Identification of Human Macrophage Inflammatory Proteins 1α and 1β as a Native Secreted Heterodimer*

Chemokines are secreted proteins that function as chemoattractants for leukocytes. The chemokines macrophage inflammatory protein 1α and 1β (MIP-1α and MIP-1β) now have been shown to be secreted from activated human monocytes and peripheral blood lymphocytes (PBLs) as a heterodimer. Immunoprecipitation and immunoblot analysis revealed that antibodies to either MIP-1α or MIP-1β precipitated a protein complex containing both MIP-1α and MIP-1β under normal conditions from culture supernatants and lysates of these cells. Mass spectrometry of the complexes, precipitated from the culture supernatants of monocytes and PBLs, revealed the presence of NH2-terminal truncated MIP-1α (residues 5–70) together with either intact MIP-1β or NH2-terminal truncated MIP-1β (residues 3–69), respectively. The secreted MIP-1α/β heterodimers were dissociated into their component monomers under acidic conditions. Exposure of monocytes or PBLs to monensin induced the accumulation of heterodimers composed of NH2-terminal truncated MIP-1α and full-length MIP-1β in the Golgi complex. The mixing of recombinant chemokines in vitro demonstrated that heterodimerization of MIP-1α and MIP-1β is specific and that it occurs at physiological conditions, pH 7.4, and in the range of nanomolar concentrations. The data presented here provide the first biochemical evidence for the existence of chemokine heterodimers under natural conditions. Formation of heterodimers of MIP-1α/β may have an impact on intracellular signaling events that contribute to CCR5 and possibly to other chemokine receptor functions.

To date, at least 50 chemokines have been identified. Despite its large size, the chemokine family is remarkably homogeneous. Chemokines are divided into two subfamilies, ␣(CXC) and ␤(CC), on the basis of conserved cysteine residues. Four conserved cysteines form two essential disulfide bonds, Cys1-Cys3 and Cys2-Cys4, in all chemokines. The three-dimensional structures of three ␣ chemokines (interleukin (IL)-8, growthrelated oncogene-␣, and platelet factor 4) and four ␤ chemokines (macrophage inflammatory protein (MIP)-1␣, MIP-1␤, RANTES (regulated on activation normal T cell expressed), and macrophage chemoattractant protein-1 (MCP-1)) have been determined either by multidimensional nuclear magnetic resonance (NMR) or by x-ray crystallography (6 -11). These studies have revealed that chemokines possess a short NH 2terminal domain preceding the first cysteine, a backbone that comprises three antiparallel ␤ strands and a COOH-terminal ␣-helix. Whereas the backbone exhibits a well ordered structure, the structure of the NH 2 terminus is disordered. The similarity in the three-dimensional structures of the chemokine monomers is consistent with the marked sequence homology of these proteins.
The quaternary structures of ␣ and ␤ chemokines, however, differ markedly from each other, and the dimer interfaces are formed by distinct sets of residues. Whereas the IL-8 dimer is globular, the homodimers formed by MIP-1␣, MIP-1␤, and RANTES are cylindrical (6,7,11,12). Calculation of the solvation-free energies of dimerization and analysis of hydrophobic clusters of amino acids suggest that the formation and stabilization of the two different types of dimers result from the burial of hydrophobic residues and that the distinct quaternary structures of ␣ and ␤ chemokine dimers are preserved throughout the two subfamilies. The biological existence and significance of dimeric forms of chemokines, especially that of chemokine heterodimers, have remained unclear.

EXPERIMENTAL PROCEDURES
Reagents-Recombinant human MIP-1␣ (full-length), RANTES, MDC, MCP-1, and IP-10 were obtained from Peprotech (Rocky Hill, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  NJ); recombinant human MIP-1␤ was from Sigma; and recombinant human (Ϫ4)MIP-1␣ and antigen affinity purified antibodies to chemokines were from R&D Systems (Minneapolis, MN).
Immunoprecipitation and Immunoblot Analysis-Cell culture supernatants were incubated at 4°C for 2 h with antibodies to anti-MIP-1␣ or anti-MIP-1␤, and then for an additional 2 h, they were incubated with protein G-Sepharose beads (Amersham Pharmacia Biotech). The beads were washed three times with phosphate-buffered saline containing 0.3% Triton X-100, after which proteins were eluted with 25 mM Tris-glycine, pH 3.0, subjected to SDS-polyacrylamide gel electrophoresis under reducing conditions on a 4 -20% gradient gel in the presence of Tricine, and transferred to a polyvinylidene difluoride membrane. The membrane was incubated for 2 h at room temperature with Trisbuffered saline containing 3% bovine serum albumin and then exposed to either anti-MIP-1␣ or anti-MIP1␤. Immune complexes were detected with alkaline phosphatase-conjugated goat antibodies to mouse immunoglobulin G, alkaline phosphatase substrate, and Lumi-Phos TM 530 (Roche Molecular Biochemicals).
