Mutational Pathways and Genetic Barriers to CXCR4-Mediated Entry by Human Immunodeficiency Virus Type 1

Abstract

To examine mutational pathways that lead to CXCR4 use of HIV-1, we analyzed the genotypic and phenotypic characteristics of envelope sequences from a large panel of patient virus populations and individual clones containing different V3 mutations. Basic amino acid substitutions at position 11 were strong determinants of CXCR4-mediated entry, but required multiple compensatory mutations to overcome associated reductions in infectivity. In contrast, basic amino acid substitutions at position 25, or substitutions at position 6–8 resulting in the loss of a potential N-linked glycosylation site, contributed to CXCR4-mediated entry, but required additional substitutions acting cooperatively to confer efficient CXCR4 use. Our assumptions, based upon examination of patient viruses, were largely confirmed by characterizing the coreceptor utilization of five distinct panels of isogenic envelope sequences containing V3 amino acid substitutions introduced by site-directed mutagenesis. These results further define the mutational pathways leading to CXCR4 use and their associated genetic barriers.

INTRODUCTION

As the principal coreceptors of human immunodeficiency virus type 1 (HIV-1) infection, chemokine receptors CCR5 and CXCR4 cooperate with the CD4 receptor to mediate virus entry (Berger, Murphy, and Farber, 1999). HIV-1 coreceptor tropism is defined as the ability of a virus to use CCR5 only (R5), CXCR4 only (X4), or both of these coreceptors (dual). R5 virus populations represent the majority of HIV-1 infections, followed by dual/mixed (R5 plus dual and/or X4) virus populations; X4 virus populations are relatively rare. R5 variants predominate during acute and early infection and generally persist throughout the entire course of HIV-1 infection (Berger, Murphy, and Farber, 1999). In contrast, CXCR4-using variants (dual and, to a lesser extent, X4 viruses) are detected infrequently in recent infections (Eshleman et al., 2007; Huang et al., 2009; Markowitz et al., 2005; Masquelier et al., 2007), but gradually increase in prevalence as chronic infection ensues (Berger, Murphy, and Farber, 1999; Connor et al., 1997; Melby et al., 2006; Schuitemaker et al., 1992; Wilkin et al., 2007). Generally, CXCR4-using variants are found in approximately 20% of antiretroviral (ARV) treatment-naïve individuals, characterized by higher CD4+ T-cell counts (Brumme et al., 2005; Moyle et al., 2005), and approximately 50% of highly treatment-experienced patients, characterized by low CD4+ T-cell counts (Coakley et al., 2006; Demarest et al., 2004; Hunt et al., 2006; Melby et al., 2006; Wilkin et al., 2007). In the absence of ARV treatment, the presence of CXCR4-using viruses is strongly associated with faster depletion of CD4+ T cells and accelerated disease progression (Daar et al., 2007; Koot et al., 1993; Richman and Bozzette, 1994; Schuitemaker et al., 1992; Shepherd et al., 2008; Tersmette et al., 1989). Whether these associations are causal or consequential has not been firmly established and is the subject of ongoing debate.

More recently, the development of CCR5 antagonists as new therapeutic agents has drawn further attention to HIV-1 coreceptor tropism. The CCR5 antagonist maraviroc (Pfizer) is approved for the treatment of HIV-1 infection in patients that have failed conventional treatment regimens comprised of nucleoside, non-nucleoside and/or protease inhibitors (Gulick et al., 2008), and more recently in the treatment naïve setting (Cooper et al., 2010). Several additional therapeutic candidates have advanced to late stage clinical evaluations, including the small molecule CCR5 antagonists vicriviroc (Gulick et al., 2007; Su et al., 2009; Suleiman et al., 2010) and INCB9471 (Erickson-Viitanen et al., 15^th Conference on Retroviruses and Opportunistic Infections, Boston, MA, Feb 3–6, 2008), as well as the CCR5 monoclonal antibody Pro-140 (Jacobson et al., 17^th Conference on Retroviruses and Opportunistic Infections, San Francisco, CA, Feb.16–19, 2010). In each case these CCR5 inhibitors effectively suppress the replication of R5 virus populations, but not dual/mixed or X4 virus populations (Lewis et al., 2007; Saag et al., 2009; Tsibris et al., 2009; Westby et al., 2006). Although initial attempts to develop therapeutic agents that block CXCR4-mediated entry have met with less success (Hendrix et al., 2004; Moyle et al., 2009), several new candidates have recently entered early-stage clinical evaluations (Jenkinson et al., 2010; Murakami et al., 2009). Since CXCR4 inhibitors effectively suppress the replication of X4 virus populations, but not R5 or dual/mixed virus populations (Fransen et al., 2008; Hendrix et al., 2004), their use in the clinic may be linked to co-administration with CCR5 inhibitors.

Numerous studies have firmly established the third variable region (V3) of the HIV-1 envelope glycoprotein (Env) as a primary determinant for coreceptor tropism (Cocchi et al., 1996; Deng et al., 1996; Hoffman and Doms, 1999; Hung, Vander Heyden, and Ratner, 1999; Hwang et al., 1991; O'Brien et al., 1990; Shioda, Levy, and Cheng-Mayer, 1991; Speck et al., 1997; Trkola et al., 1996; Trujillo et al., 1996; Westervelt et al., 1992; Wu et al., 1996). The substitution of as few as one or two amino acids in the V3 region of R5 viruses has been reported to confer CXCR4 use (Chesebro et al., 1996; De Jong et al., 1992; Mosier et al., 1999). However, more complex patterns of amino acid substitutions in the V3 region are typically observed among the CXCR4-using variants within patient virus populations that also contain an R5 subpopulation for comparison (Huang et al., 2007; Jensen et al., 2003; Lewis et al., 2007; Tsibris et al., 2009; Westby et al., 2006). Previous studies have defined specific determinants of coreceptor tropism within the V3 region, including the acquisition of positively-charged amino acids, particularly at positions 11 and 25 (De Jong et al., 1992; Fouchier et al., 1992; Resch, Hoffman, and Swanstrom, 2001), and, more recently, position 24 (Cardozo et al., 2007), as well as the loss of a potential N-linked glycosylation site (PNGS) at position 6–8 (Pollakis et al., 2001). A recent study has demonstrated that the ability to accurately predict CXCR4 use is CD4 independent for viruses with a positively-charged amino acid at position 11, but is CD4 dependent for viruses with positively-charged amino acids at position 25, or that lack a PNGS at position 6–8 (Low et al., 2008). Precisely how substitutions at position 11, 25 and 6–8 impact CXCR4 use remains to be defined.

