The evolution of HIV-1 entry phenotypes as a guide to changing target cells

evolutionaryprecursortoX4Tcell-tropicand macrophage-tropicvariants. Abstract Through a twist of fate the most common form of HIV-1, as defined by entry phenotype, was not appreciated until recently. The entry phenotype is closely linked to the target cell and thus to virus–host interactions and pathogenesis. The most abundant form of HIV-1 uses CCR5 as the coreceptor and requires a high density of CD4 for efficient entry, defining its target cell as the CD4 + memory T cell. This is the transmitted form of the virus, the form that is found in the blood, andtheformthatreboundsfromthelatentreservoir.WhenCD4 + /CCR5 + Tcellsbecomelimiting the virus evolves to use alternative target cells to support viral replication. In the CNS, the virus can evolve to use a cell that displays only a low density of CD4,while maintaining the use of CCR5 as the coreceptor. When this evolutionary variant evolves, it must be sustaining its replication in either macrophages or microglial cells, which display only a low density of CD4 relative to that on Tcells.Inthebloodandlymphoidsystem,themajorswitchlateindiseaseisfromTcellsexpressing CD4andCCR5toTcellsexpressingCD4andCXCR4,withachangeincoreceptorspecificity.Thus the virus responds in two different ways to different environments when its preferred target cell becomes limiting. there is active replication during that virus in lymph with therapy discontinuation.


INTRODUCTION
It is a special treat to write a review for the Society of Leukocyte Biology. There is a long history of viruses giving insights into fundamental cell processes. For example, one of us (R.S.) was present many years ago at a Cold Spring Harbor Retroviruses Meeting where an experiment was presented of annealing adenoviral mRNA to viral DNA with the product analyzed by EM; there were big loops in the DNA strand giving birth to the idea of splicing. The story we have to tell today is not as grandiose, but it is of fundamental importance to understanding HIV-1 replication and pathogenesis. Ironically, it is a story that got off track 30 years ago and we have taken up the task of setting the story, and the biology, straight. In this review, we will explain why, until recently, there has been no name for the most abundant form of HIV-1 as defined by its cell entry phenotype. This is because the nature of the protein. The Env protein is synthesized on the rough ER and is a type 1 integral membrane protein. There is an N-terminal signal peptide that is cleaved off leaving a 160-kDa precursor protein as the initial translation product; this protein is heavily glycosylated giving rise to the alternative name of gp160 (i.e., glycoprotein of 160 kDa). The gp160 form of the protein trimerizes and is transported through the ER/Golgi.
In the Golgi, the individual subunits of the trimer are cleaved by a furinlike protease to give rise to two proteins, gp120 and gp41, that stay together noncovalently. gp120 is completely outside of the virus particle and is called the Surface or SU protein. It is responsible for receptor and coreceptor binding. gp41 remains as a transmembrane protein giving rise to its name, TM, with an extracellular domain, a transmembrane domain, and a cytoplasmic domain; the gp41 protein mediates fusion between the viral and host membranes during entry. Like gp160, the gp120/gp41 complex remains as a trimer first on the surface of the host cell and then as part of the budded virus particle. The heavy glycosylation of Env results in one-half of the molecular mass of gp120 being carbohydrate, [2][3][4] with the role of these carbohydrates being to shield the Env protein from becoming the target of host-neutralizing antibodies that would otherwise use the surface of the protein as neutralization epitopes. 5-7

