Altered purinergic signaling in CD73-deficient mice inhibits tumor progression


  • Gennady G. Yegutkin,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Fumiko Marttila-Ichihara,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Marika Karikoski,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Jussi Niemelä,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Juha P. Laurila,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    Search for more papers by this author
  • Kati Elima,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    2. Department of Medical Biochemistry and Genetics, University of Turku, Turku, Finland
    Search for more papers by this author
  • Sirpa Jalkanen,

    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    Search for more papers by this author
  • Marko Salmi

    Corresponding author
    1. MediCity Research Laboratory, University of Turku, and National Institute of Health and Welfare, Turku, Finland
    2. Department of Medical Biochemistry and Genetics, University of Turku, Turku, Finland
    • MediCity Research Laboratory, University of Turku, Tykistökatu 6 A, FIN-20520 Turku, Finland Fax: +358-2-333-7000
    Search for more papers by this author


CD73/ecto-5′-nucleotidase dephosphorylates extracellular AMP into adenosine, and it is a key enzyme in the regulation of adenosinergic signaling. The contribution of host CD73 to tumor growth and anti-tumor immunity has not been studied. Here, we show that under physiological conditions CD73-deficient mice had significantly elevated ATPase and ADPase activities in LN T cells. In a melanoma model, the growth of primary tumors and formation of metastasis were significantly attenuated in mice lacking CD73. Among tumor-infiltrating leukocytes there were fewer Tregs and mannose receptor-positive macrophages, and increased IFN-γ and NOS2 mRNA production in CD73-deficient mice. Treatment of tumor-bearing animals with soluble apyrase, an enzyme hydrolyzing ATP and ADP, significantly inhibited tumor growth and accumulation of intratumoral Tregs and mannose receptor-positive macrophages in the WT C57BL/6 mice but not in the CD73-deficient mice. Pharmacological inhibition of CD73 with α,β-methylene-adenosine-5′-diphosphate in WT mice retarded tumor progression similarly to the genetic deletion of CD73. Together these data show that increased pericellular ATP degradation in the absence of CD73 activity in the host cells is a novel mechanism controlling anti-tumor immunity and tumor progression, and that the purinergic balance can be manipulated therapeutically to inhibit tumor growth.


Extracellular ATP, ADP and adenosine are powerful signaling molecules known to play key roles in controlling platelet aggregation, vascular tone and inflammatory responses 1–3. The purines released from damaged cells during pathological conditions function as a classical danger signal for the immune system. However, purines are also released from normal cells to the extracellular environment through several active mechanisms. Ecto-nucleoside triphosphate diphosphohydrolase (NTPDase) and ecto-5′-nucleotidase/CD73 are the two dominant cell-surface-expressed molecules dephosphorylating extracellular purines. In lymphoid tissues ATP and ADP are primarily hydrolyzed to AMP by NTPDase1/CD39, and further to adenosine by CD73. To trigger signaling cascades in the responding cells ATP and ADP bind to a series of ligand-gated (P2X) and G-protein-coupled (P2Y) receptors, whereas adenosine binds to one of the four adenosine receptors. Intriguingly, ATP and ADP generally evoke proinflammatory signals, whereas adenosine shows opposite effects by acting as an anti-inflammatory mediator. Along with the “classical” nucleotide-inactivating chain, the counteracting adenylate kinase (AK) and nucleoside diphosphate (NDP) kinase enzymes co-exist on the cell surface. The balance between these opposing nucleotide-scavenging and ATP-regenerating pathways may represent a key element in controlling the duration and magnitude of purinergic signaling 1–3.

CD73 is a glycosylphosphatidylinositol-linked surface protein expressed on subsets of leukocytes, vascular endothelial cells and on certain epithelial cells 4–7. The preferential expression of CD73, together with NTPDase, on CD4+CD25+FoxP3+ immunosuppressive Tregs has recently drawn much attention 8–11. The enzymatic activity of CD73 modulates leukocyte–endothelial cell contacts and it improves barrier functions of the vascular lining 12–14. Altered inflammatory reactions have been reported in CD73-deficient mice in multiple different models, including ischemia-reperfusion injuries and autoimmune diseases 13, 15–19.

