Biochemical Journal

Research article

Secreted CXCL12 (SDF-1) forms dimers under physiological conditions

Paramita Ray , Sarah A. Lewin , Laura Anne Mihalko , Sasha-Cai Lesher-Perez , Shuichi Takayama , Kathryn E. Luker , Gary D. Luker


Chemokine CXCL12 (CXC chemokine ligand 12) signalling through CXCR (CXC chemokine receptor) 4 and CXCR7 has essential functions in development and underlies diseases including cancer, atherosclerosis and autoimmunity. Chemokines may form homodimers that regulate receptor binding and signalling, but previous studies with synthetic CXCL12 have produced conflicting evidence for homodimerization. We used bioluminescence imaging with GL (Gaussia luciferase) fusions to investigate dimerization of CXCL12 secreted from mammalian cells. Using column chromatography and GL complementation, we established that CXCL12 was secreted from mammalian cells as both monomers and dimers. Secreted CXCL12 also formed homodimers in the extracellular space. Monomeric CXCL12 preferentially activated CXCR4 signalling through Gαi and Akt, whereas dimeric CXCL12 more effectively promoted recruitment of β-arrestin 2 to CXCR4 and chemotaxis of CXCR4-expressing breast cancer cells. We also showed that CXCR7 preferentially sequestered monomeric CXCL12 from the extracellular space and had minimal effects on dimeric CXCL12 in cell-based assays and an orthotopic tumour xenograft model of human breast cancer. These studies establish that CXCL12 secreted from mammalian cells forms homodimers under physiological conditions. Since monomeric and dimeric CXCL12 have distinct effects on cell signalling and function, our results have important implications for ongoing efforts to target CXCL12 pathways for therapy.

  • bioluminescence
  • breast cancer
  • chemokine
  • chemokine receptor
  • luciferase
  • protein fragment complementation


Chemokine CXCL12 (CXC chemokine ligand 12) [SDF-1 (stromal-cell-derived factor 1)] was originally identified as a growth factor for B-lymphocytes and a chemoattractant molecule for T-lymphocytes and monocytes [1]. In addition to effects on proliferation and trafficking of immune cells, CXCL12 has numerous other functions in development and normal physiology. Mice lacking CXCL12 die in utero with multiple abnormalities, including deficient vascularization of the GI (gastrointestinal) tract, heart defects, impaired myelopoiesis and perturbed migration of neurons in the CNS (central nervous system) [2,3]. CXCL12 also is essential for normal development of alveoli in the lung [4]. This chemokine is required for homing of haemopoietic stem cells to bone marrow, and inhibition of CXCL12 signalling through CXCR (CXC chemokine receptor) 4 is used to mobilize stem cells for bone marrow transplant [5]. Effects of CXCL12 on multiple organs and tissues are mediated through its receptors CXCR4 and CXCR7, which independently or collectively regulate chemotaxis and invasion of cells, increase cell adhesion and activate intracellular signalling pathways that control cell proliferation and survival.

Beyond critical functions in normal development and physiology, CXCL12 and its signalling pathways appear to underlie the pathogenesis of numerous diseases that are challenging to treat with current therapies. CXCL12 has been implicated in growth and organ-specific metastasis of more than 20 different human cancers, including lung, breast, prostate and ovarian [6]. Elevated levels of CXCL12 and its receptors are associated with poor prognosis and poor overall survival in many of these malignancies [7,8]. CXCL12 regulates progression of atherosclerosis, and this molecule recruits stem and progenitor cell populations to sites of ischaemic or infarcted tissue in sites including heart and brain [9,10]. CXCL12 also is associated with pathophysiology and progression of autoimmune diseases including rheumatoid arthritis and multiple sclerosis [11,12]. These studies highlight the rationale for developing CXCL12-targeted therapies and emphasize the need to understand the biology of CXCL12 to optimally utilize new drugs regulating this chemokine pathway.

Several chemokines form homodimers and heterodimers that activate signalling pathways distinct from monomeric proteins in vitro and in vivo [13,14]. However, data about homodimerization of CXCL12 are unclear. Crystal structures show CXCL12 as dimers, but NMR studies detect monomers at concentrations less than 5 mM in solution [1517]. The monomer–dimer equilibrium of CXCL12 is regulated by pH, phosphate and oligosaccharides with heparan sulfate and similar proteoglycans present on cell membranes and the extracellular space promoting dimerization [1820]. In the presence of heparin oligosaccharides, CXCL12 forms dimers at low micromolar concentrations, which are substantially less than concentrations required for dimerization of pure protein [21].

Studies using recombinant mutants of CXCL12 that favour dimers or monomers have produced inconsistent results for signalling and function. Using a monocytic leukaemia cell line, Veldkamp et al. [22] concluded that monomeric CXCL12 was the active form, whereas a dimeric mutant was a partial agonist that opposed chemotaxis. This research group also determined that only monomeric CXCL12 protected the heart from ischaemic damage in an ex vivo model [23]. However, a mutant of CXCL12 deficient in oligosaccharide binding and dimerization was less effective than wild-type chemokine as a chemoattractant for hepatoma cells, suggesting that dimeric CXCL12 increased migration of these cells [21]. Although these studies had different conclusions about the activities of monomers against dimers, the data support homodimerization of CXCL12 under physiological conditions and indicate that monomers and dimers have distinct effects on CXCL12-dependent signalling and function.