The old view of HIV-1: HIV-1 can be X4 T cell-tropic or R5 M-tropic
All viruses infect target cells based on the presence of a receptor on the cell surface. For HIV-1, the receptor was defined early on based on the observation that an antibody to the cell surface protein CD4 could block HIV-1 infection. 8,9 CD4 is an accessory protein on T helper cells that helps to stabilize the binding of the T cell receptor to peptidepresenting MHC Class II proteins on the surface of antigen presenting cells. This was the initial hint that CD4+ T helper cells are the target of HIV-1 infection, an idea that was subsequently supported by numerous lines of evidence including the fact that these cells are lost during the course of infection (see, e.g., Ref. 10) and the fact that the HIV-1 Env protein binds to 11 and has a high affinity for CD4. 12,13 However, even early on it was clear that not all HIV-1 isolates could enter all CD4-expressing cells, 14 but the initial assumptions in interpreting these early infection experiments generated misconceptions about the nature of these restrictions in HIV-1 entry phenotypes.
Misconceptions in our understanding of HIV-1 entry phenotypes remain prevalent and can largely be traced back to the fact that we virologists are usually virologists first and cell biologists a distant second (with some notable exceptions). As a result, most virology studies do not use the primary cells that viruses infect in vivo. The first shortcut we take is to figure out what continuous cell line the virus will grow in, hopefully similar to the target cell in vivo, but we must all acknowledge that transformed cell lines do not always recapitulate a phenomenon that exists in a whole organism. However, when you mix 100 virus particles with cells in culture and a week later you have 10 million particles and dead cells you usually feel that important parts of the biology of viral replication are in place. Such cells were found in the form of cell lines that had been derived from CD4+ T cell leukemias.
These cell lines in combination with viral isolates capable of growing in these cells were important tools for developing much of our initial understanding of HIV-1. This was enough to allow the development of diagnostic tests, cloning of the viral genome followed by the determination of its entire genomic sequence, the generation of infectious virus from cloned DNA to allow mutagenesis, and the definition of all the viral genes and their gene products. Given the relatively small size of the HIV-1 genome (around 10,000 bases), all of this happened within just a few years (although our knowledge of how the virus interacts with host cell proteins is still incomplete). As these tools and reagents became available, it was possible to develop antiviral drugs, initially to the viral DNA polymerase reverse transcriptase and the viral protease, and more recently to the viral integrase, accomplishments that are dramatically changing the nature of the HIV-1 epidemic. All good.
Like all good mystery stories, there is an additional narrative embedded in this truly amazing history. The early isolates of HIV-1 came from people who were diagnosed with immunodeficiency. One thing that often happens as people progress to immunodeficiency is the virus undergoes a "coreceptor switch" from using CCR5 to using CXCR4 as the coreceptor. [15][16][17] At the time the first isolates were made nothing was known about the coreceptor so the fact that the virus existed as a mixture of entry phenotypes could not be appreciated. Also, unknown at the time of these early experiments, the transformed T cell lines that were being used expressed CD4 and CXCR4, but not CCR5 18 (this is true of most CD4+ transformed T cell lines). All of this adds up to the fact that some of our early "insights" into HIV-1 entry phenotypes were based on CXCR4-using viruses which we now know are an unusual subset of HIV-1 variants that primarily emerge late in disease.
There was great positive reinforcement in using viral isolates from late in infection and transformed T cell lines because they revealed distinct viral phenotypes. One example is in a phenomenon termed syncytium formation. When infected cells express Env on their surface they can behave like a virus particle and can fuse with an adjacent cell that is expressing the CD4 receptor (and appropriate coreceptor) to form multicell syncytia, sometimes swelling with water to look like distended balloons. Because of this behavior in CD4+ T cell lines, some of the earliest HIV-1 isolates were given two names: one descriptive, syncytium-inducing (SI), and one mechanistic, T cell-tropic (for the ability to grow in a T cell line). By default, those viruses that did not form syncytia were non-syncytium inducing (NSI), and since they apparently did not grow in a T cell line, they were not T cell-tropic. However, in hindsight these conclusions and designations are limited by properties of the T cell lines that could not have been appreciated at that time. This early classification of HIV-1 as one of two types persists as a significant misunderstanding to this day, and even some of the erroneous conclusions about viral populations persist from the use of these T cell lines. The first important clue that we were on the wrong track was that all of these viruses grew in PBMCs (think CD4+ T cells) but given the framework of available information the failure to grow in a T cell line was viewed as a limiting feature of the NSI viruses (making them less pathogenic) and not a defective feature of the cell line that prevented it from recapitulating real CD4+ T cells (i.e., the absence of CCR5 expression).
In this world of T cell-tropic/SI versus NSI viruses came the use of monocyte-derived macrophages (MDM) in HIV-1 infection experiments. Since viral pathogenesis is intimately tied to the target cells that support viral replication, efforts were made to identify the targets of NSI HIV-1. Given that these variants replicated poorly in transformed T cell lines, 14,19 myeloid lineage cells were identified as possible targets. In humans, some myeloid cells (monocytes and macrophages) have been shown to express CD4, 20 but at a much lower density than that found on CD4+ T cells. 21 When NSI viruses were cultured on MDM, many were found to replicate suggesting that these isolates were macrophage-tropic 22,23 ; however, they varied widely in how well they replicated. 22 Thus, the two types of HIV-1 were further defined as SI/T cell-tropic and NSI/M-tropic. While this served as a convenient designation, it was an oversimplification that ignored the fact that NSI viruses replicated well in PBMCs (which generally do not contain macrophages) and did not always grow in MDM. However, the concept of two types of HIV-1 was quickly accepted as dogma and any phenotype not associated with SI viruses was attributed to M-tropic HIV-1. This designation that most of HIV-1 was M-tropic helped fit the virus into another (erroneous) paradigm that the entire genus of lentiviruses in the retrovirus family was based on macrophage tropism, a (false) "truism" that many of us parroted for too long.
Given that CD4+ T cell lines express CD4 but do not support infection of all HIV-1 variants, this led some to reason that these cell lines were lacking something most HIV-1 isolates needed. This conclusion was confirmed by studies showing that HIV-1 cannot fuse with nonhuman cells expressing CD4, 24-28 but can fuse with human/animal cell hybrids that express CD4. 25  and CD4, and that antibodies to CXCR4 blocked this process. 29 This identified CXCR4 as one of the elusive HIV-1 coreceptors. 29 CXCR4 is a member of the large superfamily of G-protein-coupled receptors (GPCRs) 30 and is expressed on many transformed CD4+ T cell lines. 18 In this way, CXCR4 became the coreceptor for the SI/T cell-tropic viruses, later named X4 viruses. CCR5, another GPCR, was soon identified as the coreceptor for NSI/M-tropic HIV-1, 31-35 later named R5 viruses. The addition of coreceptors as part of the entry phenotype was shoehorned into the dogmatic view of two types of HIV-1, which were subsequently defined as X4 T cell-tropic and R5 M-tropic. However, in the same way the term T cell-tropic was too simplistically applied to X4 viruses; M-tropic has been too simplistically applied to R5 viruses.

