An Essential Role of the Cytoplasmic Tail of CXCR4 in G-Protein Signaling and Organogenesis

CXCR4 regulates cell proliferation, enhances cell survival and induces chemotaxis, yet molecular mechanisms underlying its signaling remain elusive. Like all other G-protein coupled receptors (GPCRs), CXCR4 delivers signals through G-protein-dependent and -independent pathways, the latter involving its serine-rich cytoplasmic tail. To evaluate the signaling and biological contribution of this G-protein-independent pathway, we generated mutant mice that express cytoplasmic tail-truncated CXCR4 (ΔT) by a gene knock-in approach. We found that ΔT mice exhibited multiple developmental defects, with not only G-protein-independent but also G-protein-dependent signaling events completely abolished, despite ΔT's ability to still associate with G-proteins. These results reveal an essential positive regulatory role of the cytoplasmic tail in CXCR4 signaling and suggest the tail is crucial for mediating G-protein activation and initiating crosstalk between G-protein-dependent and G-protein-independent pathways for correct GPCR signaling.


Introduction
CXCR4 is a seven-transmembrane GPCR for the chemokine, CXCL12. Both CXCR4 and CXCL12 are broadly expressed by cells of multiple tissues and play an indispensable role in embryogenesis [1]. Genetic ablation of CXCR4 or CXCL12 leads to embryonic lethality, as a result of defects in cardiogenesis, vascular development, hematopoiesis, and the CNS [2][3][4][5][6]. In adulthood, CXCR4 and CXCL12 have been implicated in pathogenesis of autoimmune diseases and tumor metastasis [7][8][9][10]. However, the precise molecular mechanisms that underlie these diverse physiological and pathological functions remain obscure.
Like the majority of GPCRs, CXCR4 contains a highly conserved DRY motif (Asp-Arg-Tyr) located in the second intracellular loop. Extensive studies using rhodopsin and adrenergic receptors as models have established a general paradigm for GPCR activation. It proposes that ligation of GPCR triggers protonation of the Asp residue in the DRY motif, inducing conformational changes of the GPCR and activation of the interacting G proteins [11,12]. Mutation of the DRY motif of chemokine receptors prevents ligand-induced activation of the pertussis toxin (PTX)-sensitive Gai proteins and abolishes generation of second messengers and chemotaxis, indicating a pivotal role of the DRY motif in G-protein mediated signaling [13][14][15][16]. Increasing evidence shows GPCRs may also exert biological effects independent of G-protein function. The Cterminal (CT) tail of GPCRs is rich in serines and threonines, and truncation of the tail of several chemokine receptors abrogates ligand-activated receptor phosphorylation, demonstrating that the tail of these receptors is the only phosphorylation target of GPCR kinases (GRKs) [17,18]. Phosphorylated GPCR tail binds to barrestins, leading to rapid internalization and desensitization of the ligand-activated receptor [19,20]. In addition to mediating receptor internalization, b-arrestins also serve as scaffold proteins, recruiting Src family tyrosine kinases to the phosphorylated GPCRs and consequently activate MAP kinases [21]. Given that GPCRs may deliver signals through the DRY motif and its cytoplasmic tail, it is important to determine whether the DRY motif and the tail of CXCR4 act as independent signaling transduction modules that carry out distinct cellular functions.
The functional importance of the CT tail of CXCR4 has been underscored by identification of truncating mutations of CXCR4 in patients with WHIM (warts, hypogammaglobulinemia, immunodeficiency, and myelokathexis) syndrome. WHIM patients carry autosomal dominant mutations in Cxcr4 that eliminate a part of the serine-rich CT tail [22]. Considerable studies have been conducted using mutant cells from WHIM patients or a variety of cell lines transfected with truncational mutants of CXCR4 to investigate WHIM pathogenesis. While all these data show that deletion of the tail impairs ligand induced receptor internalization, the biochemical and cellular responses, however, seem to be highly variable in these in vitro systems, with variations in MAPK activation and chemotaxis [23][24][25][26][27][28]. These discrepancies could be attributed to different expression levels of the transgenic CXCR4 as well as to different signaling machinery available in the utilized cell lines. To overcome these problems, we generated mutant mice that express tail-truncated CXCR4 by a ''knock-in'' approach and used these mice to investigate the developmental, cellular, and biochemical functions of the CXCR4 tail under physiological conditions. Results of the present study reveal that truncation of the CT tail of CXCR4 not only obliterates G-protein independent signaling pathways mediated by tail-associated factors, but also prevents signaling through Gai, resulting in similar developmental defects as seen in CXCR4-null mice.