We used bioluminescence imaging strategies to investigate dimerization of CXCL12 secreted from mammalian cells. Using a GL (Gaussia luciferase) fusion to CXCL12 and GL complementation, we established that secreted CXCL12 forms dimers under physiological conditions. Monomeric and dimeric CXCL12 activated downstream signalling pathways and cell migration to differing extents in cell-based assays. CXCL12 monomers also were preferentially scavenged by CXCR7. These results advance our understanding of CXCL12 structure and function and inform ongoing efforts to target and utilize this chemokine for therapy.



pGloSensor-20F firefly luciferase reporter plasmid for cAMP was from Promega. The reporter was excised with HindIII and BamHI and transferred to the XbaI site of lentiviral vector FUGW by blunt end ligation.


HEK (human embryonic kidney)-293T cells and MDA-MB-231 breast cancer cells were obtained from the A.T.C.C. (Manassas, VA, U.S.A.) HEK-293T cells stably expressing unfused GL or CXCL12-GL (CXCL12 fused to GL) and MDA-MB-231 cells stably transduced with CXCR4, CXCR7 or vector control have been described previously [24]. HEK-293T and MDA-MB-231 cells were stably transduced with lentiviruses expressing CXCL12-NG (N-terminal fragment), CXCL12-CG (C-terminal fragment), unfused secreted CG, or both CXCL12-NG and CXCL12-CG complementation reporters [25]. MDA-MB-231 cells expressing both CXCL12-NG and CXCL12-CG were transduced with constitutively expressed eqFP650, a far-red fluorescent protein [26]. We also generated MDA-MB-231 cells expressing CXCL12-GL and pGlo Sensor 20F. Cells were cultured in DMEM (Dulbecco's modified Eagle's medium; Invitrogen), 10% fetal bovine serum, 1% glutamine and 0.1% penicillin/streptomycin/gentamicin. Cells were grown in a 37°C incubator with 5% CO2.

Column chromatography and analysis of GL activity in recovered fractions

HEK-293T-CXCL12-GL or HEK-293T-GL cells were cultured overnight in Phenol Red-free DMEM (Invitrogen). Supernatants were filtered through a PVDF membrane (Corning) and concentrated using an Amicon filter unit with a 10 kDa cut-off. A 1 ml volume of concentrated supernatants was applied to a 50 cm long, 1.5 cm inner diameter Kontes Flex column (Fisher Scientific) packed with Superdex 75 (Sigma) and equilibrated with column buffer [50 mM Tris/HCl (pH 7.5) and 100 mM KCl]. Size standards for elution were albumin (66 kDa), carbonic anhydrase (29 kDa) and cytochrome c (12 kDa). The void volume of the column was determined using Dextran Blue (200 kDa) (Sigma). The column flow rate was 800 μl/min. GL activity in column fractions was measured by bioluminescence imaging with coelenterazine as described in [24]. Fractions with monomeric and dimeric CXCL12-GL were determined based on protein standards. Fractionation was performed at 4°C, and recovered fractions were kept on ice until used in assays that began 3–4 h later.

Spheroid culture and imaging

Three-dimensional spheroids were cultured in a custom polystyrene HDAP (hanging drop array plate) coated with an amphiphilic solution (0.1% Pluronic F108, BASF) [27]. Plates were sterilized with UV light. Hanging drops for culturing spheroids were formed by pipetting 5×104 cells in 15 μl of complete medium from the top side of the access holes, allowing the sample liquid to form a hanging droplet on the bottom surface of the HDAP. Co-cultures were prepared with 2.5×104 cells of each type. Cells were cultured for 48 h before imaging.

Spheroids were washed four times in stepwise fashion with Phenol Red-free DMEM with 0.2% BSA 8 h before imaging. To quantify bioluminescence from secreted CXCL12 fusion proteins, 6 μl of supernatant from each hanging drop was collected before imaging spheroids. GL activity in spheroids and supernatants was measured on an IVIS 100 system (Caliper) immediately after adding 1 μg/ml coelenterazine. Data were quantified as photon flux [24].

Transwell migration assay

Cell migration was assayed in a 96-well MultiScreen-Mic (Millipore) transwell cell culture system. MDA-MB-231-CXCR4 cells were starved overnight, and 5×104 cells were plated in the top chamber in 100 μl of DMEM with 0.2% BSA. Isolated dimeric or monomeric CXCL12-GL from HEK-293T-CXCL12 cells was added to the lower chamber of each transwell. A fraction with no CXCL12-GL bioluminescence or medium with 5% serum was added to lower wells as negative and positive controls respectively. Assays were terminated after 8 h of migration, and cells in the top chamber were removed gently with cotton swabs. Cells that migrated to the lower surface of the membrane were fixed and stained with 0.2% Crystal Violet in methanol. Images were taken using an inverted microscope (Olympus 1×70), and data for percentage membrane surface occupied by transmigrated cells were quantified using ImageJ (NIH).

Western blotting

Cells were cultured in serum-free medium overnight and then stimulated for 10 min with equal amounts of bioluminescence from dimer and monomer fractions of CXCL12. Cell lysates were blotted for phosphorylated Akt (Cell Signaling Technology) as described previously [24]. Blots were stripped and re-probed for total Akt as a loading control. Relative intensities of bands were quantified with ImageJ and expressed as the ratio of phosphorylated to total Akt.