CD73 can be expressed on several cancer types such as leukemia, glioblastoma, melanoma, and ovarian, bladder, thyroid, eosophageal, gastric, colon, prostate and breast cancer 3. The ecto-nucleotidase activity on the malignant breast cancer cells is known to enhance the migration, invasion and neovascularization of these cells and to support the growth of tumors 20, 21. CD73 expression has even been suggested to serve as a prognostic marker in certain cancer types, such as breast cancer 21. Although the functions of CD73 in cancer cells have been studied to some extent, the contribution of host CD73 activity to cancer progression has not been addressed. Here, we report that CD73-deficient T cells show up-regulated NTPDase activity, and that tumor progression and intratumoral accumulation of Tregs and mannose receptor (MR)+ macrophages, which are typically considered to be type 2 macrophages 22–24, are attenuated in CD73-deficient mice. Moreover, the composition of intratumoral leukocyte populations and tumor growth can be therapeutically manipulated by targeting CD73 and NTPDase. These data indicate that suppression of the host's CD73 activity might be a new tool to keep cancer cells under the control of anti-tumor immune responses.


Lymphoid ATPase and ADPase activities are increased in CD73-deficient mice

To analyze the whole purinergic enzymatic cascade in CD73-deficient mice we isolated lymphocytes from the LNs and spleen of unchallenged donors, and determined the ATPase, ADPase, ecto-5′-nucleotidase, adenosine deaminase (ADA) and AK activities (Fig. 1A). The cells from CD73-deficient mice were practically devoid of ecto-5′-nucleotidase activity, thus confirming that CD73 is the predominant enzyme conferring this activity in lymphoid cells (Fig. 1B). Interestingly, we found significant increases in the rates of 3H–ATP and 3H–ADP hydrolyses by lymphocytes isolated from peripheral LNs (PLNs), but not spleens of CD73-deficient mice, as compared with WT controls. In contrast, the ADA and AK activities did not differ between the two genotypes. Thus, the absence of CD73/ecto-5′-nucleotidase activity leads to a selective compensatory increase in the ATP- and ADP-hydrolyzing activities (=NTPDase) in LN lymphocytes.

Figure 1.

Altered purinergic metabolism in CD73-deficient mice. (A) The scheme of major purine-converting pathways on the lymphoid cell surface. The elements of inactivating cascade are composed of (1, 2) NTPDase, (3) ecto-5′-nucleotidase, (4) ADA, whereas the backward ATP-generating pathway is represented by (5) AK and (6) NDP kinase. (B) Total cell suspensions and (C) purified CD3+ T- and B220+ B-cell subsets were isolated from PLNs and spleens of CD73-deficient and WT mice and assayed for major purinergic activities by radio thin layer chromatography (TLC). The numbers in the panels refer to the particular catalytic reaction in (A). n, number of mice in each group. Data are mean+SEM. *p<0.05 as compared with WT controls (two-tailed Student's t test).

Since CD73 expression varies between different lymphocyte subpopulations and lymphoid organs 11, 25, 26, we further separated LN T and B cells and analyzed the enzyme activities separately in the two populations. The activities of 5′-nucleotidase in the two lymphocyte populations confirmed its preferential expression in T cells (Fig. 1C). T cells isolated from CD73-deficient mice displayed significantly higher ATPase and ADPase activities than those from WT mice. In contrast, there were no differences in the ADA or AK activities in T cells isolated from WT or CD73-deficient mice. Moreover, all purinergic enzymatic activities measured were comparable in B cells obtained from either genotype. These data indicate that there is a selective increase in the ATPase and ADPase activities in T cells in CD73-deficient mice. This probably explains why the increase in NTPDase was seen in LNs but not in spleen, which contains mainly B cells.

Loss of CD73 does not alter leukocyte composition in lymphoid organs

Loss of adenosine production and/or compensatory alterations in the purinergic signaling could affect the homing, differentiation or survival of lymphocytes. In PLNs 63±6% of CD4+ and 88±2% of CD8+ T cells expressed CD73, whereas in the spleen the numbers of double-positive cells were 66±3 and 78±1% (n=3) respectively. In both organs only about 10% of B cells expressed CD73. However, we found no differences between the genotypes in the overall percentages of either CD4+ or CD8+ cells in either LNs or spleen (data not shown). More detailed analyses of CD4+ cells revealed similar subpopulations of CD62 low (a marker for effector memory cells) in both genotypes (data not shown). Moreover, lymphocytes from both genotypes were equally responsive to the induction of L-selectin shedding by exogenous ATP (data not shown). Thus, under physiological conditions CD73 is not necessary for maintaining normal levels of lymphocytes in PLNs, even though many of these cells express this ecto-5′-nucleotidase.