Generation of CXCR4-DT mice
The cytoplasmic tail of CXCR4 contains 16 serine residues which are the putative targets of GRKs. In order to evaluate the precise biological functions mediated by the CT tail of CXCR4 we removed the last 42 amino acids from CXCR4 (aa 318-359), thereby completely eliminating the sequences that confer GRK activity and b-arrestin recruitment (Fig. 1A, mutant CXCR4 is hereafter denoted as DT). To ensure a physiological expression of DT, we generated ''knock-in'' mice ( Fig. S1). Our results showed that DT was expressed at an identical level to that of wildtype (WT) CXCR4 on myeloid cells in embryonic day (E) 18.5 fetal liver (Fig. 1B). Furthermore, DT bound CXCL12 with an affinity similar to that of WT CXCR4 (Fig. 1C). In this report, we refer to the mutant mice carrying the CXCR4 tail-truncated allele as ''DT mice''.
DT mutants phenocopy the developmental defects of CXCR4-null mice Previous reports have shown that CXCR4-null mice exhibit multiple developmental defects and die perinatally [3,4,29]. Such an embryonic lethal phenotype was not rescued in DT mice, however, unlike CXCR4-null mice of which only 50% survived till E17.5 and all with reduced body sizes [4], DT embryos were present at Mendelian ratios till E18.5 with grossly normal body sizes and morphology (data not shown).
To determine to what extent the DT mutation affected organogenesis, we examined those organs of which formation was critically dependent on CXCR4. We found that, similar to that observed in CXCR4-null mice, DT mice lacked large vessels in the gastrointestinal area ( Fig. 2A); had displaced cells of the external granule layer (EGL) in the cerebellum (Fig. 2B); defective B cell genesis (Fig. 2C) and severely impaired bone marrow (BM) myelopoiesis (Fig. 2D). These results together indicate that, although the overall impact on animal development is relatively milder, the DT mutation imposes almost identical effects as the CXCR4-null mutation on the development of neural, vascular and hematopoietic systems.
Since DT mice were embryonic lethal, we generated CXCR4 f/DT mice and crossed them to transgenic CD19-Cre (B cell specific) and ROSA ET2-CRE mice (tamoxifen induced) to delete the floxed Cxcr4 allele and conditionally express DT in specific lineages of cells. We also established pro-B cell lines by Abelson virus transformation for in vitro studies to determine cellular and biochemical functions of the CXCR4 tail.