Cell-based imaging for cAMP

MDA-MB-231 cells stably expressing the cAMP reporter plasmid pGloSensor-20F were plated at 2×104 cells per well in black wall 96-well plates 1 day before assays. We incubated cells for 10 min with equal amounts of monomeric or dimeric CXCL12-GL fractions based on bioluminescence. Column fractions were diluted in Phenol Red-free DMEM. Immediately before imaging firefly luciferase bioluminescence, we added 5 μM forskolin (Sigma) and 15 mg/ml luciferin (Promega). We acquired bioluminescence images for 4 min with large binning on an IVIS 100 system.

Recruitment of β-arrestin 2 to CXCR4 or CXCR7

We quantified interaction of CXCR4 or CXCR7 with β-arrestin 2 by firefly luciferase complementation as described previously [28,29]. Cells were incubated with equal amounts of monomeric or dimeric CXCL12-GL fractions based on bioluminescence for 20 min (CXCR4) or 2 h (CXCR7) before quantifying firefly luciferase activity.

Cell-based assays for accumulation of bioluminescent CXCL12 dimers or monomers

MDA-MB-231-CXCR7 or MDA-MB-231-control cells were plated at 2×104 cells per well in 96-well black wall plates 1 day before assays. Cells were incubated for 1 h with monomeric or dimeric column fractions of CXCL12-GL. Monomeric and dimeric inputs were normalized to equal total amounts of CXCL12 based on bioluminescence. As a control, cells were incubated with a column fraction containing no CXCL12-GL. Bioluminescence from GL was measured as described previously and normalized to total protein per well quantified by sulforhodamine B staining [24].

For co-culture experiments, 104 MDA-MB-231-CXCR7 or MDA-MB-231-control cells were plated with equal numbers of MDA-MB-231 cells stably expressing CXCL12-GL or CXCL12-NG/CXCL12-CG. The following day, cells were washed with PBS and then incubated in Phenol Red-free DMEM. At various times, 10 μl samples of culture medium were collected and assayed for GL bioluminescence as described previously [24].

Animal studies

All animal procedures were approved by the University of Michigan Committee for the Use and Care of Animals. MDA-MB-231-CXCL12-NG/CXCL12-CG cells (5×105) were implanted with equal numbers of either MDA-MB-231-CXCR7 or MDA-MB-231-control cells into fourth inguinal mammary fat pads of 6-week-old female NOD/SCID (non-obese diabetic/severe combined immunodeficiency) IL2rγ−/− mice (Taconic). Blood samples of 20 μl were collected by retro-orbital puncture using heparinized capillary tubes. Blood was kept in these tubes for ~10 s before transfer to microfuge tubes. Blood recovered from capillary tubes clotted within 30 s of transfer to standard microfuge tubes. Samples from mice without tumours were used as negative controls. Collected blood was centrifuged in a microfuge at 100000 g for 10 min, and bioluminescence was quantified in 10 μl serum samples [30,31]. Bioluminescence in negative control samples was subtracted from values obtained from tumour-bearing mice.

Animal imaging

Bioluminescence imaging was performed on an IVIS Spectrum instrument (Caliper). For GL imaging, mice were injected intravenously via the tail vein with 4 mg/kg coelenterazine [30]. Mice were imaged immediately after injection using 3 min acquisition and large binning. Fluorescence imaging for eqFP650 was performed as described previously [26]. Data were quantified as photon flux (bioluminescence) or radiant efficiency (fluorescence) with Living Image software (Caliper).


Graphs and statistical analyses were prepared with GraphPad Prism. Results are plotted as means±S.E.M. Pairs of data were analysed by Mann–Whitney U test to determine statistically significant differences.


Secreted CXCL12 forms monomers and dimers

Studies analysing CXCL12 as monomers or dimers have used recombinant or synthetic fragments or full-length forms of chemokine. These studies show that dimers of CXCL12 only form at concentrations greater than amounts measured in cell culture supernatants or serum [15,16,19]. However, local concentrations of CXCL12 in the secretory pathway in mammalian cells may be higher than concentrations of chemokine in the extracellular space, which may promote dimerization. Binding of CXCL12 to heparan sulfate proteoglycans or other glycosaminoglycans on the surface of cells or released into solution also may increase local concentrations of chemokine to promote dimerization under physiological conditions [11,21,32].

To investigate monomers and dimers in secreted CXCL12, we used HEK-293T cells expressing CXCL12-GL. We demonstrated previously that CXCL12-GL is secreted from cells and activates signalling through CXCR4 to the same extent as synthetic CXCL12 [24]. We collected supernatants containing CXCL12-GL from HEK-293T cells and separated proteins based on size by gel-filtration chromatography. CXCL12-GL produced two peaks of bioluminescence in fractions at ~58 and 29 kDa respectively (Figure 1A and Supplementary Figure S1 at These molecular masses correspond to predicted sizes of dimers and monomers of CXCL12-GL respectively. There was approximately 10-fold more bioluminescence in the monomeric fraction, indicating that the CXCL12 secreted into the extracellular space is predominantly a monomer (Table 1). To exclude the possibility that GL promoted dimerization of CXCL12-GL, we also separated secreted GL by gel filtration. On the basis of bioluminescence, GL produced a single peak at the expected size (~20 kDa) for a monomer, showing that GL alone does not form dimers (Figure 1B). We note that recovered bioluminescence from GL was higher than combined values for both dimer and monomer fractions of CXCL12-GL because fusing another protein to this luciferase decreases its activity [33].

