PHENOTYPE DISTRIBUTION (PAPER I TO V)
In paper I, co-receptor use was determined for all the 143 isolates derived from the 24 children (table 5). The isolates were derived from samples taken at different time points after birth. Of these isolates, 86 were derived from PBMC and 57 from plasma.
The distribution of phenotypes showed a high abundance of CCR5-using viruses (n=121) while the remaining isolates (n=22), used CXCR4 alone (n=14) or used multiple co-receptors. Eight isolates were dual-tropic and used both CCR5 and CXCR4. Some viruses in paper I (n=6) also used orphan receptors BOB or BONZO.
In paper II, 11 children from paper I (table 5) were coupled with their infected mothers to evaluate the distribution of phenotypes between mother and child. Isolates derived from the maternal samples (n=55), taken during and after their pregnancies (ranging from the second trimester up to six months post-delivery), also displayed a dominance of CCR5 using viruses (n=46). The remaining 9 isolates used CXCR4 and were found in only two mothers carrying subtype D and CRF01_AE, respectively.
In paper III, the two mothers carrying CXCR4 using viruses from paper II were studied (table 5) together with their children, who also carried this virus phenotype. We wanted to perform a detailed analysis of their viral populations to be able to trace the origin of X4 in the children.
In paper IV, the one mother from papers II and II carrying CXCR4 using virus was matched to four new mothers who had given birth to uninfected children, all whom carried subtype D virus (table 5). This included 30 primary isolates of which 29 grew out, enabling co-receptor determination. Among these isolates, the majority were CXCR4 using viruses (n=20) while the remaining isolates used either CCR5 only (n=8) or in conjunction with CXCR4 (n=1). As observed previously (De Wolf et al., 1994;
Tscherning et al., 1998; Björndal et al., 1999; Tscherning-Casper et al., 2000), this higher degree of X4 viruses was expected in this cohort.
Paper V involved the 176 gathered sequences which were randomly selected from a larger date set of 1015 sequences (table 5). During our preliminary selection for sequences to analyze, we omitted all those with a co-receptor other than CCR5 and/or CXCR4. However, multiple sequences from our first selection carried identical patient unique numbers, so we performed a second round of selection where we randomly selected one sequence from each pool of patient unique sequences, independently of the co-receptor use. We ended up with a genetically diverse set of sequences in which 133 used CCR5, 29 used CXCR4, and the remaining 14 were dual-tropic. This provided a rather large set of data representative of the circulating HIV-1 clades, both in regards to the genetic and biological diversity.
Table 5. Overview of phenotypes identified in all papers
Paper Subjects Phenotypes: R5 X4 R5X4 Total
PBMC Plasma PBMC Plasma PBMC Plasma
I 24 children 73 49 8 6 5 2 143
II 11 mother-child pairs 53 45 11 6 1 0 116
III 2 mother-child pairs 1 4 10 5 1 0 21
IV 5 mothers 4 4 12 8 1 0 21
V 176 individuals 133† 29† 14† 176
† Information on isolate source (PBMC or Plasma) was not available.
SUBTYPES & PHENOTYPES (PAPER I TO V)
As reported previously (Tscherning et al., 1998; Björndal et al., 1999; Tscherning-Casper et al., 2000), the co-receptor use of the subtypes analyzed in papers I to V showed a possible tendency of higher CXCR4 use among subtype D clades. I papers I and II, we observed a clustering of X4 isolates in subtypes A, D and CRF01_AE, that despite not being statistically significant in this cohort, provided us with an incentive to pursue further studies on subtype D isolates (papers III and IV). In paper IV, we
analyzed samples from four mothers of uninfected children infected exclusively with subtype D, and where the co-receptor use of the available isolates had not yet been determined in the previous papers. As we had hoped, we found a high abundance of X4 virus among the isolates (64%), in these mothers. This was even higher when including the previously phenotypes isolates from the mother with the infected child (69%).