The CXCR4 tail is critical for CXCL12-induced chemotaxis but dispensable for integrin-mediated adhesion At the late gestation, the primary site of hematopoiesis shifts from the fetal liver (FL) to BM. This process is dependent on CXCR4-mediated migration of hematopoietic stem cells (HSCs) [2,4,29]. Since myelopoiesis was relatively normal in the FL but defective in BM of DT mice (Fig. 2D), it was conceivable that the DT mutation impaired BM homing of hematopoietic cells. To determine whether the CXCR4 tail was required for chemotaxis, we measured the chemotactic response of DT pro-B cells toward CXCL12 by a transwell assay. We found that while 15-20% of WT pro-B cells underwent chemotaxis toward CXCL12, less than 2% of DT cells migrated under the same condition (Fig. 3A). Abolishment of CXCL12-induced chemotaxis was also found for the FL hematopoietic cells (Fig. 3B), demonstrating that the CXCR4 tail is essential for CXCL12-mediated chemotaxis.
We have shown previously that CXCR4 signaling promotes retention of B-cell precursors in the BM [30], presumably through VCAM-1/VLA-4 mediated firm adhesion [31,32] and lack of CXCR4 leads to the release of these precursors into the periphery.
To examine whether DT is still able to facilitate the retention of B-cell precursors in the BM, we analyzed B-cell subsets in peripheral blood and spleen of CD19-Cre-CXCR4 f/DT (WT), CD19-Cre+ CXCR4 f/DT (DT), and CD19-Cre+ CXCR4 f/f mice (CXCR4 null) by flow cytometry. In WT animals, B cell precursors (B220 lo IgM 2 ) were barely detectable in PB (Fig. S2). When Cxcr4 was deleted, a substantial amount of B cell precursors were released from the BM to the periphery (Fig. 4A) and this population was significantly reduced in the in the blood and spleen of DT mice, suggesting that DT partially restored BM retention ( Fig. 4A and Fig. S2).
To directly examine whether the cytoplasmic tail of CXCR4 was required for triggering integrin-mediated cell adhesion, we decided to test in vitro the ability of DT to promote B cell adherence to VCAM-1. To exclude the possible interference from CXCR7, the other receptor for CXCL12, we used DT, CXCR4-null and WT pro-B cells in a static adhesion assay. These pro-B cells have an identical cell-surface level of VLA-4 (receptor for VCAM-1) but do not express CXCR7 (Fig. S3). As shown in Fig. 4B, WT pro-B cells exhibited a 4-fold increase in VCAM-1 adhesion after CXCL12 stimulation, whereas CXCR4-null cells exhibited only basal VCAM-1-binidng activity. Notably, CXCL12 induced adherence of DT cells to VCAM-1 was comparable to that of WT (Fig. 4B). These results indicate that tail-truncated CXCR4 is still able to mediate inside-out signaling that activates VLA-4. Interestingly, the adhesion of WT and DT mutant pro-B cells to VCAM-1 was not affected by PTX pretreatment (Fig. S4), suggesting that activation of VLA-4 was independent of Gai signals.

