Chemotherapy-induced antitumor immunity requires formyl peptide receptor 1

See allHide authors and affiliations

Science  20 Nov 2015:
Vol. 350, Issue 6263, pp. 972-978
DOI: 10.1126/science.aad0779

How dying tumor cells get noticed

Besides killing tumor cells directly, some chemotherapies, such as anthracyclines, also activate the immune system to kill tumors. Vacchelli et al. discovered that in mice, anthracycline-induced antitumor immunity requires immune cells to express the protein formyl peptide receptor 1 (FPR1). Dendritic cells (DCs) near tumors expressed especially high amounts of FPR1. DCs normally capture fragments of dying tumor cells and use them to activate nearby T cells to kill tumors, but DCs lacking FPR1 failed to do this effectively. Individuals with breast or colon cancer expressing a variant of FPR1 and treated with anthracyclines showed poor metastasis-free and overall survival. Thus, FPR1 may affect anti-tumor immunity in people, too.

Science, this issue p. 972


Antitumor immunity driven by intratumoral dendritic cells contributes to the efficacy of anthracycline-based chemotherapy in cancer. We identified a loss-of-function allele of the gene coding for formyl peptide receptor 1 (FPR1) that was associated with poor metastasis-free and overall survival in breast and colorectal cancer patients receiving adjuvant chemotherapy. The therapeutic effects of anthracyclines were abrogated in tumor-bearing Fpr1−/− mice due to impaired antitumor immunity. Fpr1-deficient dendritic cells failed to approach dying cancer cells and, as a result, could not elicit antitumor T cell immunity. Experiments performed in a microfluidic device confirmed that FPR1 and its ligand, annexin-1, promoted stable interactions between dying cancer cells and human or murine leukocytes. Altogether, these results highlight the importance of FPR1 in chemotherapy-induced anticancer immune responses.

The success of anticancer chemotherapy is linked to a durable tumor-targeting immune response (1). Accordingly, the presence of tumor-infiltrating dendritic cells (DCs) and CD8+ T lymphocytes at diagnosis increases the likelihood of breast cancer patients responding to anthracyclines (26). One mechanism through which anthracyclines can stimulate an antitumor immunity is by inducing immunogenic cell death (ICD), and this mechanism implies signaling through innate immune receptors including Toll-like receptors 3 (TLR3) and 4 (TLR4) (710).

We designed a screen for identifying candidate genetic defects that negatively affect chemotherapeutic responses. To this aim, we identified 328 nonsynonymous single-nucleotide polymorphisms (SNPs) affecting proteins involved in the recognition of dying cells by innate immune effectors (11) or influencing the incidence or prognosis of major human cancers and occurring in Caucasians with an allelic frequency >5% (table S1). DNA samples from breast cancer patients treated with adjuvant anthracycline-based chemotherapy (belonging to two independent cohorts) (table S2 and S3) were genotyped, and the effect of each SNP on overall survival (OS) was determined. When a dominant genetic model was applied, only one SNP was significantly (raw P < 0.05) associated with OS independent of major clinicopathological variables in both patient cohorts: rs867228 (fig. S1A, C to H). This SNP (1037A>C) affects exon 2 of the gene coding for formyl peptide receptor 1 (FPR1), which promotes the chemotactic interaction of neutrophils with necrotic hepatocytes (12, 13). rs867228 provokes an amino acid substitution (Glu346Ala) that suppresses FPR1 signaling (14, 15). Other SNPs with less drastic effects on FPR1 function (15, 16) had no effect on OS (fig. S1B and fig. S2, A to C). As opposed to patients bearing the most frequent FPR1 allele (17) in homozygosis (FPR1CC), women bearing the rs867228 loss-of-function allele of FPR1 in homozygosis (FPR1AA) or heterozygosis (FPR1CA) exhibited reduced OS and metastasis-free survival (Fig. 1, A and B; fig. S2, D and E; and tables S4 and S5). Similarly, colorectal cancer patients treated with oxaliplatin-based chemotherapy (table S6) and bearing the rs867228 loss-of-function allele of FPR1 in homozygosis (FPR1AA) exhibited reduced OS and progression-free survival (fig. S3, A to D, and tables S7 and S8). rs867228 had no effect on the OS of breast cancer patients bearing loss-of-function alleles of TLR4 (rs4986790, Asp299Gly) (7, 18) (Fig. 1C and table S9) or TLR3 (rs3775291, Leu412Phe) (Fig. 1E and table S10). The negative effect of the FPR1 polymorphism on OS was only evident in patients with normal TLR3 and TLR4 (Fig. 1, D and F, and tables S9 and S10), suggesting that FPR1 participates in the same therapeutically relevant pathway as the one influenced by the two TLRs.