Figure 1 Secreted CXCL12-GL fractionates as dimers and monomers

(A) Supernatants from HEK-293T cells expressing CXCL12 fused to GL (CXCL12-GL) were fractionated by column chromatography. Results are bioluminescence values measured in each sample. Closed and open arrows denote fractions with bioluminescence from dimeric and monomeric CXCL12-GL at ~58 and 29 kDa respectively. (B) Bioluminescence in fractions from unfused GL secreted from HEK-293T cells. Grey arrow shows bioluminescence from monomeric GL at ~20 kDa.

View this table:
Table 1 Column chromatography for CXCL12-GL

We measured stability of isolated monomer and dimer fractions of CXCL12 in solution. After collecting the monomeric fraction of CXCL12-GL, we incubated this fraction for either 1 or 3 h at 37°C before repeating separation of this chemokine. The monomer fraction of CXCL12-GL was very stable within these periods. Following 1 h at 37°C, all recovered bioluminescence was in the monomer fraction, while the ratio of luminescence for dimer to monomer fractions was ~1:7 after 3 h (Table 1). Since we observed a small conversion of monomer into dimer of CXCL12-GL after 3 h at 37°C, we extended the incubation period to 3 days at either 37°C or 4°C before repeating the fractionation procedure. For samples incubated for 3 days at 37°C, substantially more CXCL12-GL was isolated as dimer upon repeat fractionation. Under these conditions, the ratio of dimer to monomer was ~3.5:1. Conversion of monomer into dimer of CXCL12-GL was substantially less in samples maintained for 3 days at 4°C with a dimer to monomer ratio of 1:1.3. By comparison, the isolated dimer fraction did not produce any monomer bioluminescence following incubation under these same conditions (results not shown). Stability of the dimer fraction may be because CXCL12-GL molecules remain associated with glycosaminoglycans released from HEK-293T cells secreting this chemokine. Collectively, these results demonstrate that CXCL12 secreted from mammalian cells under physiological conditions exists as monomer and dimers.

CXCL12 forms dimers in the extracellular space

We used GL protein fragment complementation to further investigate formation of dimers by CXCL12. GL complementation is based on dividing this enzyme into inactive N-terminal and C-terminal fragments (NG and CG) that do not associate spontaneously [34]. NG and CG are fused to potential interacting proteins of interest. When brought together by specific protein–protein interactions, NG and CG reconstitute an active enzyme and produce bioluminescence. Complementation between NG and CG fragments is completely reversible, so bioluminescence ceases when proteins dissociate. Since bioluminescence is produced only when separate NG and CG are brought together by interacting proteins, GL complementation provides a quantitative assay for protein association and dissociation.

To use GL complementation to analyse CXCL12 dimers, we fused CXCL12-α to either NG or CG of GL (CXCL12-NG and CXCL12-CG respectively). As a control, we also used unfused secreted CG [25]. We generated stable populations of HEK-293T cells that express CXCL12-NG, CXCL12-CG, CG or both CXCL12-NG and CXCL12-CG. Cells expressing both CXCL12-NG and CXCL12-CG were referred to as co-transduced. We initially co-cultured HEK-293T-CXCL12-NG cells with equal numbers of cells expressing either CXCL12-CG or CG in 96-well plates. Although we readily detected bioluminescence from co-transduced cells, we could not detect bioluminescence above background levels from two-dimensional co-cultures of CXCL12-NG and CXCL12-CG cells (results not shown).

We hypothesized that the relatively large volume of extracellular medium present in two-dimensional cultures lowered local concentrations of secreted CXCL12 and prevented dimerization. To reduce the extracellular volume and simulate compact three-dimensional intercellular relationships present in tumours and other sites in vivo, we used hanging drop cultures to generate spheroids from various combinations of these cells. We quantified GL bioluminescence in spheroids and cell supernatants after 2 days in culture.

Cells co-transduced with CXCL12-NG and CXCL12-CG produced the greatest bioluminescence in spheroids and supernatants, showing that complementation occurred between CXCL12-NG and CXCL12-CG. Co-expression of both fusion proteins in the same cell probably generates a high signal because dimers of CXCL12 fusion proteins form in the secretory pathway before release into the extracellular space (Figure 2 and Supplementary Figure S2 at By comparison, cells co-expressing CXCL12-NG and unfused CG did not produce detectable bioluminescence above background, showing specificity of the complementation signal (results not shown). We also detected GL complementation in spheroids formed from co-cultures of cells expressing CXCL12-NG or CXCL12-CG, albeit at lower levels than cells co-transduced with both constructs. Supernatants from spheroids combining CXCL12-NG with CXCL12-CG cells also produced GL bioluminescence (Figure 2B). GL signal in co-cultures is only produced by dimerization of secreted CXCL12-NG and CXCL12-CG proteins in the extracellular space, so bioluminescence selectively measures steady-state levels of CXCL12 dimers. These dimers remain associated with spheroids or are released into culture supernatants. By comparison, combinations of cells expressing CXCL12-NG and unfused CG had no detectable bioluminescence in spheroids or supernatants, confirming that non-specific association of NG and CG does not produce bioluminescence. Spheroids with only CXCL12-NG or CXCL12-CG cells also had no GL signal associated with cells or supernatants. These data establish that secreted CXCL12 forms dimers in the extracellular environment under physiological conditions.