CT tail deletion abolishes b-arrestin binding and impairs CXCR4 internalization
Whether the tail of CXCR4 is the only cytoplasmic domain that recruits b-arrestins remains controversial. Whereas a number of studies have shown that CXCR4 is primarily associated with barrestins through its CT tail [27,33], others can co-IP tailtruncated CXCR4 with b-arrestins [26,34] or detect it co-localized with b-arrestins by microscopy [25]. To detect the interaction between b-arrestins and DT, we performed a Tango assay [35] in which ligand-induced rapid and transient interaction between CXCR4 and b-arrestins is converted into an accumulative transcriptional response (Fig. 5A). We found that reporter cells expressing WTCXCR4-tTA exhibited an increased luciferase activity after CXCL12 stimulation. In contrast, such CXCL12induced luciferase gene activation was not detected in DTCXCR4-tTA-expressing reporter cells (Fig. 5B). This provides strong evidence that association between CXCR4 and b-arrestin2 absolutely depends on its cytoplasmic tail.
Interaction of b-arrestins and the phosphorylated CT tail of GPCRs has been shown to cause agonist-induced receptor internalization, a process considered as an important step in signal desensitization [36]. As shown in Fig. 5C, for pro-B cells, CXCL12 treatment induced internalization of 35-40% of cell surface WT CXCR4 within 5 min, and maximal internalization of 50% was achieved at 60 min. On the contrary, CXCL12 binding to DT did not trigger endocytosis at all. Interestingly, FL cells downmodulated surface WT CXCR4 in response to CXCL12 with slower kinetics but stronger magnitude compared to that in pro-B cells. Similar to that seen in mutant pro-B cells, there was no obvious internalization of DT in FL cells 30 min after CXCL12 stimulation. However, after 2 h of CXCL12 incubation ,50% of DT had been internalized in FL cells compared to 80% for WT (Fig. 5D). Heterozygous DT mutant cells displayed an intermediate level of receptor internalization (Fig. 5E), a phenotype previously seen in cells from some patients with WHIM syndrome [23]. Taken together, our data indicate that the truncation of the CXCR4 tail severely compromises CXCL12-induced receptor internalization in primary cells. The disparity in results observed between FL and pro-B cells could be explained by FL cells expressing CXCR7 that can form heterodimers with CXCR4 [37]. Therefore, internalization of ligand-bound CXCR7 could result in partial downmodulation of heterodimerized DT, a process that would not occur in pro-B cells which do not express CXCR7.
The CT tail is critical for CXCR4 G-protein-dependent and -independent signaling It remains to be determined whether signals initiated from G proteins and the cytoplasmic tail-associated molecules may act independently or must cross-talk to coordinately regulate proper cellular functions. Therefore, we decided to use our DT mutant cells to dissect the role of the CXCR4 tail in the CXCR4 signaling cascade.
To obtain sufficient primary DT cells for biochemical studies, we generated ROSA ET2-CRE CXCR4-DT/F mice in which the floxed Cxcr4 gene (Cxcr4 F ) can be deleted by tamoxifen induced Cre, allowing only DT expression in adult tissues (referred to as conditional DT mice). Treatment of the conditional DT mice with tamoxifen led to nearly complete deletion of the Cxcr4 F allele in splenocytes, and the overall lymphoid and myeloid compartments of the spleen in the mutant and WT mice were comparable (data not shown). To determine whether the DT mutation impaired CXCR4 signaling, we first examined activation of ERK and PKB, which have been shown to be downstream of both barrestin and G protein signaling pathways [38]. We found that both ERK and PKB were rapidly and transiently phosphorylated in WT cells after CXCL12 stimulation (Fig. 6A). This signaling event was dependent on Gai protein activation, as PTX treatment blocked their phosphorylation completely (Fig. 6B). To our surprise, stimulation of DT splenocytes with CXCL12 failed to induce measurable levels of phosphorylation of either ERK or PKB, suggesting that the lack of the CXCR4 tail impairs G-protein activation (Fig. 6A). Similar results were also obtained using FL cells and Abelson transformed pro-B cells (Fig. S5), indicating that the observed defect in the activation of these signaling pathways is a ubiquitous rather than a cell-lineage specific phenomenon.
To further evaluate whether the tail contributed to CXCR4 mediated G-protein signaling, we examined CXCL12 induced intracellular calcium mobilization, a key signaling event known to be activated only by Gbc [39]. Surprisingly, we found that CXCL12 stimulation was not able to elicit calcium flux in DT cells (Fig. 6C). These results suggest that the CXCR4 tail may play an essential role in G-protein activation.