Course correction: HIV-1 can be R5 T cell-tropic, X4 T cell-tropic, or (R5) macrophage-tropic
There were several inconsistencies that had to be ignored in the "two types" simplification of HIV-1 entry phenotypes. First, all viruses grow in PBMCs (which contain CD4+ T cells, but not macrophages), so why are some HIV-1 variants "T cell-tropic" and others not? Second, not all NSI viruses grow well in macrophages (in fact most do not 22 ), so why are they collectively called by a single name? Third, a subset of viruses, often linked to HIV-associated dementia and present in the brain, could use a low density of CD4 much more efficiently than the typical R5 virus. [36][37][38][39][40][41] This low CD4 density is similar to that present on macrophages. 21  When clones, and then inhibitors, of the two coreceptors became available it was possible to assess entry phenotypes with respect to coreceptor usage in a straightforward way. In contrast, assessing macrophage tropism was/is more difficult. Entry into macrophages uses CD4 (at a low density) and typically CCR5, just like entry into real The use of single genome amplification (SGA, or template endpoint dilution PCR) to amplify and clone full-length HIV-1 env genes from patient samples without PCR recombination [43][44][45][46][47] was essential in generating biologically relavant env clones for analyses of entry phenotypes. In contrast, many early studies of HIV-1 tropism used viral isolates that were generated by culturing patient samples with PBMCs and/or cell lines. As a result, these early studies did not examine a random sample of variants from each patient, but rather examined the tropism of viruses that were able to grow in these cells (typically CXCR4-using if grown on T cell lines or with artificially reduced population complexity when passaged in PBMCs). In addition, the culturing process used in these early studies may have allowed viruses to adapt to culture conditions which could have altered their entry phenotype. In contrast, the SGA method generates full-length env clones that represent randomly chosen env genes from the viral population within a person (without passage in culture) that can be used in entry assays to assess tropism. This approach also avoids the artifact of PCR-mediated recombination, which scrambles the different lineages within a population to create genomes that never existed in vivo. There are now hundreds of env clones available that were generated by SGA performed using patient samples collected from different stages of disease and from different anatomical sites.
Another advancement that improved entry phenotype assessment was the introduction of Affinofile cells. 48 48 can be induced to express different CD4 densities, similar to the high level expressed on CD4+ T cells or the low levels expressed on MDM and monocytes. This reagent was used to examine the ability of Env proteins to facilitate entry into cells expressing low CD4 relative to their ability to infect high CD4 cells. 21 This approach revealed that both CXCR4-using T cell-tropic variants (B) and CCR5-using T cell-tropic variants (C) are inefficient at entering cells expressing low levels of CD4 and require high CD4 levels (like those on T cells) for efficient entry. (D) In contrast, M-tropic Env proteins are able to efficiently enter cells expressing low CD4 levels, similar to those found on macrophage, while still being able to infect cells with a high density of CD4 efficiently to adapt to the local CNS cells that express CD4 (i.e., macrophages and/or resident microglial cells), even if they express a low density of CD4. This is the adaptation we measure as the ability of these viral Env proteins to mediate entry into cells with a low density of . Thus it appears that as CD4+ T cells are lost systemically in the human infection the virus starts to be exposed to sufficient selective pressure to begin the move toward becoming M-tropic.