This trend was further strengthened when analyzing our collected data in paper V, where despite a random selection of sequences, unbiased towards co-receptor use, provided a high degree of X4 among subtype D sequences (9 out of the 11 subtype D sequences were characterized as X4 = 82%). In contrast, subtype C, which has been suggested to use CXC4 less readily (Tscherning et al., 1998; Björndal et al., 1999;
Tscherning-Casper et al., 2000), was also reflected in our cohorts. In our Swedish children and mothers (paper I and II), no subtype C or B classified isolate used
CXCR4. In the larger perspective in paper V, only one subtype C sequence, of a total of 28 (3.5%), used CXCR4. Similarly, only one out of 22 (4.5%) subtype A sequences was classified as an X4.
Taken together, our results spanning a large Swedish cohort of most group-M subtypes and our sequence cohort from the Los Alamos HIV database would support the notion of subtype specific differences of co-receptor use. Despite the lack of statistically significant confirmations and conflicting studies, many other studies discussed in this thesis and our own results point to an actual difference between subtypes that may very well have implications for transmission and disease development.
CONCORDANCE OF PHENOTYPES BETWEEN MOTHER AND CHILD (PAPER II AND III)
Our motive for our analysis in paper II was to investigate how the distribution of phenotypes was in the mothers to the HIV-1 infected children, where samples from the mothers were available. We wanted to relate the co-receptor use of the isolates (n= 55) from the mothers with what was observed in the children. The 11 mothers and their children were infected with all the subtypes identified in paper I. In that study, 4 out of 24 children (17%; P=0.006 according to the binomial distribution) harbored CXCR4 using virus. Similarly, in the mothers of paper II, 2 out of 11 (18%), carried CXCR4 using virus. Furthermore, there was an accumulation of X4 among these two mothers where only one isolate was R5 (P=0.01 according to the binomial distribution).
The mothers to the two children whom carried X4 viruses also had X4. One
additional child developed the X4 phenotype at around 6 years of age. Only one sample time point was available for this child’s mother, and was determined as R5. The one mother, infected with CRF01_AE, had X4 isolates exclusively during the time of monitoring while the mother infected with subtype D hade X4 in all isolates except the one derived from plasma at delivery. R5 virus was first observed in both children but later developed to X4 in both children. In the child infected with subtype D, all consecutive isolates, except the one taken at 54 months, were phenotyped as X4. For the other child, infected with CRF01_AE, a dual-tropic variant was identified at 48 months of age from a PBMC isolate, and was at 79 months phenotyped as X4 in both PBMC and plasma.
The remaining mothers and their children carried only R5 isolates during the whole period of monitoring. Taken together, these results displayed a similar pattern in HIV-1 co-receptor use between mother and child. When comparing mother-child pairs where the children had been followed for a similar length of time, as those who developed the X4 phenotype, this demonstrated a statistically significant link (P=0.048) between the development of the X4 phenotype in children with the presence of this phenotype in mothers, whether resulting from transmission of X4 or of an independent evolution from R5.
DEVELOPMENT OF THE X4 PHENOTYPE IN VERTICALLY INFECTED CHILDREN (PAPER II AND III)
To identify the nature of the link between the X4 viruses in the two mothers and their children (paper II), we created molecular PCR clones which we sequenced together with the isolates to accurately reconstruct the population dynamics over time using phylogeny. As our available isolates and PBMC/plasma RNA samples ranged from the second trimester of pregnancy and up to multiple years of age in the children, we were successful in creating detailed phylogenetic trees of both mother and child populations. The isolates in our phylogenetic trees were used as phenotype identity markers in which our clone sequences were the bulk of the populations, and the trees
themselves were rooted with maternal sequences from the first available time point to provide time linearity, where applicable.