CXCR4 mediated Gai protein activation is dependent on the CXCR4 tail
Mutational studies showed that a cluster of amino acid residues located at the N-terminal of the cytoplasmic tail of GPCRs was critical for G-protein activation [40]. However, it is not clear whether the CT tail was actually involved in G-protein coupling. Since our data showed that DT failed to activate multiple signaling pathways downstream of G proteins, we sought to determine whether the DT mutation altered the physical interaction between CXCR4 and heterotrimeric G proteins. We found that anti-CXCR4 immunoprecipitated an equivalent amount of Gai in WT and DT cells (Fig. 7A), indicating that heterotrimeric G-proteins are still physically associated with DT.
Although the CXCR4 tail was not required for G-protein association, it may control G-protein activity. To assess this  possibility, we examined Gai activity by measuring intracellular cAMP levels after CXCL12 stimulation. In WT cells, CXCL12activated Gai proteins inhibited forskolin-induced adenylyl cyclase activity, leading to the suppression in intracellular cAMP production. However, the forskolin elevated intracellular cAMP level was not attenuated in DT cells after CXCL12 stimulation (Fig. 7B). These data demonstrate that the CXCR4 tail, although dispensable for G-protein association, is required for heterotrimeric G-protein activation.

Discussion
Models regarding GPCR activation and signaling have been rapidly modified to accommodate new data derived from structural and mutational analysis of GPCRs. In a classical model, GPCRs mediate virtually all signaling events through their coupled G proteins, and the cytoplasmic tail of GPCRs plays only a negative regulatory role through the interacting b-arrestins to downmodulate ligand-activated receptors and desensitize cells from further ligand stimulation [41]. This view has been changed after the finding of ''ligand bias'', namely, a GPCR engaged to different ligands can activate G proteins without inducing receptor internalization or can activate downstream MAPK pathways through b-arrestins in the absence of detectable G protein activities [42][43][44]. In a revised model, different agonists may induce the same GPCR to assume different active states, resulting in exposure of distinct cytoplasmic interfaces to different interacting molecules (the G proteins or b-arrestins) that contribute to various cellular outcomes [45].
The physiological significance of these signaling mechanisms has not been delineated in vivo. Here, we used CXCR4 as a model molecule to elucidate the cellular effects mediated by individual signaling modules of GPCRs. For this reason, we generated DT mice. Our results demonstrate that the CXCR4 tail plays an essential function of CXCR4 in the developing brain, vascular and hematopoietic systems. However, the embryonic lethality of CXCR4-DT mice was unexpected, as patients with WHIM syndrome who also harbor truncating mutations at the CXCR4 C-terminus exhibited a dominant gain of function phenotype. We speculate that C-terminus truncated CXCR4 in WHIM patients might lose desensitization function but retain partial signaling properties through the remaining 8 to 12 serine residues in its tail. This postulation is supported by recent findings indicating that site-specific phosphorylation of serine residues at the CXCR4 tail is regulated by different GRKs which have either positive or negative impact on CXCR4 signaling [33]. Further studies are required to elucidate whether knock-in mice expressing the same mutant form of CXCR4 as WHIM patients recapitulate the human pathogenesis.
A question that remains unanswered is how the CXCR4 tail facilitates G-protein activation. G proteins are presumably activated through a cascade of conformational changes upon ligand binding to GPCRs. Due to the high flexibility of the Cterminus of GPCRs, this portion has been deleted to promote crystallization of GPCRs for structural studies [46,47]. Therefore, no information can be inferred from the available structures on the interdependence between the C-terminus of GPCRs and GPCRassociated G proteins. Based on biochemical studies, the tail of GPCRs is known to interact with GRKs and b-arrestins. Although it is unclear whether such an interaction facilitates G-protein activation, experiments from b-arrestin2 and GRK6 knock-out mice seem to suggest otherwise [48]. It is conceivable that the tail itself or other unknown associated molecules may help to facilitate the conformational change and activation of GPCR interacting heterotrimeric G proteins upon ligand stimulation [49]. Further mutagenesis and proteomics studies may help to resolve this issue.
It should be mentioned that DT embryos exhibited a relatively milder developmental retardation as compared to the CXCR4null mice. Consistently, the DT mutation completely abolishes the chemotaxis function but still retained the adhesion ability of WT pro-B cells to VCAM-1 upon CXCL12 stimulation. These results thus reveal the existence of a CT tail and G-protein independent pathway responsible for CXCR4-induced inside-out signaling for integrin activation. In agreement with this notion, it has been  shown that activation of CXCR7, results in integrin-mediated adhesion without evoking obvious G-protein signaling [50]. In the case of CXCR4, we and others have found that PTX could completely block CXCL12 induced chemotaxis but not adhesion [51]. At present, the nature of this tail-and G protein-independent pathway remains unclear.
In summary, we generated DT mice to directly determine the signaling properties of the CXCR4 tail and to assess the physiological relevance of this signaling module. Results from our studies have revealed an unexpected regulatory role of the tail in G-protein activation. This discovery is important not only for understanding the mechanisms of GPCR activation and function but is also pertinent for future drug development.