Tumor growth and metastasis are attenuated in the absence of host CD73

Tumor cells are active in generating high levels of extracellular ATP, which can function as a tumor-promoting molecule 3. Since CD73-deficient mice had increased ATP- and ADP-hydrolyzing activities, we studied whether the anti-tumor immune response and tumor growth would be altered. To that end we searched for a CD73 tumor cell type, and found that B16 melanoma lacks CD73 protein (Fig. 2A) and does not display any 5′-nucleotidase activity. We then inoculated a luciferase-expressing B16 tumor subcutaneously in the pinna, and followed the growth of the primary tumor for 17 days. Immunohistochemical staining of the tumors showed that tumor cells remain CD73 after in vivo growth. When measuring the tumor growth using physical volume measurements and bioimaging we saw a trend of retarded growth in the CD73-deficient hosts. When the relatively big interexperimental variation was taken into account by normalizing tumor size against the WT mice in different experiments, both volume measurements and bioimaging showed that the tumors in CD73-deficient mice were significantly smaller than those in the WT mice (Fig. 2B and C).

Figure 2.

Impaired growth of primary and metastatic melanoma in CD73-deficient mice. (A) CD73 expression in B16 cells was analyzed by flow cytometry. B16-luc cells were injected into the pinna of WT and CD73-deficient mice. The growth of the primary tumor was determined on day 17 by measuring (B) the volume of the tumors and (C) the luciferase signal from the tumor cells. The metastatic load was determined by measuring (D) the luciferase signal, (E) the volume and (F) the weight of the draining LNs in the neck. Parameters in (B–D) are normalized so that value 1.0 is assigned to the mean of WT mice in each independent experiment. The raw values for the same parameters are shown in the insets (red, tumors grown in WT hosts; blue, tumors grown in CD73-deficient hosts). n=10 mice in each group. Data are mean±SEM.*p<0.05; **p<0.01; ***p<0.001 (two-tailed Student's t test).

We then studied the occurrence of metastasis in the draining LNs in the same model. In the CD73-deficient mice, the metastasis formation was significantly attenuated when assessing the metastatic load either by the luciferase activity of the metastatic cells, by the volume of the draining LN or by the weight of the draining node (Fig. 2D–F). The presence of metastatic cells was ascertained using histological sections from the draining LNs (data not shown).

We also inoculated B16 melanoma cells into the flanks of recipient mice. CD73-deficient mice had significantly smaller tumors also in this model (Fig. 3). Together, these data thus show that the lack of normal CD73 activity of the host inhibits tumor growth and metastasis formation.

Figure 3.

CD73 supports tumor growth, which can be therapeutically reversed by apyrase treatment. Luciferase-expressing B16 melanoma cells were injected into the flanks of WT and CD73-deficient mice, and apyrase or vehicle was injected peritumorally every second day for the duration of the experiment. The (A) volume and (B) luminescence signal were measured from the mice at day 7 (each dot represents an individual mouse). Horizontal line indicates the mean. ***p<0.001 (two-tailed Student's t test). (C) Representative bioluminescence images from mice in the different groups.

CD73 on tumor vessels is not needed for neoangiogenesis

CD73 is normally expressed on endothelial cells in certain vessels 6, and adenosine has proangiogenic effects in wound healing models 27. We therefore speculated that the diminished tumor growth in CD73-deficient mice could be caused by an abnormal angiogenic switch. Immunohistochemical analyses showed that CD73 is present on a subpopulation of CD31+ neoangiogenic endothelial cells in the melanoma (Fig. 4A). CD73+ vessels were identifiable both peritumorally and intratumorally. However, the number of intratumoral PV-1+ blood vessels or LYVE-1+ lymphatic vessels was not different between the WT and CD73-deficient mice (Fig. 4B and C). Hence, although expressed in neoangiogenic vessels, CD73 does not appear to be needed for their formation.

Figure 4.