To advance our understanding of how HIV-1 populations broaden or switch host range, i.e. coreceptor tropism, either as a result of host- or drug-imposed selective pressure, we have attempted to more clearly define the associations between specific V3 determinants for CXCR4 use and their impact on Env mediated entry by analyzing patient-derived viruses. The primary goal of the present study is to specifically define the selective advantages and disadvantages of V3 mutational pathways that lead to efficient CXCR4 use. In doing so, we have examined the effect of positively-charged amino acids at positions 11 and 25, as well as the absence of a PNGS at positions 6–8 on CXCR4 use by analyzing (a) the envelope (env) sequences of 35 patient virus populations containing any one of these V3 substitutions in isolation, (b) multiple individual env clones derived from 12 patient virus populations containing mixed amino acid sequences at these positions, and (c) five different env sequences containing specific V3 amino acid substitutions introduced by site-directed mutagenesis. We found that these specific V3 substitutions differentially influence CXCR4 mediated entry, and their effects are highly context dependent and reliant on the presence of additional env amino acid substitutions that serve to compensate for reductions in env infectivity, or act cooperatively to confer efficient CXCR4 use. The degree of complexity and fluidity of mutational pathways leading to CXCR4 use that we observed is consistent with a high genetic barrier and may, in part, explain why CXCR4 use typically emerges late in the course of HIV infection, if at all.

METHODS

Patient virus selection

To obtain viruses containing positively-charged amino acids at position 11, 25 or lacking PNGS at position 6–8 in the V3 region of env, we determined the V3 nucleotide sequences of 200 patient plasma-derived viruses (submitted for routine coreceptor tropism [Trofile] testing), to obtain the 47 viruses described in this study. Based on env sequences, 39 viruses were identified as subtype B and 8 samples as non-subtype B and recombinant (A=1, C=2, AE=1, A/G=2, B/C=1, B/D=1). Thirty five of 47 patient viruses contained unambiguous V3 sequences (no mixtures) and a basic amino acid substitution at either position 11 (N=7) or 25 (N=25), or that lacked a PNGS at position 6–8 (N=3). The remaining 12 patient viruses contained mixed V3 amino acid sequences at position 11 (N=4) or 25 (N=5), or at position 6–8 (N=3). For this subset of 12 viruses, we also determined the V3 nucleotide sequences and coreceptor tropisms for approximately 10 env clones per virus population. Since these 47 patient virus samples were submitted to Monogram for routine coreceptor tropism testing, no clinical information or longitudinal samples were available.

Site-directed mutagenesis

Single amino acid substitutions (S11K, S11R, E25K, E25R) were introduced into the V3 regions of three R5 molecular env clones of HIV-1: JRCSF and BaL (AIDS Research and Reference Reagent Program), and clone c11.2 (from subject 11 in this study) using site-directed mutagenesis (Sarkar and Sommer, 1990). The V3 PNGS was removed from each of these three R5 env sequences by introducing an N6Q mutation. In addition, 11R substitutions were replaced by 11S substitutions in two dual env clones (c2.41, c3.14) derived from subjects 2 and 3. The complete gp160 nucleotide sequence of each engineered env gene was determined to verify the presence of the desired mutations and confirm the absence of other mutations,

Coreceptor tropism determinations

The coreceptor tropisms of patient virus populations, molecular env clones derived from patient virus populations, and env clones containing site-directed mutations in V3 were determined using the Trofile coreceptor tropism assay (Whitcomb et al., 2007). Briefly, HIV-1 env genes were amplified from patient plasma samples by RT-PCR and incorporated into env expression vectors. HIV-1 pseudovirions were generated by co-transfecting HEK-293 cells with patient virus-derived env expression vectors and an HIV-1 genomic vector containing a firefly luciferase reporter gene. Coreceptor tropism was determined by measuring the ability of pseudovirions to infect U87 target cells that express CD4 and either CCR5 or CXCR4. In the Trofile assay, the production of luciferase activity in CXCR4 and/or CCR5 target cells that exceeds background levels (~10² RLU in this study) and is inhibited by a CXCR4 or CCR5 inhibitor, respectively, is considered a demonstration of env-mediated virus entry.

env V3 sequencing

V3 nucleotide sequences for patient virus populations and molecular clones were determined using conventional dideoxy chain terminator chemistry (ABI, Foster City, CA). V3 amino acid sequences were deduced from nucleotide sequences. Predictions of coreceptor tropism based on derived V3 amino acid sequences were determined using two well-established algorithms; 11/25 rule and PSSM (De Jong et al., 1992; Fouchier et al., 1992; Jensen et al., 2003).

env V3 structure modeling

The V3 sequences of paired R5 and dual clones from four subjects (2, 3, 7 and 9), along with template structures 2QAD, 1U6U and 2ESX, were aligned by hand and modeled using MODELLER (Sali and Blundell, 1993). Surface images with electrostatic forces were created using Discovery Studio 2.5 software (http://accelrys.com/products/discovery-studio/).