Ethics Statement
All experiments were reviewed and approved by the Institutional Animal Care and Use Committee at the Feinstein Institute for Medical Research (IACUC Protocol #2009-023). allowing it to enter the nucleus and activate the expression of the tTA-dependent luciferase reporter gene (4). (B) HEK cells were transiently transfected with the luciferase reporter vector together with a combination of b-arrestin-TEV, WTCXCR4-tTA, or DTCXCR4-tTA as indicated. Cells were stimulated with CXCL12 and measured for luciferase activity. Results from 3 independent experiments are displayed as the luminescence index calculated by the fold increases in luciferase activity in cells 6 CXCL12 (*, P,0.01, calculated using the two-tailed Students t test). (C) The kinetics of receptor downmodulation was determined by measuring cell-surface expression of WT CXCR4 or DT on pro-B cells after CXCL12 stimulation for indicated times. Results are from 3 independent experiments (*, P,0.05, calculated using the two-tailed Student's t test). (D) CXCL12-induced receptor internalization on E18.5 FL cells was carried out as described in (C) above. (E) Representative histograms show CXCR4 expression on FL cells of WT, heterozygous and homozygous DT mice with (thick line) or without (shaded) 2 h CXCL12 incubation. doi:10.1371/journal.pone.0015397.g005 Mice Strategies for gene targeting, ES cell screening and genotyping of DT mice are provided in Fig. S1. To generate mice that express DT in adult tissues, DT mice were bred to previously described ROSA ET2-CRE Cxcr4 +/F mice [52] to obtain conditional DT mice (ROSA ET2-CRE CXCR4-DT/F) and littermate controls (CXCR4-DT/F). To express DT specifically in B-lineage cells, DT mice were bred to CD19-Cre Cxcr4 +/F mice to generate CD19-Cre+ CXCR4-DT/F mice. Previously reported CXCR4-null mice were used as negative controls.

Abelson virus transformation of pro-B cells
FL cells were isolated from E15.5 WT and DT embryos and incubated with Abelson virus suspensions (gift from Dr. Hua Gu, Columbia University) for 2 h, 37uC. Cells were then cultured in OptiMEM + 10% FBS. 7 d later, cells underwent a second round of Abelson virus infection. CXCR4-null pro-B cell lines were generated by infecting splenocytes of CXCR4 f/f CD19-Cre mice with Abelson virus. Primary transformants were analyzed by flow cytometry and confirmed to be a homogenous population of B220 + CD43 + pro-B cells.

Histology
Histology analysis was performed as described [4]. Briefly, brains of E18.5 embryos were dissected, fixed in Bouin's solution and embedded in paraffin. 6 mm brain sections were stained with hematoxylin and eosin and analyzed under a Nikon Eclipse TE2000-E microscope equipped with a Nikon DS-Fi1 color camera.

Flow cytometry and receptor internalization
Cells from the PB, spleen or FL were stained using following antibodies: anti-B220 APC, anti-CD43 FITC, anti-CD19 PE, anti-Gr-1 PE, anti-CD11b FITC and anti-IgM FITC (all from eBioscience). Data were attained on an LSR II (Becton Dickinson) and analyzed using FlowJo software (Tree Star).
For receptor internalization analysis, cells were resuspended at 4610 6 cells/ml in DMEM and stimulated with 0.5 mg/ml CXCL12 at 37uC. Reactions were terminated by addition of 4% PFA. Cells were stained for surface CXCR4 with anti-CXCR4 biotin (BD) and streptavidin-conjugated PE-Cy7 (eBioscience).