Neoangiogenic switch is normal in the absence of CD73. B16Luc cells were injected subcutaneously into the flank of WT and CD73-deficient mice treated or not with apyrase. (A) The expression of CD73 in tumors was determined using immunohistochemistry. Green, CD31, red, CD73. The number of (B) PV-1+ blood vessels and (C) LYVE-1+ lymph vessels were quantified at the end of the experiment by immunohistochemistry. Data are mean+SEM. n=4 mice/group.

CD73 supports accumulation of Tregs and MR+ macrophages into the tumor

CD73 is expressed on Tregs and other lymphocytes, which are important for mounting normal immune responses against tumors. Therefore, we next analyzed the composition of intratumoral leukocyte populations in the WT and CD73-deficient mice. To avoid any effects of mechanic and enzymatic digestions on leukocyte recovery and antigen expression, we relied on immunohistochemistry for the enumeration of the intratumoral leukocytes. The numbers of CD8+ and CD4+ cells in the tumors did not reveal any genotype-specific differences (Fig. 5A and B). However, there were significantly fewer FoxP3+ cells (Tregs) in the tumors growing in CD73-deficient host than in the WT hosts (Fig. 5C). Moreover, although the total number of macrophages (F4/80+ cells) was similar in both genotypes (Fig. 5D), the number of MR+ cells was significantly lower in the mice lacking CD73 (Fig. 5E). The decrease in the numbers of these cells was not merely a consequence of smaller tumor volumes, since tumors of overlapping sizes (from different experiments) still showed a selective reduction of MR+ cells in the CD73-deficient host (Fig. 5E). Staining for Clever-1/stabilin-1, which is also highly enriched in type 2 macrophages 22, confirmed this observation of CD73-dependent macrophage differentiation defect (Fig. 5F). Additional staining of intratumoral cells for FIZZ/RELM-α did not reveal differences between the genotypes (132±11 and 145±13 cells/mm2 in WT and CD73-deficient mice respectively). In this context it should be noted that although FIZZ/RELM-α is considered to be a type 2 macrophage marker, it is also expressed on other hematopoietic and non-hematopoietic cells such as adipocytes, epithelial cells and eosinophils 22–24, 28. We found fewer intratumoral macrophages expressing CD169 (sialoadhesin), which has been proposed to be central in cross-presentation of tumor antigens to T cells 29, in the tumors growing in CD73-deficient mice (28±1 cells/mm2) than in WT mice (53±2 cells/mm2, p<0.01). Together, these data show that the numbers of macrophages expressing MR and Clever-1, markers compatible with the type 2 phenotype 22–24, are decreased within the tumors, if the host lacks CD73.

Figure 5.

CD73 supports accumulation of Tregs and MR+ macrophages in the tumors in an ATPase/ADPase dependent manner. Melanoma cells were injected into the flanks of WT and CD73-deficient mice, and apyrase or vehicle was injected peritumorally. The number of (A) CD8+ (mainly cytotoxic T cells), (B) CD4+ (mainly helper T cells), (C) FoxP3+ (mainly Tregs), (D) F4/80+ (a pan macrophage marker), (E) macrophage MR+ (type 2 macrophages) and (F) Clever-1+ (type 2 macrophages) was enumerated from the flank tumors using immunohistochemistry. Data are mean+SEM. n=4 mice/group (the whole tumor area was counted in each section), *p<0.05; **p<0.01; ***p<0.001 (two-tailed Student's t test). In (E) the correlation of the sizes of tumors grown in WT (black diamonds) and CD73-deficient hosts (grey diamonds) to the numbers of intratumoral MR+ cells (data pooled from two independent experiments) are also shown. (G) mRNA expression of the indicated target genes in CD45+ tumor-infiltrating leukocytes of CD73-deficient mice. The expression level in WT mice is 1.0 by definition (n=5 mice in both groups). Only the genes with ≥2-fold difference are depicted.

We used the tumor-infiltrating leukocytes for quantitative PCR analyses of immune-related genes. The results showed that intratumoral CD45+ cells isolated from CD73-deficient mice had twofold more IFN-γ mRNA and also the expression of several INF-γ-inducible genes such as Smad 3, Smad 7 and Socs 2 was induced (Fig. 5G, and Supporting Information Table 1). Notably, intratumoral leukocytes from CD73-deficient mice had more than eight times higher expression of Nos2 when compared with those from WT controls. The level of IL-10 mRNA was not different between the genotypes, and IL-4 was not detectable in any sample. IFN-γ and Nos2 are well-established markers of type 1 macrophage polarization 22. Therefore, these results are in line with our immunohistological data that in the absence of CD73 activity fewer tumor macrophages show a type 2-like phenotype (and consequently, since there is no difference in the total numbers of all macrophages (F4/80+ cells), more macrophages exhibit the type 1-like phenotype).