RESULTS

Patient-derived viruses that contain basic amino acids at V3 position 11 or 25, or that lack an N-linked glycosylation site in V3, vary in their ability to use CXCR4 and CCR5

To better define the contributions of positively-charged amino acid substitutions and the loss of PNGS in the env V3 region to CXCR4-mediated entry, we inspected the V3 sequences of 200 patient virus populations that were submitted for routine coreceptor tropism testing (Trofile). This initial inspection defined an unambiguous subset of 35 viruses that either contained a basic amino acid substitution at position 11 (N=7) or 25 (N=25), or lacked a PNGS at position 6–8 (N=3) (Table 1). The majority of these viruses (31/35) were subtype B (C=1, B/D=1, A/G=2). All seven of the basic substitutions at position 11 were arginine (11R); no lysine (11K) was observed. In contrast, basic substitutions at position 25 were equally represented by arginine (25R, N=15) and lysine (25K, N=11). The coreceptor tropism (R5, DM, X4) and infectivity (luciferase reporter gene activity expressed as relative light units, RLU) of these 35 viruses are summarized in Table 1. All seven viruses containing 11R substitutions utilized CXCR4 efficiently relative to CCR5, i.e. infectivity was comparable or higher in CXCR4+ cells than in CCR5+ cells (median ratio of CXCR4 RLU:CCR5 RLU = 4.673, range = 0.720 to 14.971), thus no 11R viruses were classified R5. In contrast, roughly one third (8/25) of the 25K/R virus populations were R5 viruses. The remaining 25K/R virus populations (17/25) were classified DM or X4 but varied (>4 log₁₀ RLU) in there ability to utilize CXCR4 (median ratio of CXCR4 RLU:CCR5 RLU = 1.034, range = 0.001 to 129.905). Similarly, CXCR4 utilization varied (>4 log₁₀ RLU) of 2/3 viruses that lacked the PNGS within the V3 region (median CXCR4 RLU:CCR5 RLU = 225.169, range 1.083 to 449.225), while the remaining one virus was R5.

Table 1.

Comparison of coreceptor tropism of patient virus populations containing different amino acid substitutions at position 11, 25 or 6–8 assigned by phenotype (Trofile) and genotype (PSSM and 11KR/25KR)

	Position in V3a			Infectivity (RLU)b			Coreceptor Tropismc
Samples (subtype)	11	25	No. of PNGS	CXCR4+ cells	CCR5+ cells	Ratio CXCR4 RLU to CCR5 RLU	Trofile	PSSM	11KR/25KR
1(B)	R	Q	1	65,466	90,981	0.720	DM	X4	X4
2(B)	R	A	1	14,483	11,418	1.268	DM	R5	X4
3(B)	R	G	1	481,460	115,108	4.183	DM	X4	X4
4(B)	R	S	1	23,201	4,494	5.163	DM	X4	X4
5(B)	R	E	1	73,866	9,646	7.658	DM	X4	X4
6(B)	R	E	1	138,510	9,252	14.971	DM	X4	X4
7(A/G)	R	D	1	30,551	69	ND	X4	X4	X4
8(B)	G	R	1	97	713,550	ND	R5	R5	X4
9(B)	S	R	1	80	517,002	ND	R5	R5	X4
10(B)	S	R	1	56	18,037	ND	R5	R5	X4
11(B)	G	R	1	55	97,220	ND	R5	R5	X4
12(B)	G	R	1	51	4,367	ND	R5	R5	X4
13(B)	G	R	1	582	540,398	0.001	DM	R5	X4
14(B)	G	R	1	1,905	52,584	0.036	DM	X4	X4
15(B)	G	R	1	625	167,263	0.004	DM	R5	X4
16(B/D)	S	R	1	627	173,087	0.004	DM	R5	X4
17(B)	G	R	1	1,243	24,309	0.051	DM	R5	X4
18(B)	S	R	1	118,979	104,602	1.137	DM	R5	X4
19(B)	H	R	1	42,390	25,953	1.633	DM	R5	X4
20(B)	S	R	1	428,770	133,275	3.217	DM	R5	X4
21(B)	S	R	1	21,419	57	ND	X4	X4	X4
22(C)	G	K	1	66	5,448	ND	R5	X4	X4
23(B)	S	K	1	79	9,667	ND	R5	R5	X4
24(B)	S	K	1	61	29,357	ND	R5	R5	X4
25(B)	S	K	1	589	138,904	0.004	DM	R5	X4
26(B)	G	K	1	3,000	447,328	0.007	DM	R5	X4
27(B)	G	K	1	221,510	238,062	0.930	DM	X4	X4
28(B)	S	K	1	236,387	182,683	1.294	DM	R5	X4
29(B)	G	K	1	232,884	170,395	1.367	DM	X4	X4
30(B)	G	K	1	54,503	19,738	2.761	DM	X4	X4
31(B)	G	K	1	51,865	2,890	17.946	DM	X4	X4
32(B)	G	K	1	70,019	539	129.905	DM	X4	X4
33(A/G)	S	D	0	75	13,292	ND	R5	X4	R5
34(B)	I	D	0	56,990	52,608	1.083	DM	X4	R5
35(B)	S	Q	0	200,817	447	449.255	DM	X4	R5
HXB2 (X4)	S	E	1	620,340	98	ND	X4	X4	X4
BaL (R5)	R	E	0	90	2,023,112	ND	R5	R5	R5

^aThe 11R and 25K/R substitutions and the absence of the potential N-linked glycosylation site (PNGS) at position 6–8 in V3 are shown in boldface (no mixtures were detected at these positions).

^bInfectivity was measured as luciferase activity (relative light units, RLU) using the Trofile assay. RLU <200 represents background levels of luciferase activity on CXCR5+ or CXCR+ cells. The ratios of CXCR4 RLU to CCR5 RLU are indicated for DM viruses, ND: CXCR4/CCR5 ratios were not determined because luciferase activity (RLU) was at or below background levels for either CXCR4 or CCR5 cells.

^cCoreceptor tropism predictions by genotypic algorithms that do not match the phenotypic determination by the Trofile assay are highlighted in gray.

All but one of the seven viruses containing 11R substitutions were consistently predicted to use CXCR4 based on the V3 genotype-based algorithms PSSM and 11/25 rule. These predictions were highly concordant with phenotypic measurements of CXCR4 use (Table 1). In contrast, genotypic predictions of CXCR4 use or CCR5 use for the subset of viruses containing 25K/R substitutions were often discordant with phenotypic measurements (11 of 25 viruses misclassified by PSSM, 8 of 25 viruses misclassified by 11/25 rule). Similarly, genotypic algorithms also misclassified CXCR4 use among viruses lacking the PNGS within the V3 region (1 of 3 viruses misclassified by PSSM, 2 of 3 viruses misclassified by 11/25 rule) (Table 1).