Chemotaxis and adhesion assays
Chemotaxis was performed in transwells (5 mM pore size, Costar). The upper chambers containing 2610 5 cells were placed into lower wells with 0.5 mg/ml CXCL12 (Peprotech). After 3 h incubation at 37uC, migrated cells were collected and enumerated by flow cytometry.
Calcium mobilization 5610 5 FL cells from E18.5 embryos in 0.5 ml of assay buffer (25 mM HEPES, pH 7.4, 0.1% BSA, 2.5 mM probenecid in Hank's buffer) were loaded for 30 min at 37uC with 2 mM Fluo-4 and 0.84 mM Fura-red (Molecular Probes). Cells were then washed and resuspended in pre-warmed assay buffer. Samples were first analyzed by flow cytometry for 1 min to establish a baseline before stimulation with 0.5 mg/ml CXCL12 for 2 min and then 2 mM ionomycin (Sigma) for another 2 min.

Immunoprecipitation and immunoblotting
To detect interaction between CXCR4 and G-proteins, WT, CXCR4-null and DT pro-B cells were serum-starved for 2 h at 37uC and then cross-linked with 1 mM DSP (dithio-bis-succinimidyl propionate, Pierce) for 30 min, RT. Then cells were lysed by addition of lysis buffer (0.1 M NaCl, 50 mM Tris-HCl, pH 7.4, 15 mM EGTA, 1% SDS, 1% Triton X-100, 1% Cymal-5, 50 mg/ ml DNase I, and protease inhibitors (Roche)). Lysates were centrifuged and cleared samples incubated with 10 mg/ml anti-CXCR4 (Abcam) overnight, 4uC, and then with 30 ml protein G agarose for an additional 3 h. Beads were washed, eluted in 2x SDS sample buffer, resolved by 4-12% gradient SDS-PAGE, and analyzed by immunoblotting with anti-Gai (Santa Cruz Biotechnology). For detection of pERK and pPKB, splenocytes from conditional DT mice and littermate controls were stimulated with 0.5 mg/ml CXCL12 for indicated times and immunoblots prepared from lysates. Western blot analysis was carried out with antibodies against pPKB (S473, Cell Signaling), ERK and pERK (Thr202, Tyr204) (Santa Cruz Biotechnology).

Radioligand binding assay
HEK cells stably transfected with WT CXCR4 or DT were used in a standard competitive radioligand binding assay. Briefly, cells were incubated for 2 hr at 4uC with 187 pM 125 I-CXCL12 (PerkinElmer) in the presence of increasing concentrations of cold CXCL12 (0-1000 nM). After washing 3 times in wash buffer (0.5 M NaCl, 1% BSA, 5 mM MgCl 2 , 1 mM CaCl 2 , 25 mM HEPES, pH 7.1), bound radioactivity was measured using a b counter (Beckman LS 6500SC). Data were averaged from three independent experiments, and calculated as: % of binding = 100 x {(cpm non-comp. -cpm sample )/cpm non-comp. }

Intracellular cAMP assay
Splenocytes isolated from conditional DT mice and controls were starved in serum-free medium for 1 h at 37uC. After incubation with 1 mM 3-isobutyl-1-methylxanthine (IBMX) at 37uC for 10 min, cells were stimulated with 1 mg/ml CXCL12 for 10 min, followed by 10 min incubation in the presence of 1 mM forskolin. Intracellular cAMP levels were determined using a competitive enzyme immunoassay (Amersham) following the manufacturer's instructions.