Since we found that many tumor vessels were CD73+, we studied the role of this molecule in recruitment of leukocytes into the tumor. Tumor-infiltrating leukocytes were isolated from WT melanomas, and their adherence to melanoma vessels in tumors grown either in the WT or CD73-deficient mice were analyzed. When compared to the WT vasculature (100%), the binding of tumor-infiltrating leukocytes to CD73-deficient vasculature was only 45±8% (mean±SEM, p<0.02). Thus, the lack of endothelial CD73 in CD73-deficient mice can at least partially explain the diminished leukocyte infiltration in the tumors.

Pharmacological manipulation of nucleotide-inactivating cascade inhibits melanoma growth

The attenuated growth of tumors in the CD73-deficient mice having increased lymphoid ATPase and ADPase activities is compatible with the possibility that decreased peritumoral ATP concentration is detrimental to the tumor. To study this hypothesis experimentally, we injected melanoma cells into the WT mice, and then treated the tumors locally with apyrase, which hydrolyzes ATP and ADP into AMP. We found that apyrase-treated mice had significantly smaller tumors than vehicle-treated animals (Fig. 3). In addition, the tumor size in apyrase-treated WT mice was not different from those seen in CD73-deficient mice. Strikingly, apyrase treatment had no effect in tumor-bearing CD73-deficient mice. These data strongly suggest that lowering of the peritumoral ATP levels either therapeutically by apyrase or genetically by deletion of CD73 effectively inhibits tumor growth.

In the apyrase-treated WT mice, the numbers of Tregs (FoxP3+) and MR+and Clever-1+macrophages were lower than in control-treated WT mice (Fig. 5). In fact, the numbers of these cell types in the apyrase-treated WT mice were at a similar level as in the vehicle-treated CD73-deficient mice (also having higher NTPDase activity). Apyrase treatment had no effect on these leukocyte populations in the mice lacking CD73. Moreover, apyrase treatment significantly increased the number of CD8+ T cells in the tumors in both genotypes.

Finally, we tested whether the beneficial effects of CD73 deletion on tumor progression can also be achieved by pharmacological manipulation of CD73 activity. Melanoma-bearing mice were treated peritumorally with a non-hydrolyzable nucleotide analog α,β-methylene-adenosine-5′-diphosphate (AMPCP), which selectively inhibits ecto-5′-nucleotidase. The results showed significant inhibition of tumor growth in WT animals (tumor volume 415±83 in PBS-treated mice and 121±24 mm3 in AMPCP-treated mice respectively, n=3–4 animals/group). AMPCP treatment had no effect on tumor volume in CD73-deficient mice (tumor volume 150±34 and 150±95 mm3 in PBS- and AMPCP-treated CD73-deficient mice, n=4 mice/group). Thus, chemical inhibition of CD73 activity is a therapeutically amenable option to control tumor growth.


We have shown here that tumor growth is impaired in CD73-deficient mice. This correlates with diminished intratumoral accumulation of Tregs and macrophages expressing type 2 markers (MR, Clever-1, IFN-γ and NOS2) in the absence of CD73. Lack of CD73 results in increased ATP- and ATP-hydrolyzing activity in immune cells, and we show that by reducing peritumoral ATP levels or by inhibiting CD73 activity in WT mice we can reproduce the CD73-deficient phenotype. Together, our data indicate that normally the host CD73 activity shifts the balance of purinergic signaling into a direction that favors tumor progression and accumulation of Tregs and MR+ macrophages, and that this pathway is amenable to therapeutic interventions.

