Angiogenic and HIV-Inhibitory Functions of KSHV-Encoded Chemokines

See allHide authors and affiliations

Science  10 Oct 1997:
Vol. 278, Issue 5336, pp. 290-294
DOI: 10.1126/science.278.5336.290


Unique among known human herpesviruses, Kaposi's sarcoma–associated herpesvirus (KSHV or HHV-8) encodes chemokine-like proteins (vMIP-I and vMIP-II). vMIP-II was shown to block infection of human immunodeficiency virus–type 1 (HIV-1) on a CD4-positive cell line expressing CCR3 and to a lesser extent on one expressing CCR5, whereas both vMIP-I and vMIP-II partially inhibited HIV infection of peripheral blood mononuclear cells. Like eotaxin, vMIP-II activated and chemoattracted human eosinophils by way of CCR3. vMIP-I and vMIP-II, but not cellular MIP-1α or RANTES, were highly angiogenic in the chorioallantoic assay, suggesting a possible pathogenic role in Kaposi's sarcoma.

Kaposi's sarcoma (KS) is a highly angiogenic multicentric tumor most commonly seen in immunodeficient individuals. Since the acquired immunodeficiency syndrome (AIDS) epidemic, KS has become one of the most common tumors in parts of Africa and is the most common tumor found in HIV-infected individuals (1). Compared to classic KS found in patients from Mediterranean or East European descent, KS in AIDS patients is a more fulminant disease: The angiogenic properties of the HIV-1 Tat protein have been proposed to enhance KS tumor formation (2).

KSHV DNA is present in all KS biopsies, and antibodies to this virus are detectable mainly in those with KS or at risk of developing KS (3). These data, and previous epidemiological data indicating that an infectious agent is involved in KS pathogenesis (4), suggest that KSHV is likely to play a central role in KS tumorigenesis. In KS biopsies, KSHV is present in most of the tumor cells (so-called spindle cells) and in endothelial cells (5). Although it appears that KSHV gene expression in most of these cells is restricted to latent genes (5), a proportion of endothelial and spindle cells in KS lesions harbor hundreds of viral particles, suggesting that lytic viral infection may be necessary to drive KS lesion formation (6) and the vMIPs are expressed in KS lesions (7). KSHV is also etiologically linked to multicentric Castleman's disease (MCD), a polyclonal lymphoproliferation associated with prominent vascularity (8).

Sequencing of the KSHV genome has demonstrated colinearity and similarity with the gamma-oncogenic viruses Epstein-Barr virus (EBV) and herpesvirus saimiri (HVS) (9). However, unique among known human herpesviruses, KSHV encodes two genes whose products show sequence similarity to the human CC chemokine family, with highest similarity to macrophage inflammatory protein–1α (MIP-1α) and RANTES: vMIP-I [open reading frame (ORF) K6] exhibits 37.9% amino acid sequence identity and vMIP-II (ORF K4) 41.1% amino acid identity to MIP-1α (10). The amino acid identity between vMIP-I and vMIP-II is 48%, and they are more closely related to each other phylogenetically than to cellular chemokines, suggesting that they have evolved by gene duplication within the virus genome rather than by independent acquisition from the host genome. A third KSHV ORF (K4.1) is also related to the CC chemokine family, but this gene is more distantly related and probably derived independently from another member of the CC chemokine family (9, 11). Both vMIP-I and vMIP-2 are expressed in latently infected lymphoma cells, and their expression is induced by phorbol esters (10).

Chemoattractant cytokines (chemokines) and their receptors play a fundamental role in leukocyte migration and activation and in hematopoiesis (12). The finding that chemokine receptors act as cofactors for HIV-1 entry into CD4+ cells and that the CC chemokines (MIP-1α, MIP-1β, RANTES, and eotaxin) can suppress some strains of HIV replication in peripheral blood mononuclear cells (PBMCs) and chemokine receptor–transfected cell lines has intensified interest in these proteins (13). vMIP-I can inhibit infection of some primary non–syncytium-inducing (NSI) HIV strains when cotransfected with CCR5, the recently identified coreceptor for NSI HIV-1 strains (10). We generated synthetic proteins (14) of vMIP-I and vMIP-II and assessed their capacity to inhibit HIV infection, induce calcium mobilization, and induce angiogenesis.