Figure 2 GL complementation shows dimerization of secreted CXCL12 in spheroids and culture supernatants

(A) Spheroids were prepared from control HEK-293T cells, cells stably expressing CXCL12-NG only, CXCL12-CG only, both CXCL12-NG and CXCL12-CG (Co-Tx), mixtures of cells with CXCL12-NG or unfused CG, or mixtures of cells expressing CXCL12-NG or CXCL12-CG. Results are means±S.E.M. for GL bioluminescence in spheroids after 2 days in culture (n=8 per condition). Inset shows data for all cell combinations except co-transduced spheroids that express both CXCL12-NG and CXCL12-CG. (B) Supernatants from spheroids in (A) were assayed for bioluminescence from GL complementation for CXCL12. Results are means±S.E.M. Inset shows values for all conditions except cells co-transduced with both CXCL12-NG and CXCL12-CG. **P<0.01; ***P<0.005.

Differential effects of monomeric and dimeric CXCL12 on CXCR4-dependent signalling and function

CXCL12 binding to CXCR4 activates downstream signalling pathways including inhibition of cAMP through Gαi, recruitment of the cytosolic adapter protein β-arrestin 2 and phosphorylation of Akt. We analysed the effects of dimeric and monomeric CXCL12 on these signalling events in breast cancer cells. We used MDA-MB-231 human breast cancer cells stably transduced with a firefly luciferase biosensor for cAMP. We incubated cells with equal amounts of dimeric or monomeric CXCL12-GL based on bioluminescence or vehicle only for 10 min before treatment with forskolin to elevate intracellular cAMP. Relative to control cells, monomeric CXCL12-GL suppressed intracellular cAMP to a significantly greater extent than dimeric CXCL12-GL following treatment with forskolin (P<0.05) (Figure 3A).

Figure 3 Differential regulation of CXCR4 signalling by monomeric and dimeric CXCL12

(A) MDA-MB-231-CXCR4 cells expressing a firefly luciferase reporter for cAMP were incubated with monomeric or dimeric CXCL12-GL fractions or medium alone for 10 min before treatment with forskolin. Results are means±S.E.M. for bioluminescence from the cAMP reporter. (B) MDA-MB-231 cells expressing luciferase complementation reporters for CXCR4 interaction with β-arrestin 2 were incubated for 20 min with monomeric or dimeric fractions of CXCL12-GL. Results are means±S.E.M. for bioluminescence from firefly luciferase complementation. (C) MDA-MB-231-CXCR4 cells were serum-starved overnight and then treated for 10 min with medium alone or equal amounts of monomeric or dimeric CXCL12-GL based on bioluminescence. Western blotting was used to detect phosphorylated (active) Akt. The blot then was stripped and analysed for total Akt and then GAPDH (glyceraldehyde-3-phosphate dehydrogenase) as loading controls. Numbers are quantified data for intensities of bands expressed as phosphorylated Akt divided by total Akt. *P<0.05.

We have developed a firefly luciferase complementation assay to quantify recruitment of β-arrestin 2 to CXCR4 in response to ligand binding [28]. We treated MDA-MB-231 breast cancer cells stably expressing the CXCR4 and β-arrestin 2 complementation pair with monomeric or dimeric fractions of CXCL12-GL for 20 min before quantifying bioluminescence. Both monomeric and dimeric fractions of CXCL12-GL increased association of CXCR4 with β-arrestin 2 as compared with control cells (Figure 3B). Effects of dimeric CXCL12-GL were significantly greater than the monomeric fraction of this chemokine (P<0.05). We also analysed CXCL12-dependent activation of Akt, an established downstream effector of CXCR4 signalling through G-proteins. Following overnight culture in serum-free medium, cells were treated for 10 min with either monomeric or dimeric fractions of CXCL12-GL. Although both monomeric and dimeric CXCL12-GL activated Akt above control, the monomeric fraction produced substantially greater activation (Figure 3C).

CXCL12 signalling through CXCR4 and/or CXCR7 promotes chemotaxis of multiple cell types. To determine to what extent chemotaxis is affected by dimerization of CXCL12, we tested effects of monomeric and dimeric CXCL12 on transwell migration of MDA-MB-231 breast cancer cells stably transduced with CXCR4 (MDA-MB-231-CXCR4) [24]. We isolated monomer and dimer fractions of CXCL12-GL and added equal amounts of light for each fraction to lower wells of a transwell migration system. This strategy adds equivalent numbers of molecules of CXCL12 to each well since the chemokine is fused directly to GL. As negative and positive controls, we added a column fraction containing no CXCL12-GL bioluminescence or medium containing 5% serum respectively. MDA-MB-231-CXCR4 cells showed robust migration towards the fraction with dimeric CXCL12-GL, which was essentially the same as migration towards 5% serum (Figure 4). However, migration of MDA-MB-231-CXCR4 cells towards monomeric CXCL12-GL was only marginally greater than a fraction with no CXCL12-GL activity. Collectively, these studies show that dimeric and monomeric CXCL12 secreted from mammalian cells preferentially activate distinct aspects of CXCR4 signalling and function in breast cancer.