Tango assay
Plasmids b-arrestin-TEV and VTT (TEV cleavage site-tTA fusion construct) were kindly provided by the Richard Axel lab (Columbia University). Luciferase reporter vector (pBI-GI) was from Invitrogen. WT CXCR4 and DT were obtained by PCR using primers outlined in Text S1, and cloned into the VTT plasmid to generate WT or DT fused to the TEV cleavage site ENLYFQL followed by tTA (plasmids WTCXCR4-tTA or DTCXCR4-tTA).
HEK cells were plated into a 96-well flat-bottomed plate at 50,000 cells/well and transfected with b-arrestin-TEV, WTCXCR4-tTA or DTCXCR4-tTA, and pBI-GI at a ratio of 1:5:1 and incubated overnight at 37uC. Supernatant was then replaced with fresh DMEM 61 mg/ml CXCL12 and incubated for 6 hr, 37uC. Media was aspirated and BrightGlo luciferase substrate (Promega) added to each well and activity measured by a Berthold TriStar LB941 plate reader. Figure S1 Generation of Mutant Mice. (A) Structure of the wildtype Cxcr4 locus, map of the targeting construct, the targeted Cxcr4 allele before and after Cre/loxP-mediated neo gene deletion. The tail-truncation mutation was introduced into the second exon by PCR (Exon II). (B) Southern blot analysis of tail DNA from wildtype, heterozygous mice carrying both the mutated exon II and the neo gene (+/2), and heterozygous mice carrying mutant Cxcr4 allele after neo gene deletion (+/mu). The allele detected by probe a: wildtype or mutant Cxcr4 allele without the neo gene: 3.8 kb; the mutant Cxcr4 allele with the neo gene: 5.0 kb. The allele detected by probe b: wildtype: 7.2 kb; the mutant Cxcr4 allele with or without the neo gene: 4.9 kb. The following PCR primers are used for genotyping, which produces a 405-bp wildtype band and a 303-bp tail-truncated Cxcr4 band: 005-D59: 59-CTTCTT-CCACTGTTGCCTGAACC-39 005-D3-end: 59-GCCA-CAGGTCCCTGCCTAGAC-39 (TIF) Figure S2 B cell precursors in the spleen and peripheral blood. Splenocytes and peripheral blood cells were isolated from 6-week-old wildtype (DT/F), CD19-Cre+ CXCR4 F/DT (DT/-) and CD19-Cre+ CXCR4 F/F (2/2) mice and stained with antibodies against CD19, B220 and IgM. CD19 + -gated cells were plotted with IgM and B220. Numbers (mean 6 SD) represent percentages of B220 lo IgM 2 B-cell precursors within the CD19 + Bcell population (n = 5-8). (TIF) Figure S3 Expression of CXCR4, CXCR7 and CD29 in pro-B cells. (A) Cell surface expression of VLA-4 (a4b1) on Abelson-transformed pro-B cells derived from the wildtype (shaded) and CXCR4-DT mice (red line) were detected with anti-CD29 (integrin b1) and analyzed by flow cytometry. (B) Cell surface expression of CXCR4 on Abelson-transformed pro-B cells derived from wildtype, CXCR4-null and CXCR4-DT mice. (C) Expression of CXCR7 was determined by qRT-PCR. RNA was purified from fetal liver cells of E18.5 wildtype (WT) or CXCR7 2/2 embryos, bone marrow cells of wildtype 8-week-old mice, or Abelson-transformed WT pro B cells using an RNeasy kit (Qiagen). cDNA was then prepared using SuperScriptIII First Strand Synthesis kit (Invitrogen). qRT-PCR was performed using primers specific for CXCR7. Each value was normalized to b-actin expression levels. (TIF) Figure S4 Pro-B cell adhesion to VCAM-1 is insensitive to pertussis toxin treatment. Pro-B cells were treated with 100 ng/ml of pertussis toxin (PTX) for 3 h at 37uC. PTX-treated and untreated cells were respectively seeded in 96-well plates coated with 1 mg/ml of mouse VCAM-1-Fc with or without 0.5 mg/ml of CXCL12, and incubated at 37uC for an additional 2 min. Non-adherent cells were then removed by washing, adherent cells were collected and counted. Bar graphs represent the percentages of pro-B cells attached to VCAM-1 in the presence or absence of PTX. Wildtype and CXCR4-DT cells are indicated by open and gray bars, respectively. (TIF) Figure S5 Impaired signaling in CXCR4 CXCR4-DT pro-B cells. Abelson-transformed pro-B cells were stimulated with 0.5 mg/ml CXCL12 for indicated times. Cell lysates were prepared and Western blot analysis for phosphorylated PKB at serine 473 (pPKB) or phosphorylated ERK (pERK) were performed. Equal loading of protein was confirmed by re-blotting with anti-ERK antibody. Results are representative of at least 3 independent experiments. (TIF) Text S1 Primers. (DOC)