Mass Spectrometry-Culture supernatants were subjected to immunoprecipitation with anti-MIP-1␣ or anti-MIP-1␤. The molecular size of precipitated proteins was determined by matrix-assisted laser desorption ionization and time-of-flight (MALDI-TOF) mass spectrometry (PerSeptive Biosystems, Boston, MA). ␣-Cyano-4-hydroxycinnamic acid (Sigma) and recombinant MIP-1␣ or MIP-1␤ were used as matrix and internal standards, respectively. For some experiments, cells were cultured in the presence of monensin (GolgiStop, Pharmingen, San Diego, CA) for 10 h. They were then washed and lysed for 1 h on ice at a density of 1 ϫ 10 8 cells/ml in a solution containing 1% Triton X-100, 50 mM Tris-HCl, pH 7.4, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 g/ml E64, and 40 g/ml bestatin (Roche Molecular Biochemicals). After centrifugation of lysates at 6700 ϫ g for 15 min, the resulting supernatants were subjected to immunoprecipitation with either anti-MIP-1␣ or anti-MIP-1␤ for mass spectrometric analysis.

Identification of a Heterodimer of MIP-1␣/␤ in Culture Supernatants of Human Monocytes and PBLs-Supernatants
were collected from cultures of either LPS-stimulated monocytes or PBLs stimulated with IL-2 and IL-12 after incubation for 1 or 6 days, respectively, times that correspond to peak chemokine secretion (data not shown). Enzyme-linked immunosorbent assays revealed that the culture supernatants contained large amounts of MIP-1␣ and MIP-1␤ but only a low concentration of RANTES (data not shown) (13). Immunoprecipitation of culture supernatants from both cell types with either anti-MIP-1␣ or anti-MIP-1␤ followed by immunoblot analysis with each of these antibodies revealed that each immunoprecipitate contained both MIP-1␣ and MIP-1␤ (Fig. 1A). The antigen affinity purified antibodies to each MIP-1 protein were specific and did not cross-react with the other isoform on Western blot analysis (Fig. 1B) or with immunoprecipitation followed by mass spectrometric analysis (Fig. 2B).
To determine the molecular size of the chemokines precipitated by anti-MIP-1␣ or anti-MIP-1␤, we analyzed the precipitated proteins by mass spectrometry. The two antibodies precipitated identical complexes from the culture supernatant of LPS-stimulated monocytes (Fig. 2). These complexes consisted predominantly of two polypeptides with molecular masses of 7459 and 7826 Da, which correspond to MIP-1␣ lacking the four NH 2 -terminal residues ((Ϫ4)MIP-1␣) and intact MIP-1␤, respectively. The complexes immunoprecipitated by each of the two antibodies from the culture supernatant of IL-2-and IL-12-stimulated PBLs contained (Ϫ4)MIP-1␣ and a protein with a molecular mass of 7658 Da, which corresponds to MIP-1␤ lacking the two NH 2 -terminal residues ((Ϫ2)MIP-1␤). Similar results were obtained with culture supernatants of PBLs stimulated with IL-2 and phytohemagglutinin (data not shown). Immunoprecipitates prepared from the monocyte and PBL culture supernatants with anti-RANTES or anti-MDC were also analyzed by mass spectrometry. However, this approach failed to detect either RANTES or MDC in the supernatants (data not shown). Thus, these data demonstrate that LPS-treated monocytes and PBLs stimulated with IL-2 and IL-12 release a MIP-1␣/␤ heterodimer into the culture supernatant.
MIP-1␣ and MIP-1␤ Are Secreted Predominantly as a Heterodimer-We next examined whether secreted MIP-1␣ and MIP-1␤ were present in the culture supernatant of LPS-stimulated monocytes predominantly as a heterodimer, or whether they were also present as monomers. The supernatant was subjected to immunoprecipitation first with anti-MIP-1␣ and then either with anti-MIP-1␤ or again with anti-MIP-1␣. The various immunoprecipitates were then subjected to immunoblot analysis with anti-MIP-1␣ or anti-MIP-1␤ (Fig. 3). Only a small amount of residual MIP-1␣ was detected in the second precipitate prepared with anti-MIP-1␣, indicating that the first immunoprecipitation with this antibody efficiently removed most of the target antigen. The observation that the second precipitate prepared with anti-MIP-1␤ contained little MIP-1␤ indicated that most of this protein had been previously precipitated as a heterodimer with MIP-1␣.