Fig. 1 Effects of FPR1 deficiency on the survival of breast cancer patients.

(A and B) Kaplan-Meier of the OS (A) or metastasis-free survival (MFS) (B) estimated in the combined cohort of patients (n = 731) treated with adjuvant anthracycline-based chemotherapy and bearing FPR1 rs867228 with CC (wild type) or CA (heterozygous) + AA (mutated homozygous) genotypes. The analysis of the effect of FPR1 rs867228 on OS yields a false discovery rate of 0.1090. (C and D) OS estimated in breast cancer patients bearing mutated (Asp299Gly) (C) or wild-type (Asp299Asp) (D) Toll-like receptor 4 (TLR4) (SNP rs4986790) genotypes according to the illustrated FPR1 SNP rs867228 genotype. (E and F) Kaplan-Meier of the OS estimated in breast cancer patients bearing mutated (Leu412Phe) (E) or wild-type (Leu412Leu) (F) TLR3 genotypes (SNP rs3775291) according to the FPR1 SNP rs867228 genotype. Statistical significance was determined by likelihood ratio test (LRT). HR, hazard ratio.

FPR1 is expressed by myeloid cells (;, as well as by some carcinomas (19). Genetic inactivation of Fpr1 (fig. S4A) did not affect the capacity of MCA205 fibrosarcoma cells treated with mitoxantrone or doxorubicin in vitro to vaccinate syngeneic, immunocompetent mice against a rechallenge with living cells of the same type (Fig. 2, A and B, and fig. S4A). Similar results were obtained with wild-type (WT) and Fpr1−/− TC-1 lung carcinoma cells (fig. S4A; fig. S6, A and B; and fig. S7, A and B). However, Fpr1−/− mice failed to mount an immune response against anthracycline-treated MCA205 (Fig. 2, A and B, and fig. S5A) or TC-1 cells (fig. S6, A and B, and fig. S7, A and B) under conditions in which WT mice did so, suggesting that Fpr1 must be expressed by host cells, not cancer cells, for anthracyclines to elicit antitumor immunity. Accordingly, Fpr1−/− cancer cells responded to anthracycline-based chemotherapy in vivo as WT cancer cells did, whereas WT cancer cells growing in Fpr1−/− hosts were resistant to anthracyclines (Fig. 2, C and D, and figs. S5B, S6C, S6D, S7C, and S7D).

Fig. 2 Contribution of Fpr1 and Annexin A1 to chemotherapy-induced tumor growth reduction.

(A and B) WT or Fpr1−/− MCA205 fibrosarcoma cells cultured overnight with mitoxantrone (MTX) or doxorubicin (DOXO) were inoculated subcutaneously into the flanks of WT C57BL/6 or Fpr1−/− mice. Seven days later, animals were rechallenged into the opposite flank with live WT MCA205 cells. Representative experiment (A) and quantitative analysis of the data [from (A) and fig. S5A] in (B). (C and D) WT and Fpr1−/− C57BL/6 mice bearing WT or Fpr1−/− MCA205 fibrosarcomas were treated with MTX, DOXO, or phosphate-buffered saline (PBS) (as a vehicle control). Representative experiment in (C) and quantitative analysis of the data [from (C) and fig. S5B]. (E and F) WT, Mtfmt−/−, Ctsg−/−, F19a4−/−, or Anxa1−/− MCA205 tumors were established in WT mice and treated with MTX, DOXO, or PBS (as a vehicle control). Results are representative for one experiment out of three involving at least five mice per group [(A), (C), and (E)] or are pooled from at least three independent experiments [(B), (D), and (F)]. Statistical significance was calculated by means of the likelihood ratio test (A), the chi-square test (B), the Wald test [(C) and (E)], and the contrast t test [(D) and (F)] when comparing to PBS controls. ns, not significant; ***P < 0.001. Arrows in (C) and (E) indicate the initiation of chemotherapy.