Among the 35 viruses evaluated, six additional amino acid substitutions--glutamate (E), aspartate (D), glutamine (Q), glycine (G), alanine (A), and serine (S)--were observed at position 25 (10 viruses), in addition to K and R (25 viruses). Four amino acid substitutions--predominantly S (14/28) and G (11/28), and rarely histidine (H, 1/28) and isoleucine (I, 1/28)--were observed at position 11 in 28 viruses, in addition to R in seven viruses (Table 1). These data suggest that HIV-1 can accommodate a greater diversity of amino acid substitutions at position 25 versus position 11 in V3. Notably, in addition to the 200 viruses surveyed in the current study, 11K was also not observed upon surveying 1,148 env sequences of patient viruses compiled by our laboratory (Stawiski et al; unpublished observation).

11R substitutions in V3 strongly promote, but are not essential for efficient use of CXCR4

To further examine the contribution of the 11R substitution to CXCR4 usage, we next characterized the coreceptor tropism and V3 sequences of multiple env molecular clones derived from four patient virus populations that had mixed amino acid sequences at position 11 (env subtype B=2, B/C=1, AE=1), (Table 2). Three virus populations (subjects 1–3) were comprised of mixtures of R5 and dual variants, while the remaining virus population (subject 4) was comprised solely of dual variants. In all four subjects, 11R variants were mixed with either 11S (subjects 1, 2 and 4) or 11G (subject 3) variants (consistent with the position 11 substitutions that we observed for the 35 viruses listed Table 1). Based upon infectivity, every 11R env clone isolated from all four subjects utilized CXCR4 efficiently relative to CCR5. This phenotype of 11R env clones is consistent with the 11R virus populations described in Table 1 and with our previous description of a dual-X virus sub-classification (Huang et al., 2007).

Table 2.

env clonal analysis of four patient virus populations containing mixed amino acid substitutions at position 11 in V3

Subject (subtype)	Virusesa	Position 11b	V3 amino acid sequencesc	No.of AA changesd	Coreceptor tropisme	Infectivity (RLU)f
						CXCR4+ cells	CCR5+ cells
1(B/C)	Population	S/R			DM	14,188	523,551
	clone (n=4)	S		0	R5	83	951,905
	clone (n=1)	S		4	R5	63	556,160
	clone (n=1)	S		5	R5	157	1,015,357
	clone (n=2)	R		8	Dual-X	564,997	432,732
2(B)	Population	G/R			DM	73,347	55,808
	clone (n=1)	G		0	R5	177	69,366
	clone (n=2)	G		0	Dual-R	660	1,456,992
	clone (n=4)	R		5	Dual-X	681,919	350,020
	clone (n=3)	R		6	Dual-X	730,141	174,252
3(AE)	Population	S/R			DM	6,091	142,291
	clone (n=4)	S		0	R5	121	603,308
	clone (n=1)	S		2	R5	74	1,211,595
	clone (n=1)	S		2	Dual-X	16,534	13,848
	clone (n=3)	R		5	Dual-X	1,026,300	693,476
4(B)	Population	S/R			DM	560,500	745,641
	clone (n=6)	S		0	Dual-X	879,283	531,108
	clone (n=1)	R		1	Dual-X	173,666	126,407
	clone (n=1)	R		2	Dual-X	29,439	40,645

^aClones with identical V3 sequences and the same coreceptor tropism are grouped together. One dual clone from subject 2 and one from subject 3 (indicated in bold) were chosen for reverse site directed mutagenesis.

^bThe 11R substitutions in V3 are shown in boldface

^cFor V3 amino acid sequences, dots represent amino acids identical to the first sequence, dashes in the first sequence indicate gaps to accommodate insertions. Positions 11, 25 and 6–8 (for PNGS) are underlined.

^dThe number of amino acids that differ from the most prevalent V3 sequence of R5 clones in the same sample is indicated, except for subject 4. Subject 4 contained only dual clones, and the number of amino acid differences between clones with or without 11R is shown.

^eCoreceptor tropism was determined by the Trofile assay. Dual clones were further classified as dual-X and dual-R based on efficiently or inefficiently utilized CXCR4, as defined previously (Huang et al., 2007). R5 and dual-X clones from subject 3 that shared identical V3 amino acid sequences are highlighted (bold and italics). Dual-X clones from subject 4 that shared identical V3 amino acid sequences, except at position 11 (S or R), are highlighted (bold and italics).

^fInfectivity was measured as luciferase activity (relative light units, RLU) in the Trofile assay. RLU <200 on CXCR5+ or CXCR+ cells was considered background luciferase activity in this study. Infectivity of multiple clones with the same V3 sequence is expressed as the average RLU.

It is important to note that we also identified dual-X variants that lacked 11R substitutions in two of the four patient viruses evaluated (Table 2). In subject 3, we identified one dual-X variant that shared an identical V3 sequence (including 11S, 25E) with R5 clones. In this case, CXCR4 use is determined entirely by sequence changes outside of the V3 region. In subject 4, all eight clones that were evaluated displayed dual-X tropism and shared similar V3 sequences, yet only two clones contained the 11R substitution, whereas the remaining six clones had 11S. We conclude from these findings that the 11R-containing variants frequently contribute to efficient CXCR4 use, but this substitution is not a requisite determinant.

CXCR4 utilization by env sequences containing 25K or 25R substitutions in V3 is context dependent

Evaluation of the contribution of basic amino acid substitutions at position 25 to CXCR4 use was similarly conducted by characterizing the tropism and V3 sequences of multiple env molecular clones derived from five patient virus populations (env subtype B=4, C=1) that exhibited mixed amino acid sequences at position 25 (Table 3). In subjects 5–9, virus subpopulations containing either 25K or 25R co-existed with subpopulations containing 25E, 25D, or 25G. In subjects 5 and 6, all of the clones evaluated shared similar V3 sequences and did not utilize CXCR4, i.e. were R5, irrespective of the presence or absence of a basic amino acid substitution (K or R) at position 25. Conversely, in subjects 7, 8 and 9, all clones containing 25K or 25R were able to use CXCR4, i.e. were dual or X4 (10⁵–10⁶ RLU). Notably, the V3 sequences of these variants also contained numerous additional substitutions relative to the V3 sequence of the most predominant R5 variants.