In both mother-child pairs, it was clear that R5 virus was initially transmitted. This was of course already demonstrated in paper I. However, both R5 and X4 populations were identified in the mothers, but their proportions varied. For the subtype D-infected mother, the X4-associated clone sequences dominated, suggesting it to be the dominant population in blood. In the second mother, infected with CRF01_AE, most clones had no clear phenotype association, and instead, the R5 and X4 populations were observed as separate minor groups of isolates and clone sequences. However, when examining the V3 amino acid sequences of the clone sequences lacking any phenotype association, they carried sequence motifs that were more similar to the sequence from the R5 isolate the associated clones. This suggested, under the assumption that these clone sequences were R5-associated, that R5 was the dominating virus in this mother.
When subsequently expanding the phylogenetic analysis with the inclusion of sequences from the children, one time point at a time, a clear picture emerged. This showed that the maternal R5-associated sequences were closest associated with the R5 sequences, first identified in both children. This was expected as R5 was the first isolate identified in the children. However, the clear phylogenetic relationship seen for the transmission chain in both mother-child pairs was surprising. This relationship, clear in both pairs, was best presented for the subtype D infected mother-child pair. This was because the maternal R5 isolate, taken during delivery, grouped together with the child’s R5-associated clone sequences (fig 12).
In addition to the expected chain of transmission of R5, addition of sequences from later time points in the children showed that X4 developed from the R5 population in each child, and had no observed relation to the maternal X4 sequences (fig 12).
This provided evidence of a transmission-independent and unique evolution of X4 from the established R5 populations in the children. However, in the subtype D-infected child, there was an indication of a second R5-associated virus population, which had a close association with ancestral phylogenetic nodes when looking at all time points together in one phylogenetic tree. This may suggest that possible two R5 viruses were transmitted. This is particularly striking as the transmission of one R5 strain from a dominance of X4 viruses is in contradiction to random events.
The independent evolution of X4 from R5 was further strengthened when
comparing V3 loop amino-acid patterns between sequences from the mothers and their children. The children had unique differences in their X4 virus sequences when
compared to the maternal X4 sequences. The differences between maternal R5 sequences and the R5 sequences from the children showed the opposite (i.e. a striking similarity).
Despite the fact that only two mother-child pairs were analyzed in this paper (III), the high resolution given by the phylogenetic trees, of an abundance of sequences, demonstrating the chain of transmission and the population dynamics in both mother and child. Thus it was clearly shown how X4 evolved, independently of transmission, from an established R5 infection. A selective transmission of R5 was suggested in the subtype D infected mother-child pair as the vast majority of the maternal clone sequences had an X4-association. We could also demonstrate how in this mother, two
R5 populations were transmitted and co-evolved over time, and that in both mother-child pairs, R5 and X4 coexisted during the time of observation.
Fig. 12. A) Phylogenetic tree of the subtype D infected mother and the clones and isolate sequences from the child’s first time point, showing the transmission of R5. Arrow identifies the maternal R5 plasma isolate taken at delivery. B) The second time point in the child together with the maternal sequences showing the branching of the child’s X4 sequences from the maternal R5 population (paper III). Prefix letter denotes subject (M = mother; C = child) and number denotes time point in order.
CO-RECEPTOR USE AND DISEASE PROGRESSION (PAPER I)
In paper I, the clinical and immunological stages were recorded during the time of blood sampling. This provided us at each time point with both, plasma RNA levels, and CD4+ cell counts, as well as the clinical stager of the children, according to the
guidelines set by the Centers for Disease Control and Prevention (CDC) for children of ages 13 and under. This classification is represented with letter codes; A (mild
symptoms), B (moderate symptoms), and C (severe symptoms). The immunological stages, based on CD4+ cell counts, provides a numerical class; 1 (no immune
suppression), 2 (moderate immune suppression), and 3 (severe immune suppression) (CDC, 1994).
When comparing the clinical and immunological stages of HIV-1 infection for the children with the identified phenotype of the virus isolated at that time point, it was clear that in all 4 children who developed X4 virus, it occurred after immune
suppression had advanced. Therefore, the X4 virus was probably not the direct cause of the immune deficiency syndrome (fig 13). Perhaps instead, it would appear as a
possible consequence of the immune deficiency. This could be an alternative
interpretation to the suggested model of X4 development presented by Shankarappa et al., 1999, suggesting X4 to induce immunodeficiency.