Next, we investigated how FPR1 may influence the immune response against dying cancer cells by making bone marrow chimeras. WT mice receiving Fpr1−/− bone marrow could not mount a protective immune response after vaccination with anthracycline-treated MCA205 cells (fig. S8A). Moreover, MCA205 cell–derived tumors became refractory to chemotherapy when they grew in WT mice that had received Fpr1−/− bone marrow (fig. S8B). Conversely, tumors growing in Fpr1−/− hosts reconstituted with WT bone marrow responded normally to anthracycline-based chemotherapy (fig. S8C). Thus, expression of FPR1 by the host immune system contributes to the antineoplastic effects of chemotherapy.

Four ligands for FPR1 have been described, namely: (i) cathepsin G (CTSG) (20); (ii) family with sequence similarity 19 [chemokine (C-C motif)–like], member A4 (FAM19A4) (21); (iii) N-formylated peptides contained in bacteria and mitochondria (22); and (iv) annexin A1 (ANXA1), a ubiquitously expressed cytosolic protein (23). Knocking out the genes coding for Ctsg (fig. S4D), Fam19a4 (fig. S4E), or mitochondrial methionyl-tRNA formyltransferase (Mtfmt), an enzyme that catalyzes formylation reactions in mitochondria (24) (fig. S4B), had no effect on the capacity of MCA205 cells to elicit anticancer immune responses upon exposure to anthracyclines in vitro and to respond to mitoxantrone and doxorubicin in vivo (Fig. 2, E and F; fig. S5, C and D; fig. S6, E to H; fig. S7, E to H; fig. S8, D and E; and fig. S9). Conversely, anthracycline-treated Anxa1−/− MCA205 and TC-1 cells (fig. S4C) were unable to drive protective, tumor-targeting immune responses and formed tumors that were resistant to doxorubicin and mitoxantrone (Fig. 2, E and F; fig. S5, C and D; fig. S6, E to H; fig. S7, E to H; fig. S8, D and E). Both anthracyclines stimulated the secretion of Anxa1 by cancer cells (fig. S4F). Moreover, blockade of extracellular ANXA1 with a neutralizing antibody abrogated tumor growth reduction by anthracyclines in vivo (fig. S10), supporting the idea that ANXA1 is the functionally relevant FPR1 ligand.

Injection of the FPR1-specific antagonist cyclosporin H (CsH) (25) either concomitant with chemotherapy or 2 (but not 7) days after chemotherapy, abolished the anticancer effects of anthracyclines (fig. S11). Thus, FPR1 is likely to participate in the initial phase of the anticancer immune response. Microarray analyses of tumors collected from WT versus Fpr1−/− mice 2 days after chemotherapy led to the identification of multiple genes that were induced by doxorubicin in WT (but not in Fpr1−/−) hosts (Fig. 3, A to C, fig. S12, and table S11). These genes included several that are relevant for type I interferon responses, DC maturation and antigen processing/presentation, and cytotoxic T cell effector functions. Fpr1 was expressed by intratumoral CD45+ leukocytes (not by CD45 cells)—in particular, myeloid Ly6ChighLy6G and Ly6G+ cells after chemotherapy in vivo (Fig. 3B and fig. S13, A to D). Moreover, Fpr1 expression levels increased in bone marrow–derived DC confronted with dying (but not living) cancer cells in vitro (fig. S13, E to G). Ly6ChighLy6G cells do (whereas Ly6G+ cells do not) contribute to the efficacy of anthracycline-based chemotherapy (26). Our results therefore suggest the importance of FPR1 expression in the Ly6ChighLy6G myeloid cell population, which comprises DCs and their precursors (26). Elevated expression of the human orthologs of all immune-relevant genes influenced by Fpr1 in mice (Fig. 3A) had a significant (P < 0.05) positive effect on the response of breast cancer patients to anthracyclines in four out of five cohorts studied (Fig. 3D), suggesting that these genes are clinically relevant. Moreover, the mutational status of FPR1 influenced the expression level of several genes involved in antigen presentation within neoplastic lesions of breast cancer patients (fig. S14).

Fig. 3 Effect of FPR1 on the anticancer immune response and the intratumoral positioning of dendritic cells after chemotherapy.