Table 3.

env clonal analysis of five patient virus populations containing mixed amino acid substitutions at position 25 in V3

Subject (subtype)	Viruses a	Position 25 b	V3 amino acid sequences c	No.of AA changes d	Coreceptor tropism e	Infectivity (RLU)f
						CXCR4+ cells	CCR5+ cells
5(B)	Population	E/K			R5	62	383,806
	clone (n=4)	E		0	R5	70	1,796,486
	clone (n=1)	E		1	R5	92	1,559,819
	clone (n=1)	D		1	R5	79	1,799,031
	clone (n=2)	K		1	R5	51	552,861
6(B)	Population	G/R			R5	66	148,806
	clone (n=3)	G		0	R5	54	359,517
	clone (n=8)	R		2	R5	81	149,570
7(C)	Population	E/K			DM	709,584	661,721
	clone (n=3)	E		0	R5	182	298,585
	clone (n=1)	E		3	R5	172	128,256
	clone (n=4)	K		3	Dual-X	630,016	660,567
8(B)	Population	D/K			DM	2,701	27,905
	clone (n=5)	D		0	R5	79	559,320
	clone (n=3)	K		10	X4	497,805	133
9(B)	Population	E/R			DM	62,552	392,667
	clone (n=4)	E		0	R5	85	503,528
	clone (n=1)	E		1	R5	63	1,029,078
	clone (n=1)	R		11	Dual-X	1,487,002	1,754,137
	clone (n=1)	R		13	X4	356,894	122
	clone (n=1)	R		15	X4	137,117	86

^aClones with identical V3 sequences and the same coreceptor tropism are grouped together.

^bThe 25K/R substitutions in V3 are shown in boldface.

^cFor V3 amino acid sequences, dots represent amino acids identical to the first sequence, dashes indicate gaps to accommodate insertions. Positions 11, 25 and 6–8 (for PNGS) are underlined.

^dThe number of amino acids that differ from the most prevalent V3 sequence of R5 clones in the same sample is indicated.

^eCoreceptor tropism was determined by the Trofile assay. Dual clones were further classified as dual-X and dual-R based on efficiently or inefficiently utilized CXCR4, as defined previously (Huang et al., 2007).

^fInfectivity was measured as luciferase activity (relative light units, RLU) in the Trofile assay. RLU <200 on CXCR5+ or CXCR+ cells was considered background luciferase activity in this study. Infectivity of multiple clones with the same V3 sequence is expressed as the average RLU.

CXCR4 utilization by env sequences lacking a PNGS at position 6–8 in V3 is context dependent

We isolated and characterized the tropism and V3 sequences of multiple env molecular clones derived from three patient virus populations (env subtype B=2, A=1) that exhibited mixed PNGS at V3 position 6–8 (Table 4). In subjects 10 and 11, none of the clones evaluated utilized CXCR4, i.e. were R5, irrespective of the presence or absence of a PNGS at position 6–8, and in the case of subject 11, the presence of 25K as well. Conversely, in subject 12, dual-X variants lacking the PNGS utilized CXCR4 efficiently (>10⁵ RLU) when they also contained numerous additional substitutions relative to the V3 sequences of the most predominant R5 variants. Based on the clonal analyses presented in Tables 3 and 4, we conclude that the contribution of basic amino acid substitutions at position 25, or the absence of PNGS at positions 6–8, to CXCR4 use is highly dependent on the context of the env sequence, in particular, the V3 region.

Table 4.

env clonal analysis of three patient virus populations containing mixed amino acid substitutions that disrupt a PNGS at position 6–8 in V3

Subject (subtype)	Viruses a	No. of PNGS at position 6–8 b	V3 amino acid sequences c	No.of AA changes d	Coreceptor tropism e	Infectivity (RLU) f
						CXCR4+ cells	CCR5+ cells
10(B)	Population	0/1			R5	54	18,687
	clone (n=8)	0		0	R5	74	335,296
	clone (n=1)	0		1	R5	100	201,829
	clone (n=1)	1		4	R5	107	520,939
11(B)	Population	0/1			R5	163	59,137
	clone (n=5)	1		0	R5	66	580,001
	clone (n=1)	1		1	R5	101	715,276
	clone (n=2)	0		4	R5	72	34,875
	clone (n=1)	0		7	R5	74	1,563
	clone (n=1)	0		6	R5	55	6,777
12(A)	Population	0/1			DM	20,924	216,146
	clone (n=5)	1		0	R5	101	950,986
	clone (n=2)	1		0	Dual-R	898	1,148,595
	clone (n=1)	0		1	R5	118	703,962
	clone (n=2)	0		12	Dual-X	569,619	1,785

^aClones with identical V3 sequences and the same coreceptor tropism are grouped together. One R5 clone from subject 11 (indicated in bold) was chosen for site directed mutagenesis.

^bThe absence of the potential N-linked glycosylation site (PNGS) at position 6–8 in V3 are shown in boldface

^cFor V3 amino acid sequences, dots represent amino acids identical to the first sequence. Positions 11, 25 and 6–8 (for PNGS) are underlined.

^dThe number of amino acids that differ from the most prevalent V3 sequence of R5 clones in the same sample is indicated.

^eCoreceptor tropism was determined by the Trofile assay. Dual clones were further classified as dual-X and dual-R based on efficiently or inefficiently utilized CXCR4, as previously described (Huang et al., 2007).

^fInfectivity was measured as luciferase activity (relative light units, RLU) in the Trofile assay. RLU <200 on CXCR5+ or CXCR+ cells was considered background luciferase activity in this study. Infectivity of multiple clones with the same V3 sequence is expressed as the average RLU.