We investigated whether vMIP-I and vMIP-II can suppress HIV-1 replication in U87/CD4 cells (a human glioma cell line) expressing HIV-1 coreceptors and in primary PBMCs (15). To define which chemokine receptors were most important for inhibition of HIV-1 infection by the vMIPs, we tested U87/CD4 cells stably expressing either CXCR4, CCR3, or CCR5 with the NSI strains SL-2 and SF162 and the dual-tropic syncytium-inducing (SI) strains 89.6 and 2028. SL-2 and SF162 are primary NSI, macrophage-tropic strains that use the RANTES, MIP-1α, and MIP-1β receptor CCR5 to gain entry into CD4+ cells. Strains 89.6 and 2028 are SI dual-tropic and can use CXCR4 and CCR3 in addition to CCR5 for entry. CD4-independent infection of an HIV-2 strain, ROD/B, on U87 cells expressing CXCR4 but not CD4 was also examined (16). Whereas modest levels of inhibition were observed with vMIP-II on infection of dual-tropic HIV-1 strains by way of CCR5 and CXCR4 as well as on HIV-2 ROD/B by way of CXCR4 (17), vMIP-II potently inhibited both 89.6 and 2028 infection, with geater than 95% inhibition at 200 nM on cells stably expressing CCR3 (Fig. 1A). No significant blocking of HIV infection by way of CCR3, CCR5, or CXCR4 on U87/CD4 cells by vMIP-I was observed (Fig. 1A) (17). These results suggest that vMIP-II binds predominantly to CCR3, rather than CXCR4 or CCR5. Alternatively, HIV-1 infection by way of CCR3 may be especially sensitive to chemokine inhibition.

Figure 1

Inhibition of HIV replication by vMIP-I and vMIP-II. (A) U87/CD4 cells stably expressing CCR3 were treated with the chemokines vMIP-I, vMIP-II, RANTES, and SDF at the concentrations indicated. Treated cells were challenged with the SI dualtropic HIV-1 strains 89.6 and 2028. Cells were fixed and immunostained for in situ p24 5 days later. p24-positive foci were counted, and average titers are shown. (B) PBMCs were challenged with 89.6 and NSI SL-2 HIV-1 strains in the absence or presence of chemokines vMIP-I, vMIP-II, RANTES, and MCP-1 at the concentrations indicated. Cell supernatants were harvested 9 days later and p24 levels measured. Average p24 concentrations are shown. Bars represent the standard error from triplicate wells. The results are representative of two and three separate experiments for strains 89.6 and SL-2, respectively.

We next tested whether vMIP-I or vMIP-II inhibited HIV-1 infection of PBMCs. Strains 89.6 and SL-2 were used in these studies. Both vMIP-I and vMIP-II partially inhibited SL-2 infection of PBMCs (Fig. 1B). Whereas RANTES reduced SL-2 infection of PBMCs by more than 98% at 12.5 nM, vMIP-I and vMIP-II were much less efficient, blocking infection by ∼80% and 40%, respectively, at 100 nM. No substantial inhibition of 89.6 infection was observed. vMIP-I inhibited the CCR5-using SL-2 strain on PBMCs but not on CCR5+ U87/CD4 cells. CCR5, however, is likely to be expressed at substantially lower levels on PBMCs than on cell lines such as U87/CD4. This difference may explain such differential cell type–dependent sensitivity to chemokine inhibition.