(A) WT or Fpr1−/− mice bearing murine MCA205 fibrosarcomas were treated with doxorubicin (DOXO) or phosphate-buffered saline (PBS) (as a vehicle control), and tumors were harvested 48 hours later for microarray analysis. Immune-related genes up-regulated by DOXO administration in tumors from WT mice (but not from Fpr1−/− mice) are listed. (B) Alternatively, tumors were recovered and tumor-infiltrating immune cells were sorted by fluorescence-activated cell sorting. Fpr1 expression was assessed by quantitative real-time polymerase chain reaction. AU, arbitrary unit; **P < 0.01; ***P < 0.001 (unpaired t test), as compared with PBS administration. (C) A total of 427 genes, identified as highly expressed specifically in DOXO-treated MCA205 tumors growing in WT mice (but not in Fpr1−/− mice), were subjected to gene ontology analysis. (D) Analysis of the overexpression of the set of 84 immune-related genes identified in (A) on “responsive” as compared to “nonresponsive” tumors from five independent breast cancer patient cohorts by employing gene set analysis (GSA). The combined (All) P value was obtained using Fisher’s method. The vertical dotted line represents significance thresholds (P < 0.05). (E to G) Effect of chemotherapy with DOXO on the immune infiltrate of MCA205 fibrosarcomas. Immunofluorescence stainings were performed on tumors recovered 2 days after chemotherapy. Quantitative analysis (E) of the immunostainings [(F) and (G)] are shown. In (E), the length/width ratio (elongation) was determined in 60 to 100 randomly chosen cells, using the morphometric methods illustrated in (F) and (G). Box plots report the lower and upper quartile plus the median value. ns, not significant; ***P < 0.001, one-way analysis of variance (ANOVA), Bonferroni’s multiple comparison test, as compared with tumors harvested from mice with the same genetic background and receiving PBS; ##P < 0.01, as compared with tumors harvested from WT mice treated with DOXO. [(F) and (G)] Representative fluorescence micrographs of tumor-infiltrating DCs (CD11c+CD86+) in DOXO-treated, dead-cell (Casp3a+) enriched tumor sites are reported. Enlarged windows describe the method of length/width quantification. Scale bar, 10 μm. (H and I) Representative images showing the method employed to measure the intercellular distance between the same DC (CD11c+CD86+) and the closest Casp3a (blue line) or Casp3a+ (yellow line) DOXO-treated tumor cells. Scale bar, 20 μm. (J) Quantitative analysis of the dead-cell proximity index determined for tumor-infiltrating DCs. Data are represented as scatter dot plots. ###P < 0.001 (t test), as compared with tumor cells from MTX-treated WT mice.

The presence or absence of Anxa1 (in tumor cells) or that of Fpr1 (in the host) did not affect the capacity of chemotherapy to elicit autophagy, apoptosis, and necrosis in cancer cells (fig. S15). Moreover, the absence of Fpr1 in the host did not influence the density of tumor infiltration by CD11c+CD86+ DCs, macrophages, or granulocytes after chemotherapy (fig. S16A). Morphometric analyses of the ratio between the maximum length and the perpendicular width of tumor-infiltrating myeloid cells, reflecting cellular asymmetry related to migration (27), revealed a selective defect in the chemotherapy-induced elongation of intratumoral DCs (but not macrophages and granulocytes) in Fpr1−/− hosts (Fig. 3, E to G, and figs. S16B and S17). Two days after chemotherapy, a sizeable fraction of cancer cells underwent apoptosis and hence contained immunofluorescence microscopy–detectable active caspase 3 (Casp3a). In WT mice, CD11c+CD86+ DCs recruited to neoplastic lesions by chemotherapy were closer to the most proximal Casp3a+ (dying) cancer cell than to the nearest Casp3a (live) cancer cell. Indeed, their dead-cell proximity index (DCPI; i.e., the ratio between the distance of CD11c+CD86+ DCs from Casp3a+ cancer cells and the distance of CD11c+CD86+ DCs from Casp3a cancer cells) was consistently > 1. Conversely, the DCPI was close to 1 for CD11c+CD86+ DCs infiltrating tumors treated with anthracyclines in Fpr1−/− hosts (Fig. 3, H to J). In accord with its expression pattern (Fig. 3B), Fpr1 did not influence the DCPI of macrophages, yet affected that of neutrophils (fig. S17D). Thus, FPR1 determines the spatial positioning of DCs with respect to dying cancer cells. Accordingly, Fpr1 blockade with CsH or Fpr1 knockout reduced the capacity of intratumoral CD11b+Ly6ChighLy6G cells (which include inflammatory DCs) (28) to acquire antigens from cancers engineered to express green fluorescent protein (GFP) tethered to the inner leaflet of the plasma membrane, in response to chemotherapy (Fig. 4, A and B, and fig. S18A). Moreover, Fpr1 inhibition interfered with the capacity of tumor-infiltrating CD11b+Ly6ChighLy6G cells to express increased levels of class II major histocompatibility complex (MHC) proteins (I-A/I-E) upon chemotherapy in vivo (Fig. 4, A and B, and fig. S18B). However, Fpr1 knockout did not affect major DC functions in vitro and did not interfere with the presentation of soluble protein antigen by DC (figs. S19 and S20).