An accumulation of multiple amino acid changes in V3 is required for efficient utilization of CXCR4

It should be noted that env variants that are capable of utilizing CXCR4, i.e. dual-X and X4, were characterized by multiple amino acid substitutions in V3 relative to R5 variants within the same virus population. The number of amino acid differences among env clones in each virus population is summarized in Figure 1. Furthermore, the association between the presence of basic amino acid substitutions at position 11 and 25, or the loss of a PNGS at position 6–8, and CXCR4 use appears to be reliant on the presence of additional mutations in V3 (Figure 1, Table 2–4). For example, although all 11R variants were dual-X, they also contained 5–8 additional amino acid substitutions in V3 relative to R5 clones from the same virus population (Table 2, subjects 1–3). Similarly, CXCR4-using variants containing 25K or 25R substitutions, or lacking the PNGS at position 6–8, also contained 3–15 different amino acids in V3 compared to R5 clones within the same virus population (Table 3, subjects 7, 8, 9 and Table 4, subject 12). Overall, the number of different V3 amino acid substitutions in env sequences containing 11R, 25K/R, or lacking a PNGS at position 6–8, was noticeably higher for X4 and dual-X variants (mean=9, range 3–15) compared to R5 and dual-R variants (inefficiently utilized CXCR4) within the same virus population (mean=1, range 0–5). We interpret these data as evidence that the evolution of CXCR4 use either by the acquisition of basic amino acid substitutions at position 11 or 25, or the loss of a PNGS at position 6–8, also involves the accumulation of additional V3 amino acid changes that may contribute to either efficient CXCR4 use or improvements in env infectivity.

Figure 1. The number of V3 amino acid differences among env clones from each of 12 subjects.

Each symbol represents a unique V3 sequences that differs from the most prevalent V3 sequence of R5 clones in the same sample (except for subject 4). Subject 4 contained only dual-X clones (efficiently utilize CXCR4) and the number of amino acid differences between clones with or without 11R is shown. R5 and dual-R (inefficiently utilize CXCR4) clones are indicated by open squares, dual-X and X4 clones are indicated by filled squares. The amino acid substitution or number of PNGS is shown within each symbol.

Introduction of 11K/R, 25K/R or N6Q mutations into the V3 region of R5 env sequences clarifies mutational pathways and genetic barriers to CXCR4 use

We used site-directed mutagenesis to engineer a panel of isogenic viruses that could be used to clearly establish the impact of basic amino acid substitutions at positions 11 and 25 of V3, or the loss of a PNGS site at position 6–8, on coreceptor utilization and env infectivity (Table 5). Single V3 mutations (11K, 11R, 25K, 25R, N6Q) were introduced into the env sequences of three R5 viruses: the well-characterized JRCSF and BaL strains, and molecular clone c11.2 isolated from the virus population of subject 11. Clone c11.2 (S11, E25, N6) was chosen as the third R5 env backbone sequence for these studies because all variants analyzed from the parental patient virus population were R5, irrespective the presence of a basic amino acid substation at position 25 (K) and/or the absence of a PNGS at position 6–8 (Table 4).

Table 5.

Coreceptor utilization of site-directed mutants using three R5 and two dual env backbones

			Infectivity (RLU)c
Source of env backbonesa	V3 mutationsb	Coreceptor tropism	CXCR4+ cells	CCR5+ cells	% RLU of parental (CCR5+ cells)d
JRCSF (R5)	S11, E25, PNGS(+)	R5	193	1,321,140	100
	S11K	Dual	1,929	299,175	23
	S11R	Dual	35,195	655,246	49
	E25K	Dual	22,303	829,696	63
	E25R	Dual	40,693	816,309	62
	PNGS(−)	Dual	47,206	3,234,479	245
BaL (R5)	S11, E25, PNGS(+)	R5	90	2,420,047	100
	S11K	R5	191	123,724	5
	S11R	Dual	11,597	757,409	31
	E25K	Dual	14,063	2,795,704	116
	E25R	Dual	37,674	2,079,533	86
	PNGS(−)	Dual	1,899	1,542,274	64
c11.2 (R5)	S11, E25, PNGS(+)	R5	60	308,852	100
(subject 11)	S11K	R5	87	6,157	2
	S11R	R5	78	67,408	22
	E25K	R5	77	152,441	49
	E25R	R5	70	117,422	38
	PNGS(−)	R5	78	62,047	20
c2.41 (Dual)	R11, Q25, PNGS(+)	Dual	632,384	400,342	100
(subject 2)	R11S	Dual	586,581	600,991	150
c3.14 (Dual)	R11, E25, PNGS(+)	Dual	790,017	342,895	100
(subject 3)	R11S	Dual	525,764	317,581	93

^aFive distinct env sequences were used to generate site-directed mutants: parental JRCSF, BaL and c11.2 (isolated from subject 11) env sequences are R5; parental c2.41 and c3.14 (isolated from subjects 2 and 3) env sequences are dual.

^bThe listed substitutions were introduced into the V3 region of three R5 env sequences (JRCSF, BaL and c11.2) and two dual env sequences (c2.41, c3.14). The presence or absence of PNGS at position 6–8 is indicated as + or −. PGNS(−) mutants of R5 env sequences were engineered by the introduction of N6Q substitutions.

^cInfectivity was measured as luciferase activity (relative light units, RLU) in the Trofile assay. RLU <200 on CXCR5+ or CXCR+ cells was considered background luciferase activity in this study.

^dCCR5-mediated entry of each mutant env sequence is represented as a percentage of the parental env sequence.

