MIP-1α/CCL3-mediated maintenance of leukemia-initiating cells in the initiation process of chronic myeloid leukemia

Purified KLS⁺ cells were infected with the retrovirus carrying MSCV-BCR-ABL-ires-GFP as previously described (Naka et al., 2010). At 4 d after the infection, ∼1% of c-kit⁺ cells expressed GFP reporter protein (Fig. 1 a), and the positive cell proportion did not increase by 7 d after the infection (not depicted). Because we injected ∼4 × 10⁴ total KLS⁺ cells into the each BM cavity, we estimated that 400–500 resultant BCR-ABL-expressing LICs could be injected into each mouse. In the beginning, we injected LICs directly into the BM cavity of nonirradiated WT mice. The white blood cell (WBC) number in peripheral blood (PB) increased until 2–3 wk after the transplantation, decreasing thereafter (Fig. 1 b). CD11b⁺GFP⁺ leukemia cells consistently but transiently appeared in the PB from 2–3 wk, diminishing at 4 wk (Fig. 1 c).

Given the capacity of exogenous proteins such as human BCR-ABL and GFP to induce an immune response in nonirradiated immune competent mice (Pear et al., 1998; Stripecke et al., 1999; Mumprecht et al., 2009), we presumed that BCR-ABL⁺ leukemia cells could not survive T cell–dependent immune response in the WT recipient mice. To overcome this obstacle, we injected LICs directly into the BM cavity of a nonirradiated athymic nude mouse lacking T cell–dependent immune response. When LICs were given into the BM cavity, all nude mice developed a CML-like leukocytosis (Figs. 2 a and 3 a) and GFP⁺ LIC-derived cells, particularly Gr-1⁺CD11b⁺GFP⁺ cells, predominated in the PB (Figs. 2 b and 3 a). Splenomegaly, a pathognomonic sign, was evident in all nude mice receiving LICs (Fig. 2 c), and death ensued within 2 mo in most of the nude mice (Fig. 2 d). Moreover, histopathological examination demonstrated extramedullary hematopoiesis in the SP and liver and abnormal leukocyte infiltration in the lung (Fig. 3 b). LIC injection into the BM cavity caused a progressive replacement of lineage⁻c-kit⁺ BM progenitors with BCR-ABL⁺ cells, as indicated by an increase in GFP-positive cells in the lineage⁻c-kit⁺ population (Fig. 2 e). Furthermore, BCR-ABL⁺KLS⁺ LICs were distributed in the contralateral nontransferred BM as well as the transferred BM (Fig. 2 f). Thus, the present model can recapitulate a CML-like MPN in nonirradiated recipients.

In the early phase of the present model, normal hematopoietic cells were preserved in the BM and PB. Given the accumulating evidence to indicate the effects of chemokines on BM HSPC mobilization (Pelus and Fukuda, 2008), we examined inflammatory chemokine expression in the serum of mice developing CML. Consistent with the previous report that MIP-1α/CCL3 expression is positively regulated by BCR-ABL expression (Zhang et al., 2012), we observed that serum CCL3 transiently increased at 2 wk after BM transplantation of LICs (Fig. 4 a). On the contrary, other chemokines such as CXCL2, CCL2, CCL4, and CCL5 did not change significantly (Fig. 4 a). Flow cytometric analysis demonstrated that CCL3 was mainly expressed by BCR-ABL⁺lineage⁻c-kit⁻ immature leukemia cells and, to a lesser degree, by BCR-ABL⁻lineage⁻c-kit⁻ cells in the BM (Fig. 4 b). Moreover, these populations expressed constitutively high levels of CCL3 during CML development in the BM (not depicted) despite a transient increase in the serum concentration (Fig. 4 a).