Fig. 4 Effect of FPR1 on dendritic cell function.

(A and B) MyrPalm-GFP–expressing MCA205 fibrosarcoma tumors were implanted in WT (A) or Fpr1−/− mice (B) and treated with mitoxantrone (MTX) or PBS as a vehicle control, alone or in combination with cyclosporin H (CsH) (A). Thirty-eight hours later, tumor-infiltrating CD11b+Ly6ChighLy6G myeloid cells were isolated and assessed for tumor antigen uptake (indicated by GFP fluorescence intensity) or MHC class II (I-A/I-E) expression. Representative mean fluorescence intensities (MFIs) from three independent experiments are shown. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001 (one-way ANOVA, Bonferroni’s multiple comparison test), as compared with tumors recovered from WT mice subjected to the same treatment but not exposed to MTX (A) or from mice with the same genetic background receiving PBS (B). #P < 0.05; ##P < 0.01, as compared with tumors harvested from WT mice treated with MTX alone. (C and D) Immunofluorescence detection of CD3+ T cells in MCA205 tumors harvested from WT and Fpr1−/− mice 7 days after MTX treatment. Representative fluorescence micrographs are shown in (C), and quantitative data are reported in (D). *P < 0.05; ***P < 0.001 (one-way ANOVA, Bonferroni’s multiple comparison test), as compared with tumors harvested from mice with the same genetic background receiving PBS; ###P < 0.001, as compared with tumors harvested from WT mice treated with MTX. (E and F) Representative fluorescence micrographs showing the interaction between MCA205-DOXO corpses (in red) and mouse DC (purified by means of CD11c microbeads and labeled with the PKH67 green fluorescent cell tracker) (E) or MDA-MB-231-DOXO corpses (in red) and human DC (PBMCs, labeled with PKH67) (F). White arrows indicate portions of DOXO corpses interacting with the DC. (G and H) Microfluidic time-lapse recordings of the interaction between DOXO-pretreated WT or Anxa1−/− MCA205 cells and WT or Fpr1−/− splenocytes (G) or, alternatively, between DOXO-pretreated human MDA-MB-231 breast cancer cells and PBMCs collected from healthy donors bearing FPR1 CC (WT), CA (heterozygous, HT) or AA (mutated homozygous, MT) genotypes (H). Plots represent the trajectories of the individual splenocytes or PBMC (n = 6 to 30 cells per condition) toward target tumor cells (black spots) within 24 to 48 hours (splenocytes) or 48 to 72 hours (PBMC). (I and J) Quantification of the interaction times between WT or Fpr1−/− splenocytes and WT or Anxa1−/− MCA205 cells (I) or between FPR1CC, FPR1CA, or FPR1AA PBMC and MDA-MB-231 cells (J) that had been exposed or not exposed to DOXO. Each dot represents a single spleen cell (n = 38 to 97 cells per condition) or a single PBMC (n = 20 to 40 cells per condition). ***P < 0.001 (Mann Whitney test) referred to WT splenocytes (I) or FPR1CC PBMC (J) challenged with untreated tumor cells. ###P < 0.001, compared with WT splenocytes (I) or FPR1CC PBMC (J) challenged with DOXO-treated WT tumor cells.