The introduction of the 11K substitution into all three R5 env sequences conferred detectable levels of CXCR4-mediated entry to JRCSF env (>10³ RLU), but not for BaL and c11.2 envs (Table 5). Notably, the introduction of 11K was also associated with reductions in env-mediated infectivity on CCR5+ cells (23%, 5% and 2% relative to the parental JRCSF, BaL and c11.2 parental envs, respectively). Similarly, the introduction of the 11R substitution conferred CXCR4-mediated entry to the JRCSF and BaL envs (>10⁴ RLU) but not to the clone c11.2 env. Again, the introduction of 11R was accompanied by reductions in infectivity on CCR5+ cells (49%, 31% and 22% relative to the JRCSF, BaL and c11.2 parental clones, respectively). Taken together, these observations suggests that the 11K and 11R substitutions are strong determinants of CXCR4 use, but when introduced as single mutations, their contributions are partially obscured by impairments in env infectivity in the absence of compensatory mutations.

The introduction of the 25K or 25R substitutions conferred CXCR4-mediated entry to JRCSF and BaL (>10⁴ RLU) but not clone c11.2. In contrast to the 11K and 11R substitutions, the 25K and the 25R substitutions did not severely impair infectivity on CCR5 cells, suggesting that these two substitutions do not dramatically reduce overall env infectivity (Table 5). Thus, the ability of the 25K and 25R substitutions to confer CXCR4 use to some, but not all, R5 env sequences was more apparent than the 11K and 11R substitutions.

Disruption of the PNGS in the V3 regions of the three R5 env sequences was accomplished by the introduction of an N6Q mutation. Loss of the PNGS conferred detectable levels of CXCR4 use to JRCSF (>10⁴ RLU) and BaL (>10³ RLU). However, infection of CXCR4+ cells was not observed with the c11.2 env containing the N6Q mutation (Table 5). The effects of the loss of the PNGS on CCR5-mediated entry varied widely, ranging from 20% to 245% for the c11.2 and JR-CSF mutants, respectively. The reduction in CCR5 cell infectivity of the c11.2 mutant implies that compensatory mutations may be required to obtain efficient CXCR4 use by this R5 env sequence.

Reversion of R to S at position 11 does not reduce the ability of patient viruses to use CXCR4

To further test the role of V3 substitutions on CXCR4 use by patient virus isolates, we used site-directed mutagenesis to revert R substituions to S substitution at position 11. The infectivity of two dual clones (c2.41 and c3.14) containing 11R from subjects 2 and 3, and their respective mutants with a single R11S change are shown in Table 5. Interestingly, the two mutants with the R11S substitution retained the dual-tropic phenotype, and their ability to use CXCR4 and CCR5 for entry was not reduced relative to the parental clones. These data, together with experiments described above, confirm that backbone env sequences collectively contribute to CXCR4 use and can strongly influence the effects of specific V3 determinants of CXCR4 use, such as 11R, 25K/R and loss of a PNGS.

Comparative modeling of V3 structure in paired R5 and dual clones

To further examine the influence of V3 amino acid substitutions on CXCR4 use, we modeled the structure of the V3 loops of paired R5 and dual clones from two subjects with 11R variants (G11R for subject 2 and S11R for subject 3) and two subjects with 25K/R variants (E25K for subject 7 and E25R for subject 9) (Figure 2). Compared to R5 clones, each paired dual clone exhibited a large gain in positive electrostatic charge. This is in agreement with previous reports describing higher net charges in the V3 regions of CXCR4-using viruses (De Jong et al., 1992; Fouchier et al., 1992). However, our modeling indicated that a single basic amino acid substitution (R or K) at position 11 or 25 alone was not sufficient to produce the net change in positive charge observed in the paired R5 and dual clones (data not shown). This result suggests that the other V3 amino acids substitutions observed in dual clones contribute to the observed differences in electrostatic charge and are consistent with our genotypic and phenotypic characterizations described above. Furthermore, the amino acid side chains of S or G at position 11 are considerably smaller than the corresponding R and K side chains. Consequently, the S and G side chains are free to face, and perhaps form stabilizing hydrogen bonds with the opposing V3 stem, as observed in our models. In contrast, steric constraints imposed an outward positioning of the R and K side chains relative to the opposing V3 stem. At position 25, the side chains of E, R or K are more similar in size, and as such, all are sterically restricted away from the opposing V3 stem. The disparate positioning of amino acid side chains at position 11 and the consistent positioning of amino acid side chains at position 25 in R5 and dual clones may explain the impaired env infectivity that is observed for site-directed mutants containing 11R/K mutations, but not 25R/K mutations.

Figure 2. Structural modeling of the V3 regions of paired R5 and dual env clones from four patient virus populations.

Patterns of electrostatic charge across the V3 region are depicted (positive charges in blue, negative charges in red). Gains in positive charge at positions 11 and 25 are highlighted by * and ^, respectively. The V3 amino acid sequences are provided below each modeled structure, and basic amino acids are highlighted in blue. Subject 2 (G11 vs. 11R), subject 3 (S11 vs. 11R), subject 7 (E25 vs. 25K) and subject 9 (E25 vs. 25R).

DISCUSSION

In this study, we have investigated the effect of different V3 mutational pathways on CXCR4 mediated entry, including the acquisition of positively-charged amino acids or the loss of PNGS. Specifically, we characterized coreceptor usage among patient viruses that contained K or R substitutions at position 11 or 25, or that lacked a PNGS at amino acid position 6–8. We found that all patient viruses with a basic amino acid substitution at position 11 contained R substitutions and were able to utilize CXCR4. Viruses with 11K substitutions were not observed in the group of 200 patient viruses characterized in this study, or a larger survey of >1000 patient virus sequences. Conversely, CXCR4-mediated entry was variable among patient viruses containing K or R substitutions at position 25, or lacking a PNGS at position 6–8, including a subset of viruses that exhibited an R5 phenotype. Analysis of individual env clones containing these V3 changes confirmed our observations within virus populations. All env clones containing 11R substitutions utilized CXCR4 for entry, whereas only a subset of clones that contained either a 25K or 25R substitution, or that lacked a PNGS site at position 6–8 exhibited CXCR4-mediated entry. Current V3 sequence based tropism prediction methods (PSSM and 11/25 rule) are limited in their ability to accurately predict coreceptor use of viruses that lack K or R at position 11.