Figure 4 Dimeric CXCL12 promotes chemotaxis of breast cancer cells expressing CXCR4

(A) Monomeric and dimeric fractions of CXCL12-GL were isolated, and equal amounts of bioluminescence from each fraction were added to the bottom chamber of transwells. Additional wells had a fraction with no CXCL12-GL bioluminescence or a fraction with 5% serum added. MDA-MB-231 breast cancer cells stably transduced with CXCR4 (MDA-MB-231-CXCR4) were added to the top chamber. Cells that migrated through the filter were stained with Crystal Violet. Photograph shows representative staining of breast cancer cells that migrated through the transwell membrane. (B) Percentage membrane area occupied by transmigrated breast cancer cells plotted as means±S.E.M. **P<0.01.

CXCR7 preferentially sequesters monomeric CXCL12

In addition to CXCR4, CXCL12 also binds to CXCR7 [35,36]. CXCR7 functions at least in part to sequester CXCL12 from the extracellular space and degrade it, thereby controlling levels and gradients of this chemokine available for signalling [37,38]. We used two complementary approaches to investigate uptake of dimeric and monomeric CXCL12 by CXCR7. First, we separated CXCL12 into dimer and monomer fractions by column chromatography and incubated MDA-MB-231 cells expressing CXCR7 (MDA-MB-231-CXCR7) or vector control (MDA-MB-231-control) with equal amounts of chemokine by bioluminescence [38]. We measured cell-associated GL bioluminescence after 1 h and determined that MDA-MB-231-CXCR7 cells accumulated significantly more monomeric than dimeric CXCL12-GL (P<0.05) (Figure 5A). Uptake of dimeric CXCL12-GL did not differ between MDA-MB-231-CXCR7 and MDA-MB-231-control cells, and total amounts of cell-associated bioluminescence were at background levels quantified with the control fraction containing no CXCL12. Although accumulation of monomeric CXCL12-GL was slightly higher in MDA-MB-231-control cells, MDA-MB-231-CXCR7 cells had 2.2-fold more monomeric CXCL12-GL than control cells, while uptake of dimeric CXCL12-GL did not differ significantly from control.

Figure 5 CXCR7 preferentially interacts with monomeric CXCL12

(A) MDA-MB-231 breast cancer cells stably expressing CXCR7 (MDA-MB-231-CXCR7) or vector control (MDA-MB-231-control) were incubated with equal amounts of input light from dimeric and monomeric fractions of CXCL12-GL. Cells were incubated with CXCL12-GL fractions for 1 h and then washed with an acidic solution to remove extracellular chemokine before imaging GL activity. Results are means±S.E.M. for each condition (n=4). (B) MDA-MB-231-CXCR7 or MDA-MB-231-control cells were co-cultured with MDA-MB-231 cells stably transduced with CXCL12-NG and CXCL12-CG. Bioluminescence from dimeric CXCL12-NG/CXCL12-CG in cell culture medium was quantified at times up to 6 h and plotted as means±S.E.M. (n=4 per time point). (C) Co-cultures of MDA-MB-231-CXCR7 or MDA-MB-231-control cells with MDA-MB-231 cells secreting CXCL12-GL were incubated as described in (B). Bioluminescence from CXCL12-GL in the extracellular space was quantified at various times. Results are means±S.E.M. (n=4 per time point). (D) MDA-MB-231 cells expressing luciferase complementation reporters for association of CXCR7 with β-arrestin 2 were incubated for 2 h with monomeric or dimeric fractions of CXCL12-GL. Results are means±S.E.M. *P<0.05; **P<0.01.

As a second approach, we used co-cultures of MDAMB-231-CXCR7 or MDA-MB-231-control cells with MDA-MB-231 cells co-transduced with either CXCL12-NG/CXCL12-CG or CXCL12-GL. The co-culture format models human breast tumours that contain cells expressing CXCR7 and/or secreting CXCL12 in the tumour microenvironment [39,40]. Co-cultures using MDA-MB-231 cells co-transduced with CXCL12-NG/CXCL12-CG allowed us to selectively quantify uptake of CXCL12 dimers, since GL complementation occurs only when these fusion proteins interact. We quantified bioluminescence from CXCL12-NG/CXCL12-CG dimers in supernatants at various times through 6 h. Bioluminescence increased progressively over time and did not differ between co-cultures with MDA-MB-231-CXCR7 or MDA-MB-231-control cells (Figure 5B). However, in co-cultures using MDA-MB-231 cells expressing CXCL12-GL, there was significantly less chemokine in the extracellular space when these cells were combined with MDA-MB-231-CXCR7 cells (Figure 5C). On the basis of data with column chromatography, secreted CXCL12-GL predominantly exists as a monomer, confirming data for preferential uptake of monomeric CXCL12 by CXCR7.