When KLS⁺ cells purified from CCL3^−/− mice were transduced with the BCR-ABL gene, the transduction efficiency was ∼1.5% (not depicted), similarly as observed when using WT mouse–derived KLS⁺ cells. However, CCL3^−/−-derived LICs failed to induce CML-like leukocytosis (Fig. 5 a) and splenomegaly (Fig. 5 b) when administered to the nonirradiated nude mouse BM cavity. A small number of CD11b⁺GFP⁺ leukemia cells appeared in the PB at 3 wk after BM transplantation but disappeared by 4 wk (Fig. 5 c). GFP⁺lineage^high mature leukemia cells were generated from CCL3^−/−-derived as well as WT-derived LICs at 2 wk after the transplantation, but this cell population disappeared in BM transplanted with CCL3^−/−- but not WT-derived LICs at 3 wk after the transplantation (Fig. 5 d). Moreover, CCL3^−/−-derived GFP⁺lineage⁻ immature leukemia cells in BM and KLS⁺ LICs in SP and BM markedly decreased compared with WT-derived cells (Fig. 5, d and e). To exclude the possibility of the defect in the leukemogenesis in CCL3^−/−-derived LICs by the gene modification, we extrinsically injected CCL3 into nude mice that were transplanted with CCL3^−/−-derived LICs (illustrated in Fig. 6 a). The administration of CCL3 restored most of the CML phenotypes such as CML-like leukocytosis (Fig. 6 b), splenomegaly (Fig. 6 c), and the maintenance of GFP⁺ LICs in BM (Fig. 6 d). After the termination of CCL3 administration, three out of five mice succumbed to leukemia within 8 wk, whereas all five PBS-treated mice survived (Fig. 6 e). Moreover, CCL3^−/−-derived LICs induced a CML-like MPN in irradiated recipients, accompanied with leukocytosis containing CD11b⁺GFP⁺ cells and splenomegaly, but to a lesser degree than WT LICs did (Fig. 7, a–c). However, the LIC numbers in BM were much less with the use of CCL3^−/−-derived LICs than that of WT LICs (Fig. 7 d). Thus, CCL3 can be crucial for the early phase of CML development, at least in part by maintaining LICs in the BM.

Because CCL3^−/−-derived LICs can only reproduce the CML-like MPN in irradiated WT mice (Fig. 7) but not in nonirradiated nude mice (Fig. 5), we next examined the synergistic effects of CCL3 gene loss on TKI treatment using irradiated WT mice. Daily treatment with a TKI, imatinib, from 8 d after LIC transplantation, reduced CML-like leukocytosis and splenomegaly (Fig. 7, a and c), but LICs survived in the BM until 3 wk, regardless of the loss of CCL3 gene (Fig. 7 d). However, survival after the cessation of imatinib was significantly prolonged in the mice injected with CCL3^−/− mouse–derived LICs compared with those injected with WT mouse–derived LICs (Fig. 7 e). Moreover, immediately after the cessation of imatinib treatment, CCL3^−/− mouse–derived LICs were present in the BM at similar levels as WT mouse–derived LICs, but the numbers were remarkably decreased in the mice surviving until 60 d after the injection (Fig. 7 d). Thus, CCL3 can also crucially contribute to the recurrence of CML after TKI discontinuation by supporting the maintenance of LICs in BM.

Leukocytosis persisted with the appearance of LIC-derived cells in PB until 3 wk after intra-BM injection of LICs into immune competent WT mice (Fig. 1). Moreover, when WT mouse–derived BCR-ABL–transduced KLS⁺ cells were transferred into nonirradiated WT mice, splenomegaly transiently appeared, reaching a maximum at 2 wk after the transplantation (not depicted). Concomitantly, lineage⁻GFP⁺ cells were abundant in the BM (Fig. 8 a) at 2 wk after the transplantation. On the contrary, the transplantation of CCL3^−/−-derived and BCR-ABL–transduced KLS⁺ cells yielded reduced lineage⁻GFP⁺ cell numbers in BM (Fig. 8 a). Because these observations paralleled the results with nude mice, we assume that it is reasonable to compare the graft efficiency of LICs at 2 wk after BM transplantation to nonirradiated BALB/c mice and those deficient in specific receptor genes for CCL3, CCR1, or CCR5. After transplantation, WT-derived GFP⁺lineage⁻ immature cells and GFP⁺KLS⁺ LICs were efficiently maintained in the CCL3^−/− recipient BM as well as in the WT one (Fig. 8, c and d), suggesting that donor cell–derived CCL3 is sufficient for their maintenance. In contrast, when WT-derived LICs were transferred into the BM of nonirradiated CCR1^−/− or CCR5^−/− mice, both strains failed to maintain GFP⁺lineage⁻ immature cells and GFP⁺KLS⁺ LICs (Fig. 8, c and d). Thus, the maintenance of LICs may require CCL3-mediated signals through both CCR1 and CCR5 expressed on BCR-ABL⁻ normal recipient cells. However, the transfer of CCR1^−/− or CCR5^−/− mouse–derived LICs failed to maintain GFP⁺lineage⁻ immature cells and GFP⁺KLS⁺ LICs even in the WT recipients (Fig. 8, c and d), suggesting the requirement of CCR1 and CCR5 expression on both donor and recipient cells for the maintenance of LICs. However, BCR-ABL infection generated GFP⁺lineage⁺ mature cells at a similar proportion among all strains (Fig. 8 b). These observations would indicate that LICs were successfully transplanted but could not be maintained to generate mature leukemia cells in the absence of CCL3-mediated signals.