CD73-deficient mice display enhanced leukocyte extravasation at sites of inflammation in several ischemia-reperfusion models, and also the vascular permeability is increased in the absence of CD73 27. It has been firmly established that these effects are largely mediated by diminished adenosine production in these mice. However, the other enzymes involved in the inactivation and/or transphosphorylation of ATP↔ADP↔AMP, and further degradation of AMP into adenosine and inosine have not been previously studied in the CD73-deficient mice. Here, we confirmed that CD73 was expressed both in a subpopulation of CD4+ and CD8+ T lymphocytes. T cells had significantly increased ATPase and ADPase activities in the CD73-deficient mice. This suggests that the extracellular levels of proinflammatory ATP and procoagulant ADP molecules are lower in these mice. However, since extracellular AMP hydrolysis is also largely blocked in the absence of CD73, the concentration of extracellular adenosine, which is an anti-inflammatory molecule, is actually also decreased in the absence of CD73. Thus, the net effect of CD73 deficiency may be to tilt the balance of purinergic signaling towards a state in which AMP accumulates in the body.

The tumor microenvironment is capable of diverting the inflammatory reaction in a way that paradoxically enhances tumor growth. Intratumoral infiltration of Tregs and intratumoral differentiation of type 1 macrophages into type 2 macrophages are two key events in this immune evasion process 23, 30–33. Our findings indicate that the altered purinergic balance in the absence of CD73 inhibits this detrimental process, inasmuch the tumors in CD73-deficient mice had specific decrease in the numbers of intratumoral Tregs and MR+ macrophages when compared with the WT mice. Interestingly, type 2 macrophages also show altered expression of purinergic receptors, which may link the CD73 and altered NTPDase activities to the observed phenotype 34. Moreover, tumor-infiltrating leukocytes in CD73-deficient mice showed increased IFN-γ synthesis. Since the transcription factor T-bet was actually down-regulated in tumor-infiltrating leukocytes in CD73-deficient mice, we speculate that IFN-γ is mainly produced by CD8+ cells, which in contrast to CD4+ and NK cells do not require T-bet for IFN-γ production 35. IFN-γ inhibits tumor formation and drives macrophage polarization into classically activated type 1, which show multiple anti-tumoral properties 30, 36. Notably, increased IFN-γ synthesis has also been recently reported in CD73-deficient mice during allograft rejection and in gastritis 37, 38. Interestingly, adenosine prevents IFN-γ-induced STAT phosphorylation and macrophage activation 39, and ATP has been reported to impair IFN-γ secretion in blood cells 35. Moreover, in type 2 macrophages extracellular ATP functions via its pyrophosphate chains, independently of purine receptors, to inhibit the production of active proinflammatory IL-1β 40. Several tumors produce high levels of extracellular ATP 41, 42. Extracellular ATP can have direct protumorigenic effects by activating P2 receptors on tumor cells, which increases tumor cell survival and migration 3. Thus, the up-regulated NTPDase activity in CD73-deficient mice could decrease extracellular ATP within the tumor, and together with diminished adenosine production, inhibit the development of the immune-suppressing microenvironment in the tumor.

Tumor-infiltrating leukocytes from CD73-deficient mice showed highly elevated NOS2 synthesis. Interestingly, inducible NOS is one of the best markers of classically activated M1 macrophages, and its synthesis is driven by IFN-γ 22. Functionally, these leukocytes use nitric oxide for several effector functions including signaling and killing of nitric oxide sensitive tumors 43. However, in tumors NOS2 is derived from many other sources in addition to macrophages, and it correlates positively with poor prognosis 44. Hence, although the overall pathophysiological significance of induced NOS in the absence of CD73 remains to be resolved, we suggest that normally CD73 may suppress NOS2 expression in tumor-infiltrating macrophages, which may be involved in their conversion into tumor promoting type 2 cells.

It is intriguing that tumor progression is decreased both in CD39-deficient mice, in which the hydrolysis of ATP and ADP is blocked 45, and, as shown here, in CD73-deficient mice, in which hydrolysis of ATP and ADP is enhanced. Tumor neoangiogenesis is defective in CD39 mice 45, but not in CD73-deficient mice. In CD39-deficient mice, the numbers of tumor-infiltrating macrophages were reported to be decreased, but no distinction between type 1 and 2 cells was made 45. Moreover, absence of CD39 on Tregs has been shown to inhibit metastasis formation through induction of NK cell activity 46. Thus, CD39 and CD73 activities may affect partially distinct vascular and immune cell populations. Moreover, the physical interactions of CD39 with other molecules, such as scaffolding protein RanBPM 47, which further binds to receptors for oncogenic growth factors and integrins, may exert non-purine-dependent effects on tumor growth. Taken together, we propose that the finely tuned balance between the extracellular ATP, ADP, AMP and adenosine, rather than a single purine, is decisive in the control of tumor progression. In fact, in processes such as granulocyte chemotaxis and tumor cell migration in vivo, such interdependence of ATP-mediated and adenosine-mediated signaling is known to regulate the net outcome of the response 48. For instance, both the anti-CD73 antibody treatment, which inhibits adenosine production, and apyrase treatment, which is expected to increase adenosine concentrations, decreased migration of CD73+tumors cells in vitro 49. This could explain why the two genotypes shifting ATP/ADP levels in opposite directions could both actually suppress tumor growth.