Because the HIV inhibition studies indicated that vMIP-II binds predominantly to CCR3, the eotaxin receptor, we investigated the potential of the vMIPs to act as agonists on CCR3. Because CCR3 is expressed predominantly on eosinophils (18), we used an in vitro human eosinophil activation assay that measures intracellular Ca2+ mobilization (19). vMIP-II, like eotaxin, was a potent activator of human eosinophils and was more active than MIP-1α or RANTES (Fig. 2A). In comparison, vMIP-II had no effect on Ca2+ flux in human neutrophils, whereas the CXC chemokines interleukin-8 (IL-8) and GROα were potent agonists (Fig. 2A). Despite the sequence similarities between vMIP-I and vMIP-II, vMIP-I was inactive on both eosinophils and neutrophils. We therefore examined whether vMIPs function as receptor antagonists on eosinophils and neutrophils, as has been demonstrated for synthetic NH2-terminal variants of CC (for example, Met-RANTES) and CXC (for example, IL-81 69) chemokines (20). vMIP-I had no inhibitory effect on the response of eosinophils to eotaxin, and neither vMIP-I nor vMIP-II prevented Ca2+ mobilization in neutrophils induced by IL-8 (data not shown).

Figure 2

Calcium-mobilization assay on eosinophils and neutrophils. (A) Peak increase of [Ca2+]i concentration in fura-2–loaded human eosinophils and neutrophils in response to vMIP-I and vMIP-II; the CC chemokines eotaxin, RANTES, and MIP-1α; and the CXC chemokines IL-8 and GRO. Data are expressed as the mean ± SEM of three to four separate experiments with cells purified from different individuals. (B) Receptor desensitization of Ca2+ mobilization in human eosinophils between vMIP-II and the CC chemokines eotaxin, RANTES, and MIP-1α. In these experiments the desensitizing agent was added 150 s before the response to the second agonist was measured. Data are expressed as the percentage of the maximal [Ca2+]iinduced by vMIP-II (3 nM) or eotaxin (3 nM) alone (filled columns) after treatment of the cells with various concentrations of the desensitizing agent (vMIP-II, 0.3 to 30 nM; eotaxin, 0.3 to 10 nM; RANTES, 3 to 30 nM; and MIP-1α, 3 to 30 nM). Values are the mean ± SEM of three separate experiments with eosinophils purified from different individuals. (Insets) Representative traces of the data from a single experiment showing the dose-response relation between the magnitude of the response to the desensitizing agent and the subsequent response to the second agonist. The desensitization was specific rather than global because none of the desensitizing agents reduced the response of the cells to a subsequent addition of the complement fragment C5a.

Whereas CCR3 is the predominant chemokine receptor through which eotaxin, RANTES, and other CC chemokines activate eosinophils (18), RANTES and MIP-1α can use CCR1 (21), which is also expressed on eosinophils. To determine the receptor usage by vMIP-II on eosinophils, we performed desensitization studies. vMIP-II exhibited complete cross-desensitization with eotaxin, partial desensitization with RANTES, and no desensitization with MIP-1α (Fig.2B), providing functional evidence that vMIP-II binds to CCR3 on eosinophils.

To show that vMIP-II can chemoattract eosinophils, we performed an in vitro chemotaxis assay (22). vMIP-II, but not vMIP-I, was chemotactic for human eosinophils (Fig.3). This activity of vMIP-II was comparable to that of eotaxin, which correlates with the activity to induce Ca2+ mobilization (Fig. 2A).

Figure 3

Human eosinophil transwell chemotaxis assay. vMIP-II and eotaxin are chemotactic for human eosinophils in vitro. vMIP-I shows no activity in this assay. Buffer alone is represented by a dashed line.

Although angiogenesis is central in the pathology of KS and prominent in MCD, the mechanisms by which KSHV induces new blood vessel formation are not known. Angiogenesis is a complex process involving chemotactic migration and proliferation of endothelial cells, followed by lumen formation and functional maturation of the endothelium (23). Often it is associated with the presence of activated inflammatory cells. Apart from their potent biological effects on leukocytes, some members of the chemokine family have been shown to have a direct effect on blood vessel formation. Within the CXC chemokine family, proteins containing the NH2-terminal Glu-Leu-Arg (ELR) amino acid sequence motif such as IL-8 are angiogenic factors, whereas non–ELR-containing proteins including IP-10 are angiostatic (24). To our knowledge, there is no information on the angiogenic ability of CC chemokines. We therefore examined the potential of vMIP-I and vMIP-II as mediators of angiogenesis in the chick chorioallantoic membrane (CAM), which is a standard in ovo assay to investigate angiogenic function (25, 26).