We observed that CCR1 and CCR5 expressed on both donor and recipient cells were required for the maintenance of LICs (Fig. 8 d). Considering a low efficiency of BCR-ABL transduction (Fig. 1 a), a small number of BCR-ABL⁺KLS⁺ LICs coexist with a relatively larger numbers of donor KLS⁺ cells (illustrated in Fig. 9 a). Thus, we assumed that LICs should survive the competition with both recipient- and donor-derived normal hematopoietic cells by using the CCL3/CCR1 and CCL3/CCR5 axes. To test this hypothesis, WT LICs were mixed with an equal number of nontransduced WT, CCR1^−/−, or CCR5^−/− KLS⁺ cells, and the resultant cell population was injected into the BM cavity of nonirradiated nude mice (illustrated in Fig. 9 b). WT LICs efficiently induced CML-like leukocytosis and splenomegaly in the absence of competitive cells (Fig. 9, c and d). The admixture of nontransduced CCR1^−/−- and CCR5^−/−-derived KLS⁺ cells, but not WT-derived KLS⁺ cells, attenuated leukocytosis (Fig. 9 c) and splenomegaly (Fig. 9 d) and markedly reduced the numbers of LICs in BM (Fig. 9 e). Thus, the interaction of leukemia cell–derived CCL3 with CCR1 and CCR5 expressed on normal hematopoietic cells can confer LICs the advantage to survive in BM in a paracrine manner.

At 3 wk after the transplantation with WT-derived LICs, the numbers of GFP-nonexpressing total cells and lineage⁻c-kit⁺ progenitor cells in BM were markedly reduced, which were attenuated when CCL3^−/−-derived LICs were transplanted (Fig. 10 a). Likewise, the reduction of GFP-nonexpressing BM progenitor cells was attenuated when WT-derived LICs were transplanted into CCR1^−/− and CCR5^−/− mice (not depicted). CCR1 and CCR5 were constitutively expressed on lineage⁻c-kit⁺ progenitor cells as well as lineage⁺ mature leukocytes in nontreated WT mice (Fig. 10 b). Moreover, BCR-ABL⁻ normal BM progenitor cells expressed both CCR1 and CCR5. However, CCR5 but not CCR1 expression was abrogated in BCR-ABL⁺ progenitor cells in the BM of CML mice (Fig. 10, c and d). Thus, CCL3 may interact with CCR5 and CCR1 expressed on normal BM progenitor cells. The capacity of CCL3 to mobilize HSPCs from the BM (Lord et al., 1995; Broxmeyer et al., 1999) prompted us to examine the effects of intra-BM injection of CCL3. This treatment decreased the number of lineage⁻c-kit⁺ progenitors in BM by half in WT mice, but not in CCR1^−/− or CCR5^−/− mice (Fig. 10 e). Moreover, the circulating lineage⁻c-kit⁺ progenitor cells significantly increased in the PB after intra-BM injection of CCL3 (Fig. 10 f). However, intra-BM injection of CCL3 did not change the composition of long-term HSCs, short-term HSCs, and multipotent progenitors among lineage⁻c-kit⁺ BM progenitors (not depicted). Thus, CCL3 can interact with CCR1 and CCR5 expressed on normal HSPCs to accelerate their mobilization from BM to PB.

In the present study, we established a novel mouse model developing CML-like MPN in a nonirradiated host with preserved BM architecture. Previous CML mouse models have difficulties in elucidating the interaction between normal hematopoietic cells and LICs because of irradiation-induced destruction of BM architecture, whereas our model can facilitate the analysis on this point, particularly at the early phase of CML development. By using this model, we demonstrated that an inflammatory chemokine, CCL3, was expressed abundantly at the early phase of CML development and that BCR-ABL-expressing lineage⁻ immature leukemia cells were a major source of CCL3. Moreover, we showed that CCL3-deficient BCR-ABL–expressing LICs failed to populate the BM and could not cause CML-like leukocytosis and splenomegaly when given directly into the BM of nude mice. Furthermore, WT mouse–derived LICs exhibited reduced capacity to survive in the BM of mice deficient in CCR1 or CCR5, specific receptors for CCL3, even when BCR-ABL–expressing LICs were directly injected into the BM of these mice. These observations implicate the crucial involvement of the interaction between CCL3 and its receptors, CCR1 and CCR5, in the initiation of CML development.