Tumor cells derived from human breast, prostate, thyroid and esophageal carcinomas, from melanoma and from glioblastoma can be CD73+3. The ecto-nucleotidase activity is known to be utilized by the breast cancer cells to enhance their adhesion, migration and invasion via adenosine receptor-mediated pathways 20, 21, 49, 50. Targeting of CD73 by antibodies and siRNA attenuates the growth and metastasis of CD73+tumors in a T- and/or B-cell-dependent manner 49, 50. Interestingly, anti-CD73 therapy, which results in diminished adenosine production, was inefficient against CD73 breast tumors 49. Our study is the first one to dissect the contribution of host CD73 in the progression of tumors. It strongly suggests that some of the beneficial effects seen in previous studies may actually be dependent on the inhibition of host CD73 rather than targeting the tumor. Moreover, our data show that the host CD73 is a potential therapeutic target for controlling tumor progression also in those cases in which tumor cells themselves lack or loose CD73 expression.

The altered purinergic signaling cascade can offer new therapeutic targets for inhibiting tumor growth. We showed that the scavenging of extracellular ATP in tumors by soluble apyrase treatment or CD73 blockade by AMPCP retarded growth of CD73 tumors in vivo. The phenotypes of apyrase-treated WT mice and that of control-treated CD73-deficient mice were virtually indistinguishable in terms of the kinetics of tumor growth and in the composition of intratumoral Treg and MR+ macrophage infiltrates. Moreover, apyrase treatment had no beneficial effect on tumor growth in CD73-deficient mice, and it did not alter these intratumoral leukocyte subpopulations either.

CD73 is induced by HIF-1a under hypoxic conditions 51. Because larger tumors are typically hypoxic, induction of CD73 in the stromal cells is very likely in clinical settings. Hence, it may be useful to be able to counteract the effects of inducible CD73 on intratumoral leukocyte accumulation by altering the purinergic signaling by enzyme therapy. These findings also highlight the novel fact that mechanistically the increased ATPase and ADPase activities, together with the reduced adenosine production, in CD73-deficient mice are major players in the improved control of tumor growth.

Materials and methods


WT and CD73-deficient mice on a C57BL/6 background (kindly provided by Linda Thompson) have been described earlier 13, 18. Age- and sex-matched animals were used in all experiments. All animal experiments were approved by the local animal care committee.

Tumor models

B16-F10 melanoma cells stably transfected with luciferase were obtained from Xenogen, and maintained in MEM/Earle's balanced salts medium containing 10% FCS, 200 mM L-glutamine, 1 mM sodium pyruvate, 1 mM non-essential amino acids, MEM vitamin solution and penicillin and streptomycin. Two different tumor models were used, an ear model for measuring the dissemination of tumors by lymphatic vessels and a flank model for characterizing primary tumors in detail. For the dissemination model 52, melanoma cells were injected subcutaneously into the left pinna of the mice (4×105 cells in 30 μL RPMI1640). For the local growth model 53, the same number of cells was injected subcutaneously into the flank of the mice. In both models, the growth of primary tumors was followed by measuring the luminescence signal after i.p. administration of luciferin followed by in vivo imaging system (IVIS) 50 bioimaging. The volume of the tumor was also analyzed using an electronic caliber. In the ear model, the in vivo imaging system (IVIS) signal, weight and volume of the draining LNs were also analyzed. At the end of all experiments, the tumors were isolated and used for immunohistochemistry or for cell separations.