F I G U R E 1 M-tropic HIV-1 is efficient at entering cells that express low CD4 densities. (A) Affinofile cells
Given the difficulty in finding true M-tropic viruses, it is worth noting several more examples of their detection. The isolation of macrophage-tropic virus from the blood has been described, 59,60 although the method of phenotyping the virus did not use Affinofile cells. We have found a single example (to date) of a lineage of virus in the blood that had a CD4-low entry phenotype 50 ; in this one example. we were able to sample virus before and shortly after the person started therapy and found that both the standard CD4-high (T celltropic) and the unexpected CD4-low (M-tropic) viruses were reduced equivalently in blood viral load by therapy, indicating both were replicating in short-lived T cells (Joseph and Swanstrom, unpublished data).
Similarly, we have detected M-tropic virus in the semen of one male. 61 We hypothesize that in each of these cases the M-tropic virus evolved in an isolated tissue depleted of CD4+ T cells and was introduced into the systemic pool of replicating virus in the former or into draining seminal fluid in the latter.
This brings us to the second question: Why not be an M-tropic/CD4low virus all of the time? Why bother maintaining the CD4-high entry phenotype? The question of why a CD4-high entry phenotype exists, and dominates, over the CD4-low phenotype is a difficult one, which means we do not have a clear answer. In an attempt to find phenotypic correlates of macrophage tropism, we noted that there is a trend toward these viruses being more sensitive to CD4-binding site antibodies. 49 It is possible that to become more efficient in using CD4 the virus must assume an alternative conformation making it more susceptible to antibody neutralization. This alternative conformation appears to be distinct from the "open" conformation that one obtains with tissue culture adaptation of HIV-1. Tissue cultureadapted virus is an artifact that has confounded and likely will continue to confound some studies of the Env protein. Tissue culture environments lack the selective pressures present in vivo, thus allowing Env proteins to evolve a conformation that approximates the CD4bound conformation, with both the V3 loop exposed and the CD4- ding from one macrophage has a higher probability of infecting another macrophage than a CD4+ T cell. The two explanations are not mutually exclusive, that is the same immunodeficiency that reduces target CD4+ T cells could also cause a waning of the neutralizing antibody response that selects against the altered conformation of macrophagetropic viruses.