Another chemokine, CXCL12, is abundantly present in the BM and has diverse effects on BM HSPCs (Aiuti et al., 1997; Zou et al., 1998). Basu and Broxmeyer (2009) recently demonstrated that CCL3, CCL4, and CCL5 can enhance CXCL12-induced chemotaxis and concomitantly reduce CXCL12-induced adhesion to VCAM-1 of human HSPCs. This may account for the capacity of CCL3 and its genetically engineered variant, BB-10010, to potently and rapidly mobilize HSPCs from the BM to PB (Lord et al., 1995). We also observed that intra-BM injection of CCL3 efficiently induced the mobilization of HSPCs from BM to PB in WT but not CCR1^−/− or CCR5^−/− mice. These observations would indicate that CCL3 can act cooperatively on CCR1 and CCR5 on HSPCs as observed on dendritic precursor cells (Zhang et al., 2004). Moreover, WT-derived but not CCL3^−/−-derived BCR-ABL–expressing LICs markedly reduced the numbers of BCR-ABL⁻lin⁻c-kit⁺ normal HSPCs in BM when injected into the BM of nude mice. Given the failure of BCR-ABL–expressing LICs to populate the BM of CCR1^−/− or CCR5^−/− mice, we postulated that BCR-ABL–expressing leukemia cell–derived CCL3 can displace from the BM normal CCR1- and CCR5-expressing HSPCs, which are present in both donor cell population as well as recipient cell population, thereby boosting the growth of LICs. This notion is supported by our observation that the admixture of CCR1^−/−- and CCR5^−/−-derived HSPCs but not WT-derived HSPCs to BCR-ABL–expressing LICs attenuated the development of CML-like MPN when injected into the BM of nude mice.

In the developing wing imaginal disc of Drosophila melanogaster, normal epithelial cells can eliminate a small number of transformed cells with mutations in oncogenes and/or tumor-suppressor genes within a normal epithelial monolayer (Prober and Edgar, 2000; Hogan et al., 2009). Similar observations were obtained by using in vitro mammalian cell culture system (Hogan et al., 2009). The phenomenon is proposed to be termed as “cell competition,” which can function as a potent intrinsic tumor suppression system. However, evidence is lacking to indicate the presence of cell competition in mammals in vivo. Our present observations would indicate that the CCL3 axis crucially contributes to the maintenance of BCR-ABL–expressing LICs in BM by mobilizing normal HSPCs from the BM and eventually reducing their numbers in BM. Accumulating evidence indicates that the niche construction is required for the maintenance and colonization of both normal BM stem/progenitor cells (Zhang et al., 2003) and LICs (Wertheim et al., 2002; Nair et al., 2010). Thus, our present observations would imply the presence of cell competition between normal HSPCs and LICs in the limited space of the BM cavity under the control of CCL3.

Here, we demonstrated that BCR-ABL–expressing leukemia cell–derived CCL3 can have a crucial role in CML development by mobilizing normal HSPCs from the BM and eventually making space for LICs in the BM. Several lines of evidence indicate that CCL3 has negative impacts on the proliferation of normal HSPCs. On the contrary, BCR-ABL–expressing LICs are rather resistant to CCL3-mediated growth inhibition arising from constitutive Abl tyrosine kinase activation–mediated desensitization (Wark et al., 1998). Consequently, CCL3 may provide BCR-ABL–expressing LICs with growth advantage over normal HSPCs. Thus, LICs may compete with normal HSPCs in the BM, particularly in the early phase of CML development, through these two distinct mechanisms, mobilization of normal HSPCs and inhibition of their growth.

