα have been described, but usually as one of several cytokines and acting in some general way to promote inflammation as occurs in later stages.
The details in our reports show that IFNα is more important, early and specific. Another  Table 2) and as we previously reported also reduces the HIV inhibitory beta chemokines (17). We demonstrate that in untreated HIV infected non-EC patients all peripheral blood immune cells and at all stages of an immune response show abnormalities of distribution numbers and immune phenotype.
Importantly, these abnormalities can be induced by elevated IFNα levels, as described in the companion paper or in a few specific cases reported by us and others (28,29, 30). They include induction of several inhibitory checkpoints and immunosuppressive molecules, diminishing some essential positive immune signals, and interference with T-cell homeostasis and the subsequent adaptive immune response.
We show that EC and, to a lesser extent TP avoid these abnormalities of immune cell subsets  (Fig. 4). We suggest that the remaining defects in EC and in TP are the residue of the early phase of the IR due to the known thymic damages, whether IFNα-induced or not (extended data Table 2) and further in TP before cART (31,32). 14 IFN is a different story. Even in TP, IFNλ remains elevated suggesting persistence of local mucosal tissue reservoirs of HIV. However, IFNλ does not have detrimental effects on immune cells like IFNα because immune CD4 + T-cells do not have constitutive receptors for IFNλ (33), nor does it enhance CCR5 expression ( Fig. 2A). Consequently, IFN may be a useful agent against HIV even when present at high levels.
Evidence indicates that some EC are partially protected by genetic factors, particularly their HLA genetics, and most especially HLA-B57 (34). Indeed, it is tempting to think that these are the answers for explaining the EC state because genetic factors are, of course, present at the onset of infection, and the evidence dictating progression strongly favors an early event.
However, for explaining the entirety of the EC group, this cannot be true because known genetic factors are not present in many EC, and HLA-B57 is present in many HIV non-EC who demonstrate typical disease progression (34). A reduced CCR5 genetic trait has also been described but present only in a small number of EC (35). Our data argue against a single trait for becoming an EC because their minimal abnormalities are numerous and diverse.
Consequently, we have hypothesized that there may be an additional mechanism, which would also be at the onset: a fortuitous infection with a low inoculum of HIV. This would reduce the number of competing founder viruses and favor chances for a lower producer.
From the onset, this would mean for EC low HIV input → low set point HIV/AIDS is now a treatable disease, and for >99% of patients, therapy is a lifelong, sometimes with side effects, co-morbidities, and possibly a shortened life span. A goal in the field is to develop a functional "cure" without needing further therapy, but recognizing that some HIV cell reservoirs with "silent" proviral DNA may remain that can be re-awakened with therapy interruption or by external stimuli. Consequently, something more is needed. It is unlikely that many novel standard approaches, based on our detailed knowledge of the molecular events in HIV replication, are left for us to explore other than longer-lasting drugs and perhaps specific attacks on HIV proviral DNA. In agreement with the EC consortium (13) and its leadership, we suggest it may be rewarding in the quest for functional cure to focus on pathogenesis with an eye on therapies that diminish HIV progression and imitate the EC state.
In this regard, the work of the consortium defines a unique pattern of HIV integration, including spaced mono/oligoclonal clusters and HIV proviral DNA integration in silent DNA regions with a reduced frequency of escape mutations in cytotoxic epitopes and antibody contact regions (13,37), and they have suggested therapies geared toward reproducing this pattern in non-EC. We suggest that this pattern of HIV proviral integration may be an event not originally causing the control of HIV but rather the consequences of HIV control, but which nonetheless subsequently helps maintain that control by markedly diminishing HIV expression. The key question then is what is this original control? We hypothesize and our data indicate that it is the prevention of elevated IFN in the earliest stages, whether through immediate therapy or by a low inoculum initiating infection.
To conclude, IFN of course, is a key and an immediate protector against foreign invaders.
Sometimes it is also a well-known contributor along with several other cytokines to dangerous 16 inflammation in the later stages of infection by some viruses. However, our results here are directed to a new and an early as well as continuous solo pathogenic effect. These results place elevated IFN as a key direct mediator of HIV pathogenesis, in addition to the known direct HIV effect. The IFN effect occurs along with HIV early after infection and continues to AIDS progression unless treatment is initiated, or in the case of EC, as we hypothesize here, can be avoided because of low inoculum infection. We emphasize here, that, beside ART, therapy should include temporarily targeting elevated systemic IFN, as soon as HIV seropositivity is known, in order to reduce measurable HIV proviral DNA intact sequences down to the EC level (14,37), while concomitantly providing unharmful anti-viral IFN to control other viral infections and hopefully this approach will lead to the lack of the further need for ART, ie a functional cure, as a direction promoted by B. Walker and colleagues (13).

Online Methods
Human samples. HD were obtained through Etablissement Français du Sang (EFS, Paris, France). 67 people living with HIV were recruited and subdivided into three group: EC (n=18) were obtained from the NVS cohort (Baltimore), UP (n=36) were obtained from NIH (Bethesda n=19) and from the Laboratoire de Référence SIDA (Liège n=17) and TP (n=27) from NIH (Bethesda n=19) and from Laboratoire de Référence SIDA (Liège n=8). Patient group did not significantly differ in terms of age, gender, disease status. All participants or their surrogates provided informed consent in accordance with protocols approved by the regional ethical research boards and the Declaration of Helsinki. Clinical data are indicated in extended Table   3a-c. Unsupervised analyses were performed using cytobank software and R studio software.
Composite cell phenotypic alteration score. We generate a cumulative phenotypic score for each T-cell subsets. The frequency of the following markers was used to calculate the score: and not IFNλ1 or IFNλ3, we focused only on serum IFNλ2 levels (39).

Statistical analyses.
Statistical significance of differences between group was assessed using the unpaired nonparametric Mann-Whitney. Non-parametric, paired Wilcoxon tests were used for paired data. Correlations were assessed by the nonparametric Spearman test. Analyses were performed with GraphPad-Prism, and R. Two-sided P value less than .05 was considered statistically significant (ns: nonsignificant; *P < .05; **P < .01; ***P < .001; ****P < .0001).                  extended data Fig. 3