However, the suggested increased abundance of X4 prior to AIDS could instead be represented with a stage of fluctuating levels of X4 virus under the shadow of
dominating R5 strains until a change of available targets cells occurs in conjunction with the immune deficiency, which would allow the expansion of X4 (Koning et al., 2003).
Fig. 13. Clinical and immunological stages shown in conjunction with the switch from R5 (○) to X4 (●) (paper I).
PERTINENT V3 AMINO ACIDS IN CO-RECEPTOR USE (PAPER III TO V) During our comparative amino acid analysis in paper III, the most notable difference we observed between R5 and X4 sequences was the V3 loop N-linked glycosylation motif NNT. In the two mother-child pairs, the NNT motif was
completely conserved in all R5 isolates and R5-associated clone sequences. For the X4 sequences, this motif was either conserved or it had changed. This motif was also the main difference between the X4 sequences of the subtype D infected mother and her child, which gave us further support for our conclusion of an X4 expansion from R5. In this mother, the X4 isolates, and clone sequences associated to them, had the NNT motif intact. In addition, a glutamate (Q) substitution present in almost all R5 associated sequences in the mother was present in the majority of the child’s R5-associated clone sequences. This specific difference was not present in any of the X4 sequences from both mother and child, and provided a possible identity link, together with the conserved NNT motif, between the mothers and the child’s R5 sequences.
In the second mother-child pair infected with CRF01_AE, the same conserved NNT motif among all R5-associated sequences was seen. In this case, both the maternal and the child’s X4 sequences had lost their NNT motif. However, the specific
mutations within the NNT motif differed significantly between mother and child, supporting a separate evolution of X4 from R5 in this child, as observed in the other case. Our results allowed us hypothesize that the NNT motif is a requirement for CCR5
use, and that this motif is of less importance for viruses using CXCR4 or both.
These observations led us to expand our analysis by including four mothers
infected with subtype D but who had given birth to uninfected children (paper IV). Our expectation was that this subtype would provide a high probability of obtaining X4 isolates, which in fact was the case. In analyzing the sequences, we observed the same fidelity of the NNT motif among R5 sequences with the exception of a discrepancy in two R5 isolates. It is important to note that when determining the co-receptor use of these isolates, both a syncythia induction in the astroglioma cell lines and a positive PCR result of these cells was our criteria for identification. In these two specific isolates, the PCR run on the lysed cells were positive for CXCR4 expressing cells despite no visual syncythia formation. Therefore they did not fulfill the criteria for being X4 viruses. Perhaps these criteria in some cases were too strict, and one can thus not completely exclude the possibility of these isolates being dual-tropic.
There was another observation regarding the presence of the NNT motif in these mothers, which varied considerably, not only with regards to the R5- and X4-associated populations. In the mother with the infected child, who had rather good clinical
parameters, all isolates and clones possessed the NNT motif, while the first mother with as many X4 isolates and similar clinical parameters, only 11.6% of her clones had the NNT motif. From the point of view that the NNT motif may be strongly associated with transmissibility of HIV-1, this offers a new perspective on the virological characteristics in mother-to-child transmission.
RELEVANCE OF OUR RESULTS FOR GLOBALLY CIRCULATING HIV-1 VARIANTS (PAPER V)
Our observation of the conserved nature of the NNT motif among R5 sequences in our comparatively small cohort prompted us to evaluate this in globally circulating clades. To do this, we analyzed all reported HIV-1 V3 sequences with known genetic class and co-receptor use. After an unbiased and random selection, we obtained 176 patient unique sequences in which we assessed the N-linked glycosylation motifs (sequons) for the whole V3 gene segment present in each case. Five possible sequons were identified through the HXB2R reference sequence (acc. no. K03455) and the amino acid composition of these sites were examined and assessed for a high (≥70%) probability of oligosaccharide addition.