Chemokine binding to CXCR7 recruits β-arrestin 2, which can be quantified by firefly luciferase and other complementation strategies [29,41]. We treated MDA-MB-231 cells stably expressing luciferase complementation reporters for CXCR7 and β-arrestin 2 for 2 h with equal amounts of monomeric and dimeric CXCL12-GL based on bioluminescence. Although both monomeric and dimeric fractions of CXCL12-GL increased association of CXCR7 with β-arrestin 2, monomeric CXCL12-GL produced a significantly greater effect (P<0.05) (Figure 5D). The magnitude of difference between monomeric and dimeric CXCL12-GL is greater for CXCR7-dependent accumulation of monomeric chemokine than recruitment of β-arrestin 2. We have shown previously that accumulation of CXCL12 by CXCR7 only is partially dependent on β-arrestin 2 [29]. These data suggest that β-arrestin 2-independent mechanisms for chemokine uptake by CXCR7 also favour monomeric CXCL12-GL. Collectively, these data establish that CXCR7 preferentially interacts with monomeric CXCL12-GL to remove this chemokine from the extracellular space.

CXCR7 has minimal effect on CXCL12 dimers in breast tumours

We demonstrated recently that MDA-MB-231-CXCR7 cells scavenge CXCL12-GL in orthotopic breast cancer xenografts, reducing amounts of chemokine detectable in primary tumours and serum [42]. To investigate effects of CXCR7 on amounts of dimeric CXCL12 in breast tumours, we implanted either MDA-MB-231-CXCR7 or MDA-MB-231-control cells with MDA-MB-231 cells co-transduced with CXCL12-NG/CXCL12-CG into mammary fat pads of mice. MDA-MB-231 cells with CXCL12-NG/CXCL12-CG also were transduced with a far-red fluorescent protein, eqFP650, to monitor relative numbers of these cells in each tumour [26]. After tumours reached ~8 mm diameter, we imaged bioluminescence from CXCL12-NG/CXCL12-CG in primary tumours and quantified amounts of dimeric chemokine released into sera of tumour-bearing mice.

MDA-MB-231-CXCR7 cells did not significantly lower amounts of bioluminescence from CXCL12-NG/CXCL12-CG in primary tumours relative to tumours with MDA-MB-231-control cells (Figures 6A and 6B). MDA-MB-231-CXCR7 cells in primary tumours also did not alter amounts of CXCL12-NG/CXCL12-CG released into sera obtained from these mice (Figure 6C). Although blood samples were collected in heparizined capillary tubes, blood was in these tubes for only ~10 s and clotted rapidly when transferred out of the capillary tube. Continued clotting suggests very minimal transfer of heparin from the tube to blood. Comparing the results with dimeric CXCL12-NG/CXCL12-CG, we established recently that MDA-MB-231-CXCR7 cells decrease amounts of CXCL12-GL in the tumour microenvironment and secreted into sera of tumour-bearing mice [42]. These results provide further evidence that CXCR7 is less effective at scavenging dimeric CXCL12 rather than the monomeric form of this chemokine.

Figure 6 CXCR7 minimally alters levels of CXCL12 dimers in vivo

(A) MDA-MD-231 breast cancer cells expressing both CXCL12-NG and CXCL12-CG were implanted orthotopically into mammary fat pads with either MDA-MB-231-CXCR7 or MDA-MB-231-control cells respectively. Representative images are shown for GL bioluminescence produced by CXCL12-NG/CXCL12-CG dimers in primary tumours. Fluorescence image of the same mice shows eqFP650 in MDA-MB-231-CXCL12-NG/CXCL12-CG cells. Scale bars depict range of pseudocolours used to display bioluminescence and fluorescence imaging data. (B) Quantified data for dimeric CXCL12-NG/CXCL12-CG bioluminescence in primary tumours normalized to fluorescence from eqFP650 in each tumour (n=5 mice per group). Results are means±S.E.M. (C) Blood samples were collected from mice in each group, and bioluminescence from CXCL12-NG/CXCL12-CG dimers was quantified in serum (n=5 per group). Results are means±S.E.M. for GL bioluminescence corrected for background values from mice with no tumours.


CXCL12 has key functions in pathogenesis of multiple diseases, making it a promising therapeutic target. For diseases such as cancer, treatments are focused on blocking functions of CXCL12 and its receptors CXCR4 and CXCR7. In pre-clinical models, inhibitors of CXCL12 signalling limit tumour growth and metastasis when administered as single-agent therapy, and antagonists of CXCL12–CXCR4 improve efficacy of standard chemotherapeutic drugs [36,43,44]. CXCL12 also promotes chemotaxis of stem and progenitor cells, so the chemokine is being investigated as a possible therapeutic agent to increase stem cell trafficking for tissue regeneration and repair in settings including myocardial infarction and ischaemic vascular disease [45,46]. To enable development and optimal utilization of CXCL12-targeted treatments, it is essential to identify biologically active forms of this chemokine for specific signalling pathways and cellular functions.

We established that CXCL12 secreted from mammalian cells forms dimers under physiological conditions. Using GL complementation to detect and quantify interactions between CXCL12 molecules in cell-based assays and living mice, we showed that stable dimers of this chemokine were secreted from cells. We also demonstrated that CXCL12 dimers formed in the extracellular space when cells expressing either CXCL12-NG or CXCL12-CG were co-cultured in spheroids. Since spheroid cultures model the compact three-dimensional intercellular interactions that occur in normal organs and tumours, these data suggest that CXCL12 forms dimers in vivo. Although we established that GL does not artificially promote dimerization of CXCL12, fusing GL to CXCL12 possibly could increase formation of monomers, in which case our results would underestimate the relative abundance of dimeric CXCL12.