MIP-1␣/␤ Heterodimer in the Golgi Complex-We next investigated the kinetics and site of MIP-1␣/␤ dimerization. Immunoprecipitation and mass spectrometric analysis revealed the presence of the MIP-1␣/␤ heterodimer in the culture supernatants of monocytes or PBLs as early as 7-10 h after exposure to LPS or to IL-2 and IL-12, respectively (data not shown). To determine whether the heterodimerization of MIP-1␣ and MIP-1␤ occurs intracellularly in the Golgi complex, we treated cells with monensin to prevent intracellular protein transport by inducing retention of proteins in the Golgi complex (15,16). PBLs and monocytes were incubated with IL-2 and phytohemagluttinin A or with LPS, respectively, for 20 h, the last 10 h of which they were also exposed to monensin. Under these conditions, chemokines were produced primarily during the last 10 h of culture. The cells were then lysed and subjected to immunoprecipitation with anti-MIP-1␣ or anti-MIP-1␤, and the resulting precipitates were analyzed by mass spectrometry. The immunoprecipitates prepared from both PBLs (Fig. 4A) and monocytes (Fig. 4B) with either antibody consisting predominantly of heterodimers of (Ϫ4)MIP-1␣ and full-length MIP-1␤. The results revealed that MIP-1␤ was processed from the intact form (7826 Da) detected in the Golgi complex and in the culture supernatant that was obtained after 1 day (data not shown) to a mixture of both the full-length protein and (Ϫ2)MIP-1␤ (7658 Da) in the supernatant obtained after 3 days. After incubation of PBLs for 6 days, almost all of the MIP-1␤ had been converted to the NH 2 -terminal truncated form. In contrast, analysis of the culture supernatants of monocytes obtained after 1 day (Fig. 4B) or 3 days (data not shown) did not reveal proteolytic processing of MIP-1␤. Given that monensin only blocks protein transport to the post-Golgi compartment from the trans-Golgi network, these data indicate that MIP-1␣/␤ heterodimers probably form in the endoplasmic reticulum or cis-Golgi. Other experiments using brefeldin A to disrupt the Golgi complex also revealed the presence of heterodimers in cell lysates (data not shown) and suggest that heterodimers are formed in a pre-Golgi compartment. They also suggest that the proteolytic processing of MIP-1␤ released by PBLs may occur at a late or post-Golgi step during secretion. The MIP-1␣/␤ heterodimer was also detected in immunoprecipitates prepared from both PBL and monocyte lysates without monensin treatment (Fig. 4C), indicating that MIP-1␣/␤ dimerization occurs under normal conditions.
Dissociation of Native MIP-1␣/␤ Heterodimers under Acidic Conditions-To investigate the nature of the interaction between MIP-1␣ and MIP-1␤, we adjusted the culture supernatants of stimulated PBLs or monocytes to pH 4.0 before immunoprecipitation with anti-MIP-1␣ or anti-MIP-1␤. Under these low pH conditions, each antibody precipitated only the corresponding antigen (Fig. 5). These results suggest that acidic conditions induce the dissociation of the MIP-1␣/␤ heterodimer and that electrostatic interactions among charged amino acids therefore are likely to contribute to formation of the dimer interface.

DISCUSSION
The possible existence and activity of native chemokine dimers have been controversial (21). The ␤ chemokines MIP-1␣, MIP-1␤, and RANTES tend to self-associate and thereby form homodimers, tetramers, or larger aggregates in vitro. This process is dynamic and reversible, but it has been thought that chemokine concentrations in vivo may be too low for the formation of such multimers to occur. However, high local concentrations of chemokines may occur in vivo under certain conditions, such as during platelet degranulation, inflammatory disease, and local accumulation of chemokines on cell mem-branes mediated by receptors or by glycosaminoglycans.
We have now identified a naturally occurring MIP-1␣/␤ heterodimer produced by activated monocytes and PBLs. Our data demonstrate that the MIP-1␣/␤ heterodimer forms in the endoplasmic reticulum or Golgi complex, and that these two chemokines are secreted in the form of the heterodimer. Furthermore, the combination of MIP-1␣ with MIP-1␤ in vitro showed that these two chemokines indeed form heterodimers at physiological (nanomolar) concentrations.