Our results showed both highly conserved and variable sites for likely N-linked glycosylation, homogenous between R5, X4 and dual-tropic viruses. The exception was the fifth N-linked glycosylation motif within the V3 loop, the same as identified in our previous studies (paper III and IV). Among R5 sequences, a functional N-linked glycosylation motif was highly conserved (128/133 = 96%), which was highly
statistically significant when comparing with X4 sequences (P<10 -12). The NNT motif in specific was highly conserved in its amino acid composition (120/121 = 99.2%) among the M-group sequences. The significance of the N-linked glycosylation motif in R5 sequences was less significant when comparing with the dual-tropic (P<0.058). This may not be so surprising as this group may very well reflect a possible mixed
population of R5 and X4 viruses, or the V3 loop N-linked glycan may have a more ambiguous role for co-receptor binding. A high sensitivity (≥0.98) and a high predictive value for the association between the R5 phenotype and the V3 loop N-linked
glycosylation motif strengthened the validity of these observations further.
We also performed a V3 loop net charge calculation for all sequences to compare its association with each phenotype with the N-linked glycosylation motif in the V3 loop. This was done based on a breaking point of +4.1 (at neutral pH) to distinguish R5 from X4 (Fouchier et al., 1992). Our calculations were in agreement with previously reported observations of amino acid substitutions with positively charged amino acids in X4 sequences, rendering higher net charges (Hwang et al., 1991; De Jong et al., 1992; Fouchier et al., 1992; Shioda et al., 1992; de Wolf et al., 1994; Shimizu et al., 1999; Verrier et al., 1999; Hu et al., 2000; Briggs et al., 2000). Our results gave an average lower net charge of for R5 sequences (median = 3.0; P=8.8-12) as compared to X4 (median = 5.9), and also when compared to dual-tropic sequences (median = 5.5;
P=0.002).
From these results, it is possible to conclude that among R5 viruses, the N-linked glycosylation motif within the V3 loop of gp120, and its oligosaccharide, may very well play an important part in the interaction with the co-receptor. The true nature of this interaction is likely an interaction between multiple parts of gp120 and perhaps also gp41. Interestingly, the CCR5 and CXCR4 chemokine receptors are very similar in structure (Pontow & Ratner, 2001). However, CXCR4 has two N-linked glycans that are lacking on CCR5. Considering that this is reversed on the virus side, a plausible model for the interaction between gp120 and the co-receptor is a physical fitness, governed by the oligosaccharides (Hartley et al., 2005).
These results are not directly new in the field of co-receptor research. However, the literature presents the association between the loss of the N-linked glycosylation with the transition from R5 to X4 (Pollakis et al., 2001; Polzer et al., 2001 & 2002; Tebit et al., 2002), in conjunction with a higher net positive charge of the amino acids of the V3 region. A similar study to ours, also focusing on the loss of N-inked glycans and the increase in net charge, further demonstrated this tendency (Dong et al., 2005). Another similar study published earlier, observed a high stability of the V3 region in the context of N-linked glycans and amino acid net charge compared to more variable V3-V4 regions among CRF01_AE sequences (Liang et al., 2003).
In our study, we provide a different perspective on the importance of N-linked glycans for HIV-1 co-receptor use. In specific, we present the association between the V3 loop N-linked glycan and use of CCR5 as co-receptor, as opposed to the loss of it in conjunction with a switch to X4. Our results together with those from fellow
researchers may have implications for the strategies used for the development of vaccines or therapeutic agents. It might be prudent to consider strategies to combat HIV-1 infection through specifically targeting R5 and X4 viruses. Currently,
co-receptor antagonists are used to target chemokine co-receptors, and can interfere with virus interactions, but this strategy is rather crude. However, the strategy of targeting
phenotypes of HIV-1 rather than genotypes may prove to be a sound strategy for future vaccines and antiretroviral drugs. This may be especially true for preventive measures if R5 viruses are the main transmitting variants.