Our initial observations indicate that the presence of a basic amino acid at position 11 of V3 might be a stronger determinant of CXCR4 use than a basic amino acid at position 25, or the absence of a PNGS at position 6–8. To substantiate this hypothesis, we used site-directed mutagenesis to either introduce a basic amino acid substitutions at positions 11 or 25 or remove the PNGS at position 6–8 in three distinct R5 env sequences. We found that addition of either the 11K or 11R substitution alone was sufficient to confer CXCR4 use for 2/3 R5 env sequences, but that the magnitude of this affect was partially obscured by impairments in env infectivity, as noted by reductions in CCR5 mediated entry (most notably for the 11K substitution). The addition of 25K or 25R substitutions, or disruption of the PNGS was also capable of conferring CXCR4 use, however they were not consistently accompanied by reductions in CCR5-mediated infectivity. These results illustrate that the influence of individual V3 substitutions on CXCR4 use is dependent on env sequence context.

Across three distinct R5 env backbones, larger reductions in env infectivity were consistently observed with the 11K versus the 11R substitution. We speculate that 11K substitutions may be more incompatible with CCR5 utilization than 11R substitutions and consequently subject to more stringent negative selective pressure in vivo. Consistent with the observations in this study, a survey of 506 V3 sequences obtained from the Los Alamos National Laboratory database indicates that serine (S, 90.3%) and glycine (G, 7.5%) are the most prevalent substitutions at position 11 for R5 viruses. Examination of codons reveals a preferential selection for R over K at position 11: a transition from either S or G to R requires one base change while transition to K requires two base changes. This genetic disadvantage coupled with reductions in infectivity likely account for an extremely low frequency of 11K substitutions found in patient isolates.

Overall increases in V3 sequence heterogeneity among CXCR4-using virus populations have also been reported (Low et al., 2008; Troyer et al., 2005). In our study, clonal analysis of the four patient virus populations containing 11R substitutions revealed that, in three cases (subjects 1, 2, 3), all clones bearing 11R substitutions also contained additional V3 amino acid substitutions (N=5–8) relative to R5 clones that co-existed within the same virus population lacking 11R substitutions. In the remaining case (subject 4), both 11R and 11S clones utilized CXCR4 and shared similar V3 sequences, suggesting that 11R, per se, was not (or was no longer) the primary genetic determinant of CXCR4-mediated entry. Furthermore, reversion of R to S at position 11 by site-directed mutagenesis in dual env clones from subjects 2 and 3 did not alter CXCR4 utilization relative to the parental env clones. Based on the combined results of our database survey, site-directed mutagenesis experiments, and clonal analyses of 11R viruses, we invoke the following model that describes potential mutational pathways that lead to efficient CXCR4 use by 11R/K viruses: either 11R/K substitutions occur early, but require compensatory substitutions in V3 that restore obligate losses in env infectivity, or 11R/K substitutions occur late and must be preceded by V3 substitutions that provide more accommodating env contexts for 11R/K substitutions. This model is consistent with a strong association between 11R/K substitutions and efficient CXCR4 use.

Similar to 11R viruses, CXCR4-using env clones that contained 25K or 25R substitutions, or lacked a PNGS at position 6–8, also contained additional V3 amino acid substitutions (N=3–15), relative to R5 env clones that co-existed within the same virus populations (subjects 7, 8, 9 and 12). However, unlike 11R/K substitutions, 25R/K substitutions and mutations that result in loss of the PNGS at position 6–8 are not consistently associated with efficient CXCR4 utilization or reductions in env infectivity. Once again, based on the combined results of our database survey, site-directed mutagenesis experiments, and clonal analyses of 25K/R or PNGS(−) viruses, we invoke the following model to describe potential mutational pathways that lead to efficient CXCR4 use by 25R/K viruses or viruses that lack a PNGS at position 6–8: 25R/K substitutions or loss of the PNGS may occur early or late and are not associated with severe reductions in env infectivity, but require additional V3 substitutions that act cooperatively to confer efficient CXCR4 utilization. This model is consistent with a less stringent association between the presence of 25R/K substitutions or the absence of a PNGS at position 6–8 and efficient CXCR4 use.

The results of our V3 structural modeling exercises are consistent with our experimental observations. The acquisition of efficient CXCR4 use is associated with notable increases in positive electrostatic charges along the V3 stem-loop structure relative to the V3 stem-loop structures of paired R5 viruses from the same virus population. The models indicate that such increases in electrostatic charge are not acquired through single basic amino acid substitutions at position 11 or 25, but rather require multiple amino acid substitutions in the V3 sequence.

It is important to note that previous studies have demonstrated that not all determinants of CXCR4 use reside within the V3 region. Pastore et al., reported that substitutions in the V1/V2 regions of gp120 restore losses in entry infectivity resulting from V3 mutations and play an important role during coreceptor switching from CCR5 to CXCR4 of viruses selected in vitro (Pastore et al., 2006). Previously, our laboratory has reported that substitutions in gp41 of patient-derived envs impacted CXCR4 utilization (Huang et al., 2008). Furthermore, the replacement of V3 regions of X4 or dual env sequences with V3 regions of R5 env sequence, and vice versa, provides compelling evidence for the existence of determinants of CXCR4 utilization outside of the V3 region (Huang et al., 2008). Consistent with these prior observations, in this study we describe two env sequences from subject 3 that share identical V3 sequences, but differ notably in their ability to utilize CXCR4 (R5 vs. dual-X). In addition, clones that inefficiently use CXCR4 (dual-R) and contain the same V3 sequence with R5 clones were also identified.

The detailed characterization of 11R/K viruses that utilize CXCR4 in this study provides a plausible explanation for the observation that basic amino acid substitutions at position 11 but not 25, or losses in PNGS at positions 6–8, are associated with more accurate predictions of CXCR4 use. We postulate that 11R/K viruses likely escape from CCR5 inhibitor treatment by their ability to utilize CXCR4 once compensatory mutations accumulate to restore reductions in infectivity. The phenotypic impact of V3 amino acid substitutions on CXCR4 utilization by patient viruses presented in this study provides further insight into the different mutational pathways and genetic barriers for acquisition of CXCR4 use by clinical isolates.