The treatment with TKIs against BCR-ABL exhibits marked therapeutic effects, resulting in frequent remission at cytological and even at the molecular levels and has markedly improved the prognosis of CML patients (Druker et al., 2006; Kantarjian et al., 2010; Saglio et al., 2010). However, the cessation of the treatment often leads to the relapse of leukemia after a certain period of latency (Michor et al., 2005). Evidence is accumulating to indicate that LICs are relatively resistant to TKI therapy compared with mature leukemia cells (Corbin et al., 2011). As a result, LICs can survive in the BM even after extensive TKI treatment, and a residual number of TKI-resistant LICs can proliferate massively after the cessation of the treatment. Several signaling pathways have been proposed to account for LIC survival after TKI treatment (Seke Etet et al., 2012), including Hedgehog (Dierks et al., 2008; Zhao et al., 2009; Long et al., 2011) and TGF-β–FOXO3 pathways (Naka et al., 2010). Given a potent capacity of CCL3 to provide LICs with a growth advantage over normal HSPCs in the initiation process of CML, we also assessed the effects of CCL3 ablation on the TKI treatment. Under our experimental conditions, CCL3 ablation failed to decrease LICs on the TKI treatment. However, the cessation of TKI treatment induced CML to rapidly recur, whereas lack of CCL3 remarkably prolonged the survival. This prolongation can be ascribed to a sustained decrease in LICs in BM transplanted with CCL3^−/−-derived LICs, even after the cessation of imatinib treatment. Thus, CCL3 can also crucially contribute to the maintenance of LICs of CML after TKI discontinuation.

We identified BCR-ABL⁺lineage⁻c-kit⁻ immature leukemia cells as the main CCL3-producing cells in CML mice. Moreover, we observed that after 3-mo treatment with TKI, human CML patients exhibited enhanced CCL3 mRNA expression in BM with a marked reduction in the copy number of BCR-ABL gene (unpublished data). Because immature leukemia cells are more resistant to TKI treatment than differentiated leukemia cells, at its initiation phase (Tang et al., 2011), CCL3-expressing immature leukemia cells might have an advantage in survival in the initiation phase of TKI treatment. Thus, more detailed analysis on CCL3-expressing cells in human CML BM can pave the way to develop a novel type of treatment against CML to supplement present TKI treatment.

Specific pathogen–free 5–7-wk-old male BALB/c and athymic BALB/c-nu mice were purchased from Charles River and designated as WT and nude mice, respectively. CCL3^−/− mice were obtained from the Jackson Laboratory. CCR1^−/− and CCR5^−/− mice were provided by P.M. Murphy (National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD) and K. Matsushima (University of Tokyo, Tokyo, Japan), respectively. These gene-deficient mice were backcrossed with BALB/c for more than eight generations. Animal experiments were approved by the Committee on Animal Experimentation of Kanazawa University. All mice were kept under the specific pathogen–free conditions, and all the animal experiments in this study complied with the Guidelines for the Care and Use of Laboratory Animal of Kanazawa University.

The following rat anti–mouse mAbs were used: anti-CCR1 (643854; R&D Systems), anti-CD4 (RM4-5; eBioscience), anti-CD8 (53-6.7; eBioscience), anti-CD11b (M1/70; eBioscience), anti-CD34 (RAM34; eBioscience), anti-CD45R/B220 (RA3-6B2; eBioscience), anti-CD117/c-kit (ACK2; eBioscience), anti–Ly-6A/E/Sca-1 (D7; eBioscience), anti–Ly-6G/Gr-1 (RB6-8C5; eBioscience), anti–MIP-1α/CCL3 (39624; R&D Systems), and anti–TER-119 (TER-119; eBioscience). Mouse lineage Ab cocktail with isotype control set was purchased from BD. Rabbit anti–human CCR5 polyclonal Ab with a cross reactivity to mouse CCR5 was purchased from Novus Biologicals. Isotype-matched control IgGs for individual rat mAbs and control rabbit IgG were purchased from BD and Dako, respectively.

Total BM cells were flushed out with cold magnetic-activated cell separation (MACS) buffer (PBS supplemented with 2 mM EDTA and 3% FBS) from the femoral and/or tibial bones. Total BM cells were separated by density gradient centrifugation using Histopaque-1083 reagent (Sigma-Aldrich), and then lineage marker (CD4, CD8, CD11b, Gr-1, B220, and TER-119)⁻c-kit⁺Sca-1⁺ cells were sorted by using the FACSAria cell sorter (BD) and were used as KLS⁺ BM cells. Total SP cells were isolated by mechanical digestion from SP. Flow cytometric analysis on BM, SP, and PB was conducted after erythrocytes were depleted by ammonium chloride lysing buffer containing 0.826% NH₄Cl, 0.1% KHCO3, and 0.004% EDTA-4Na.