The failure of Fpr1−/− DCs to encounter dying cancer cells and to take up tumor antigens should compromise T cell responses. Accordingly, the capacity of chemotherapy to stimulate the intratumoral proliferation of adoptively transferred T cells specific for a model antigen [ovalbumin (OVA)] was attenuated in Fpr1−/− mice harboring OVA-expressing MCA205 tumors, as compared with WT mice (fig. S21A). WT DCs purified from OVA-expressing tumors after chemotherapy stimulated OT1 responses upon their adoptive transfer into naïve mice, whereas Fpr1−/− DCs failed to do so (fig. S21, B and C). Moreover, the frequency of tumor-infiltrating CD3+ T cells was increased by anthracycline-based chemotherapy in WT mice but less so in Fpr1−/− hosts (Fig. 4, C and D).

The aforementioned results suggest that the initial defect of the Fpr1−/− immune system consists in the failure of DCs to migrate toward dying cancer cells. We attempted to explore this hypothesis in vitro. When mouse splenocytes or human peripheral blood mononuclear cells (PBMCs) expressing functional FPR1 (WT for mice, FPR1 CC for humans) were confronted with anthracycline-treated WT cancer cells in a microfluidic device (29), they migrated toward dying or dead cells treated with doxorubicin (but not to live untreated cells), and the fraction of leukocytes containing DCs exhibited a prolonged (>60 min) juxtaposition with the corpses (Fig. 4, E and F, and fig. S22A). Conversely, FPR1-deficient cells (Fpr1−/− for mice, FPR1CA or FPR1AA for humans) were largely unable to engage in such long-term interactions. Moreover, no stable conjugates were formed between DCs and dying or dead Anxa1−/− cancer cells (Fig. 4, E to J; fig. S22, A and B; fig. S23, A to C; and movies S1 to S24). Prolonged interactions of dying cancer cells with human blood-borne DCs were observed, but only if such DCs expressed functional FPR1 (fig. S22, C and D).

Altogether, our findings underscore the obligate contribution of the interaction between ANXA1 and FPR1 to the immune response against cancer cells succumbing to chemotherapy. Although it does not influence the recruitment of inflammatory DCs (with a CD11c+CD86+Ly6Chigh phenotype) to the tumor bed (which depends on other factors, including ATP and chemokines) (26, 28), FPR1 is required for DCs to come into close proximity of dying cancer cells, to establish stable contacts with corpses, to take up tumor-associated antigens, and to cross-present them to T cells. Hence, deficient FPR1 signaling results in defective intratumoral DC maturation and insufficient T cell–mediated anticancer immune responses, ultimately abolishing the efficacy of chemotherapy. Given the high allelic frequency of the loss-of-function FPR1 SNP rs867228 (17), it will be important to identify means for restoring or bypassing defective FPR1 signaling.

Supplementary Materials

Materials and Methods

Figs. S1 to S23

Tables S1 to S11

Movies S1 to S24

References (3040)

References and Notes

  1. Acknowledgments: The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. L.Z. filed a patent (WO2011131472 A1) that relates to the predictive role of SNPs in cancer therapy. G.K. and L.Z. are supported by the Ligue Nationale Contre le Cancer (Équipes Labellisées), Sites de Recherche Intégrées sur le Cancer (SIRIC) Socrates and Carpem, the Swiss Institute for Experimental Cancer Research (ISREC) Foundation, Agence Nationale pour la Recherche (AUTOPH, Emergence), Cancéropôle Ile-de-France, European Commission (ArtForce), European Research Council Advanced Investigator Grant (to G.K.), Fondation pour la Recherche Médicale, Fondation de France, the LabEx Immuno-Oncology, Institut National du Cancer (INCa), and the Paris Alliance of Cancer Research Institutes. Y.M. is supported by the LabEx Immuno-Oncologie and the Chinese National Thousand Talents Program; H.Y. by the Chinese National Thousand Talents Program; J.L.S. by the Deutsche Forschungsgemeinschaft (SFB704 and SFB645) and the Excellence Cluster ImmunoSensation; F.M. by the Italian Ministry of Health (RF-2011-02347120); G.S. by Associazione Italiana per la Ricerca sul Cancro (AIRC) (IG 14297); I.V. by AIRC (MFAG 14641), Italian Ministry of Health (RF_GR 2011-2012), and Programma per i Giovani Ricercatori “Rita Levi Montalcini” 2011; and A.S. and G.M. by AIRC.
View Abstract

Stay Connected to Science

Navigate This Article