Most studies on chemokine homodimerization have been performed in vitro, and the natural formation of either homodimers or heterodimers of chemokines has not been described previously. Native MIP-1 purified from LPS-stimulated mouse macrophage RAW 264.7 migrated on SDS-polyacrylamide gels as a doublet composed of peptides with similar physical characteristics. The NH 2 -terminal sequences of the two peptides identified them as MIP-1␣ and MIP-1␤ (22). Although it was not shown that the co-purified mouse chemokines originally existed as a heterodimer, our data now suggest that these previous results might be explained by the formation of a native mouse MIP-1␣/␤ heterodimer.
The three-dimensional structures of both ␣ and ␤ chemokines have been determined mostly with the molecules in the form of crystallized homodimers. Whereas the monomeric structures of both ␣ and ␤ chemokines are highly similar, the dimeric structures of members of these two subfamilies, as typified by IL-8 (12,20,23) and MIP-1␤ (9,11,20), respectively, differ markedly from each other with the dimer interfaces being formed by distinct sets of interacting residues. Whereas the IL-8 dimer is globular, the MIP-1␤ dimer is cylindrical. The three-dimensional structure of the ␣ chemokine PF4 revealed it to be a tetramer composed of two dimers of the IL-8 type (18, The molecular mass of the precipitated proteins was then determined by mass spectrometry. 24). NMR revealed MIP-1␤ as an end-on-end dimer with the dimer interface showing a large number of contacts between the two monomers. Other ␤ chemokines, including RANTES, MIP-1␣, MCP-1, and MCP-2, exhibit a similar homodimeric structure (6,8,17,25). MCP-3 remains monomeric at concentrations of up to 20 mg/ml (7), whereas I-309, another ␤ chemokine, was also shown to exist as a monomer at high concentrations during sedimentation (17).
A truncation mutant of MIP-1␤ lacking the NH 2 -terminal five residues forms a dimer similar to that formed by the wild-type protein, whereas a mutant lacking the NH 2 -terminal eight residues exists only as a folded monomer (11). An MCP-1 mutant lacking the NH 2 -terminal eight amino acids exists predominantly as a monomer (7). IL-8 and melanocyte growth stimulating activity each dimerize by the formation of a central six-stranded ␤ sheet, three strands of which are contributed by each subunit (26 -30). The calculation of the solvation-free energies of dimerization and analysis of clusters of hydrophobic amino acids indicates that the formation and stabilization of the two main types of chemokine homodimers result from the burial of hydrophobic residues and that the distinct quaternary structures are preserved throughout the two subfamilies (12,21). Such a scenario would explain the lack of receptor crossbinding and cross-reactivity, which is apparent between the ␣ and ␤ chemokine subfamilies.
Comparison of our present data with those of previous studies of chemokine homodimers indicates that substantial differences exist between MIP-1␣/␤ heterodimers and chemokine homodimers in terms of physical properties: (i) the formation of homodimers requires high concentrations of chemokines, whereas the MIP-1␣/␤ heterodimer forms at physiological concentrations of monomers; (ii) the dissociation of homodimers occurs under physiological conditions, whereas dissociation of MIP-1␣/␤ heterodimers is apparent only at low pH; (iii) homodimers have been detected only in solution under in vitro conditions, whereas MIP-1␣/␤ heterodimers are secreted from primary monocytes and PBLs; and (iv) the formation of the MIP-1␣/␤ heterodimer appears to be mediated by electrostatic interactions, whereas the formation of chemokine homodimers is thought to be mediated by hydrophobic interactions.
The existence of native chemokine homodimers remains to be demonstrated with the data having been obtained that are consistent or inconsistent with homodimers being the functionally active form of these proteins (12). Mutagenesis and cross-linking studies indicate that the active form of MCP-1 is a dimer (31). However, other studies have shown that IL-8 and MIP-1␤ derivatives that do not dimerize are fully active (32,33).
With regard to the functional role of MIP-1␣/␤ heterodimerization, it is possible that the formation of stable heterodimers protects these chemokines from enzymatic digestion and, thus, increases or stabilizes their activity. Preliminary data indicate that MIP-1␣/␤ heterodimer-containing mixtures have potent activity in inducing down-regulation of the CCR5 receptor. 2 The formation of the MIP-1␣/␤ heterodimer under natural conditions may induce (possibly heterologous) receptor dimerization that may have an impact on intracellular signaling events, which contribute to CCR5 and possibly other chemokine receptor functions. Production of pure homogeneous preparations of heterodimers is required for further characterization of heterodimer activities and receptor binding properties.