MSCV-BCR-ABL-ires-GFP plasmid was prepared as described previously (Naka et al., 2010). Retroviral packaging cells (Phoenix 293T) were transiently transfected with the MSCV-BCR-ABL-ires-GFP plasmid using jetPRIME transfection reagent (Polyplus transfection) to produce the retrovirus in the culture supernatant, which was subjected to the infection into KLS⁺ BM cells.

KLS⁺ cells purified from BM of nontreated WT or each gene-deficient mice were cultured in serum-free S-Clone SF-03 medium (Sanko Junyaku) supplemented with 1% BSA, 100 ng/ml stem cell factor (Wako Chemicals USA), 100 ng/ml thrombopoietin, 25 ng/ml fms-like tyrosine kinase-3 ligand (R&D Systems), 10 ng/ml IL-6 (PeproTech), and 10 ng/ml IL-3 (PeproTech) for 1 d. Cultured KLS⁺ cells were infected with the retrovirus carrying MSCV-BCR-ABL-ires-GFP using ViroMag R/L kit (OZ Bioscience) to obtain LICs. The right knee joint was flexed, and a hole was made on the top of tibial bone above the patellar ligament, using a 29-G needle. LICs in a volume of 30 µl were subsequently injected with a 29-G needle–conjugated insulin syringe (Terumo). Hence, the right and left tibia are designated as the transferred and nontransferred BM, respectively. In the imatinib (Novartis) administration experiment, LICs were intravenously injected into the mice, which were irradiated in advance, in an essentially same way as described previously (Naka et al., 2010).

Total RNAs were extracted from cells using RNeasy Mini kit (QIAGEN) and then reverse transcribed using the SuperScript III First-Strand synthesis system (Invitrogen). Quantitative real-time PCR was performed on an Applied Biosystems StepOne real-time PCR system using SYBR Select Master Mix (Applied Biosystems) and the specific primer sets for GAPDH gene (sense, 5′-GCGGCACGTCAGATCCA-3′; and antisense, 5′-CATGGCCTTCCGTGTTTCCTA-3′), CCR1 gene (sense, 5′-TTTGTGGGTGAACGGTTCTG-3′; and antisense, 5′-TGGTATAGCCACATGCCTTTGA-3′), and CCR5 gene (sense, 5′-CATCCGTTCCCCCTACAAGA-3′; and antisense, 5′-GGAACTGACCCTTGAAAATCCA-3′). Thermal cycling was initiated with an initial activation step for 2 min at 50°C and 20 s at 95°C, followed by 40 cycles of 95°C for 3 s and 60°C for 30 s. Immediately after the amplification, melt curve protocols were performed to ensure that primer-dimers and other nonspecific products had been minimized. Relative expression of target gene was analyzed by the ΔΔCt method, using the Ct value of GAPDH gene.

Isolated leukocytes were stained with various combinations of fluorescent dye–conjugated Abs in MACS buffer. For the intracellular MIP-1α/CCL3 staining, leukocytes were incubated in serum-free S-Clone SF-03 medium supplemented with 0.1% GolgiStop reagent (BD) for 4 h. Subsequently, intracellular CCL3 was stained with PE-conjugated anti-CCL3 Ab and the help of Intracellular Cytokine Staining Starter kit (BD). Expression of each molecule was determined using FACSCanto II (BD) and analyzed with FlowJo software (Tree Star).

Tissue samples were fixed in Ufix reagent (Sakura) and embedded in paraffin blocks. Each 4-µm section was stained with hematoxylin and eosin (H&E). Whole blood samples were smeared on to glass slides and fixed with 100% methanol, which were subjected to the Giemsa stain or fluorescent immunostaining. Histopathological images and fluorescent images were acquired using a fluorescence microscope (BX50; Olympus). DP Controller software (Olympus) was used for image processing.

Chemokine expression in the serum was determined on a Bioplex reader (Bio-Rad Laboratories) using a Mouse Cytokine Standard 23-plex assay kit (Bio-Rad Laboratories). All procedures were conducted according to the manufacturer’s instructions.

Data were analyzed statistically using methods indicated in each figure legend. P < 0.05 was considered statistically significant.

We would like to express our gratitude to Dr. Joost J. Oppenheim (National Cancer Institute–Frederick, Frederick, MD) for his critical review of the manuscript.

This work was supported in part by Japan Society for the Promotion of Science KAKENHI grant number 25460492.

The authors have no conflicting financial interests.