Exploring the preferences of T cell-tropic viruses
HIV-1 infection of T cells has long been thought of as infection of CD4+/CCR5+ cells early in disease and infection of CD4+/CXCR4+ cells late in disease, but it is now appreciated to be more complicated.
After completing development, naive CD4+ T cells exit the thymus and, if stimulated by exposure to their cognate antigen, may become activated, acquire effector functions and undergo proliferation. While the majority of these effector cells will die, a subset will develop a memory phenotype. A distinguishing feature of naive and memory CD4+ T cells is that naive cells express CXCR4, whereas memory T cells express both CXCR4 and CCR5. 20  cally lower than that of memory T cells. 66,67 The fact that naive cells are less likely to be latently infected after a period of ART may be due to them being less susceptible to infection prior to therapy or due to them turning over more rapidly after the initiation of therapy. An additional factor that may reduce the number of latently infected naive cells is that over time they might develop a memory phenotype.
While memory CD4+ T cells are generally accepted as the largest reservoir of replication-competent proviruses during ART, we are only beginning to understand the contribution that different memory subsets make to the reservoir and the processes maintaining these latently

HIV-1 and the latent reservoir
One of the truisms of HIV-1 therapy is that when therapy is stopped the virus comes back, thus necessitating the need for life-long treatment. The main culprit for the source of this rebound virus is generally thought to be resting CD4+ T cells that contain latent copies of integrated viral DNA that can be induced to produce virus. [81][82][83] When we examined the entry phenotype of the rebound virus that initially appears in the plasma of people stopping therapy, we found it was R5 T cell-tropic. 84 This is consistent with a T cell reservoir, and most efforts directed at curing HIV-1 are focused on this cell type. Yet there is significant interest in knowing if there could be a separate latent reservoir in macrophages, either in the periphery or in the CNS. One can imagine a large reservoir of virus in T cells and a smaller reservoir in macrophages. Eradication of virus from T cells would be inadequate for a cure in this case. While we have argued that the major form of HIV-1 should be considered R5 T cell-tropic, even this form of the virus can inefficiently infect macrophages with their low density of CD4. Given an entire body, there must be many macrophages that get infected fortuitously by R5 T cell-tropic virus. Thus the potential for a macrophage reservoir, even in the absence of detecting M-tropic virus, must exist. A first step in addressing this question has been made with a humanized mouse model system with only myeloid-derived target cells present to support HIV-1 replication. In this system, the virus persisted for several weeks while the mice were given antiretroviral therapy followed by rebound when therapy was stopped. 54 This striking observation reminds us that the interaction between HIV-1 and its target cell will continue to challenge us as we further move our understanding of replication and persistence from the cell culture setting to all that is happening within an infected person.

Shouting down echoes of the past
This characterization of three types of HIV-1, with R5 T cell-tropic being the most abundant, is supported by compelling evidence. We can anticipate refinements of any construct with future information, but this framework accounts for much more of the biology of HIV-1 than does the longstanding "two types" categorization. However, the drag of 20+ years of self-reinforced dogma will continue to confound the thinking in our field. As one example (there are many), we can cite our much beloved Principles of Virology (4th ed., 2015) that stops with the description of "X4 and R5 strains," makes a vague reference to "strains with enhanced neurotropism," and describes the late stage virus in the blood as becoming "relatively homogeneous and specific for the CXCR4 receptor. Properties associated with increased virulence predominate, including an expanded cellular host range, ability of the virus to cause formation of syncytia, rapid reproduction kinet-

SUMMARY
There are three types of HIV-1 not two: R5 T cell-tropic (the standard virus), and X4 T cell-tropic and M-tropic (the latter two evolved to either use a new coreceptor or to infect more efficiently cells with a low level of CD4). The analysis of genetic compartmentalization and differing entry phenotypes provides insights into local tissue environments where viral replication can occur. It is important to correctly understand the biology of the virus to be able to interpret its interaction with the host. F I G U R E 2 Viral populations present in the blood late in disease. Plasma samples from subjects with low CD4+ T cell counts ("late-stage subjects") were used as the source of viral RNA for deep sequencing of the C2 to V3 region of the viral env gene. Phylogenetic trees show the diversity of the viral population in each subject. The lineages are color coded based on the false positive rate (FPR) of predicting X4 lineages using the Geno2Pheno algorithm. We found that all lineages with values above 10% were R5 viruses phenotypically (blue and green), lineages with values less than 2% were X4 (red), and lineages with FPR values between 2% and 10% (orange and blue) were R5 75% of the time. For comparison, a group of three "early-stage subjects" are included showing limited diversity associated with the transmission of a single variant and R5-predicted phenotypes. Taken from Zhou et al. 85