We have shown previously that secreted CXCL12-GL and CXCL12 GL complementation proteins are present at 15–30 ng/ml in cell culture supernatants [24,25]. These concentrations are lower than values reported for dimerization of pure synthetic CXCL12 proteins in solution or crystals. Previous studies have shown that basic amino acids in CXCL12 confer binding to glycosaminoglycans and substantially reduce concentrations of synthetic chemokine required for dimerization in solution [47,48]. It is likely that dimerization of secreted CXCL12 is increased by binding to glycans and glycosaminoglycans on cell membranes and released into the extracellular space, which increases local concentrations of chemokine. Dimerization of CXCL12 also may be enhanced because the chemokine is concentrated in endosomes in the secretory pathway. While the present paper was under review, Drury et al. [49] showed that CXCL12 purified from bacteria also existed as monomers and dimers, although dimers did not form with less than 0.5 μg of purified protein. Overall, these studies support the conclusion that dimerization of CXCL12 occurs physiologically.

Monomeric and dimeric CXCL12 have distinct profiles for receptor interaction, signalling and cell function. Monomeric CXCL12 more effectively signalled through CXCR4 to suppress cellular cAMP and activate Akt, both of which are dependent upon G-protein pathways. By comparison, dimeric CXCL12 produced greater recruitment of β-arrestin 2 and was a more potent chemoattractant molecule for migration of breast cancer cells expressing CXCR4. Since CXCR4 signalling through β-arrestin 2 is required for chemotaxis, our data show that dimeric CXCL12 preferentially activates this component of CXCR4 signal transduction and function [50]. Our results support prior work by Fermas et al. [21], who showed reduced chemotaxis of hepatoma cells toward a synthetic mutant of CXCL12 with minimal binding to oligosaccharides [CXCL12 (3/6)] and normal affinity for CXCR4. These data suggest that dimerization of CXCL12 mediated by glycan molecules was essential for cell migration. By comparison, Veldkamp et al. [22] concluded that only monomeric CXCL12 promoted chemotaxis, whereas dimeric chemokine had no effect. Potentially, these differences may be due to use of recombinant purified CXCL12 against chemokine secreted from mammalian cells in the context of other secreted molecules. Discordant results among these studies also suggest cell-type-specific differences for biological effects of monomeric against dimeric CXCL12. Further studies are needed to establish effects of monomeric and dimeric CXCL12 on signal transduction and resultant functions of cells in vivo.

In addition to differences in signalling and chemotaxis, we also determined that CXCR7 preferentially sequesters monomeric CXCL12 both in cell culture assays and in a mouse model of breast cancer. CXCR7 functions as a scavenger receptor to remove CXCL12 from the extracellular space and degrade it in lysosomes [37,38]. By controlling levels of CXCL12 in the extracellular space, the scavenger function of CXCR7 is proposed to maintain and regulate CXCL12 signalling through CXCR4 [51,52]. Greater scavenging of monomeric against dimeric CXCL12 by CXCR7 may preferentially promote selective CXCL12–CXCR4-dependent signalling pathways, such as chemotaxis, in normal development and disease.

Our results provide evidence that dimerization of CXCL12 occurs under physiological conditions, and effects of CXCL12 dimers in cell signalling and function are distinct from those of monomeric CXCL12. Factors that regulate the monomer–dimer equilibrium for CXCL12, including pH and glycosaminoglycans on cell membranes, are known to change among normal and diseased cells in various anatomical sites. Shifts in proportions of monomeric and dimeric CXCL12 probably will substantially alter types and/or the extent of pathways activated in response to this chemokine. The GL fusion proteins used in the present study provide new approaches for detecting and quantifying dimerization of CXCL12 or other chemokines in cell-based assays and mouse models. This strategy and complementary methods will further advance understanding of CXCL12 monomers and dimers in normal physiology and disease conditions, which will facilitate ongoing efforts to target CXCL12 for therapy.


Paramita Ray, Sarah Lewin, Laura Anne Mihalko and Sasha-Cai Lesher-Perez performed experiments. Shuichi Takayama, Kathryn Luker and Gary Luker provided new reagents. Paramita Ray, Sarah Lewin, Laura Anne Mihalko, Kathryn Luker and Gary Luker analysed data. Paramita Ray and Gary Luker wrote the paper. Kathryn Luker and Gary Luker supervised the project.


This research was supported by the National Institutes of Health [grant numbers R01CA136553, R01CA136829 and P50CA093990].


We thank Dmitriy Chudakov for providing fluorescent protein eqFP650.

Abbreviations: CG, C-terminal fragment; CXCL, CXC chemokine ligand; CXCL12-GL, CXCL12 fused to GL; CXCR, CXC chemokine receptor; DMEM, Dulbecco's modified Eagle's medium; GL, Gaussia luciferase; HDAP, hanging drop array plate; HEK, human embryonic kidney cells; NG, N-terminal fragment


View Abstract