Angiogenic responses were observed and photographed 3 days after implantation of a methylcellulose disk containing test samples onto CAM. Most eggs implanted with vMIP-I or vMIP-II showed a clear angiogenic response. The positive controls used, vascular endothelial growth factor (VEGF) and phorbol 12-myristate 13-acetate (TPA), also induced angiogenesis as expected (Fig. 4and Table 1). In contrast, cellular MIP-1α and RANTES failed to induce angiogenesis in most eggs, and the few positive responses recorded were not as prominent as those for vMIPs. The potency of vMIPs to induce angiogenesis in chick embryos was demonstrated by clear angiogenic responses in some eggs at doses as low as 0.05 μg, in contrast to poor induction by human CC chemokines at 0.25 μg. Although it is difficult to interpret a comparison of angiogenic activity between chemically synthesized and recombinant proteins, the activity of the vMIPs was not as potent as that of VEGF, but more potent in the CAM assay than that of the ELR-CXC chemokine IL-8 (Table 1). Preliminary results also indicate that eotaxin is not an inducer of angiogenesis in CAM. The finding that KSHV-encoded chemokines can induce new blood vessel formation in the chick CAM is intriguing and could have important implications for the role of KSHV in KS and MCD development.

Figure 4

Induction of angiogenesis by vMIPs in chick CAM. Methylcellulose discs containing vMIP-I (A), vMIP-II (B), RANTES (C), MIP-1α (D), VEGF (E), and BSA (F) were implanted onto 10- or 11-day-old CAMs. Photographs were taken 3 days later. Representative samples are shown.

Table 1

Angiogenic effect of vMIPs on the chick CAM.

View this table:

Our results demonstrate that the vMIPs are biologically active: (i) vMIP-II binds predominantly to CCR3, resulting in potent inhibition of HIV entry by way of this receptor and activation and chemotaxis of eosinophils. (ii) vMIP-II also exhibits weak HIV inhibition through other chemokine receptors. (iii) Both vMIPs induce angiogenesis in chick embryos. These activities may have important implications in viral pathogenesis. vMIP-I and vMIP-II inhibit HIV replication in PBMCs, albeit modestly. In patients with high KSHV viral load, especially as seen in the lymph nodes of HIV-positive individuals (10), these viral chemokines may affect HIV pathogenesis. vMIP-II inhibition of HIV infection by way of CCR3 is potent. This result is consistent with the report in which only vMIP-II was studied with a single HIV-1 strain (27). CCR3 is one of the main receptors for HIV-1 entry into microglia, and an antibody to CCR3 can inhibit HIV-1 infection of microglia (28). Patients with KS or high KSHV viral load, or both, may be less likely to develop HIV-related dementias. The potent agonistic activity of vMIP-II for eosinophils contrasts with its antagonistic function shown in other cell systems (27). Although previous studies have shown that HIV Tat is angiogenic (2), this protein does not play a role in the pathogenesis of non-AIDS KS (for example, classic, African endemic, and posttransplant KS). In contrast, KSHV is present at high levels in all epidemiologic types of KS, and the demonstration of angiogenesis induced by these viral-encoded proteins is the first in vivo indication that KSHV-encoded proteins have the potential to directly induce angiogenesis. Although it is unlikely that these viral chemokines are solely responsible for the marked vascularity seen in KSHV-associated tumors, they could contribute with other angiogenic factors involved in KS spindle cell growth or MCD development—for example, VEGF, basic fibroblast growth factor and IL-6 (2,29).


View Abstract

Navigate This Article