Isolation of lymphoid cells

PLNs (axial and inguinal) and spleen were collected from unchallenged mice, and single-cell suspensions were generated by mechanical teasing. Erythrocytes were lysed from the spleen samples using a hypotonic buffer. T cells and B cells were isolated using MACS MicroBeads conjugated to monoclonal rat anti-mouse CD45R (B220) and VarioMACS As depletion columns (Miltenyi Biotech).

The tumor-infiltrating leukocytes were released from the melanomas using collagenase D digestion and gentle teasing through a metal grid, and purified with CD45–PE staining followed by anti-PE Easysep beads 53. This population was routinely found to be >80% leukocytes.

Enzymatic assays

Specific lymphoid purinergic activities were determined by using 2,83 H–ATP, 2,83 H–ADP (PerkinElmer), 2–3 H–AMP or 2-3 H–adenosine (Amersham Biosciences), as described previously 54. Briefly, the lymphocyte suspensions (5–10×104 cells) were incubated at 37°C in a final volume of 80 μL RPMI-1640 supplemented with 4 mM β-glycerophosphate with the following tracer substrates: 500 μM 3H–ATP (ATPase), 500 μM 3H–ADP–(ADPase), 300 μM 3 H–AMP (CD73), 300 μM 3 H adenosine (ADA), 400 μM 3 H–AMP plus 800 μM γ-phosphate-donating ATP (AK). The incubation times were chosen to ensure the linearity of the reaction (i.e. the amount of the enzyme products is not allowed to exceed 10–15% of the amount of the original substrate). Mixture aliquots were applied onto Alugram G/UV254 sheets (Macherey Nagel) and separated using TLC. The enzymatic activities were determined using scintillation β-counting, and expressed as nmol of the labeled substrate metabolized per 1 h by one million cells.

FACS analyses and immunostaining

Lymphocyte phenotyping by flow cytometry was done as described earlier 52, 53. For two-color staining, the isolated cells were first incubated with anti-CD73 mAb TY23, followed by FITC-conjugated anti-rat Ig, and finally by a cocktail of mAbs containing PerCP-Cy5.5-conjugated anti-CD8, Alexa647-conjugated- anti-CD4, and Pacific Blue 220-conjugated B220. In other experiments, the cells were stained with FITC-conjugated anti-mouse CD3, CD8, and CD62L (L-selectin) mAbs (BD Biosciences), in combination with R-PE-conjugated CD4 mAb (Caltag Laboratories). The number of Tregs was determined using a commercial mouse Treg kit (e-Biosciences). Tumor-infiltrating leukocytes were preincubated with Fc-block, and then stained with TY23, FITC-anti-rat Ig and APC-CD45, followed by 7-AAD for live/dead cell discrimination. The samples were analyzed using an LSRII cytometer.

Tumors were snap frozen in liquid nitrogen, and 5 μm acetone-fixed frozen sections were cut. The sections were stained with the indicated primary antibodies and fluorescent second-stage reagents. In certain experiments, anti-CD73 mAb TY23 was detected with Alexa 546-conjugated anti-ratIg, and the sections were then stained with Alexa448-conjugated anti-CD31 mAb to visualize the vessels. Anti-CD169 (AbD Serotech) and Relm-α (Abcam) antibodies were also used for immunohistochemistry. The number of intratumoral leukocytes was enumerated by microscopic counting from ≥5 randomly selected high magnification fields/sample.

Adhesion assays

Tumor-infiltrating leukocytes were isolated from WT melanomas, and their binding to vessels in tumors grown in WT and CD73-deficient mice was analyzed using the frozen section binding assay, as described earlier 55.


Isolated CD45+ tumor-infiltrating leukocytes were immediately lysed in the guanidine thiocyanate-containing lysis buffer of NucleoSpin RNAII Total RNA Isolation kit (Macherey-Nagel) for subsequent RNA isolation. Total RNA was reverse-transcribed into cDNA using iScript cDNA Synthesis Kit (BioRad). Equal amount of samples were loaded into TaqMan Mouse Immune Array micro fluidic cards (Applied Biosystems) and run using a 7900HT Fast Real-Time PCR System (Applied Biosystems) in the Finnish Microarray and Sequencing Center, Center for Biotechnology, Turku, Finland. The results were analyzed with SDS 2.3 software using relative quantitation. The normalization was performed against 18S rRNA, which was chosen as a representative house keeping gene.

In vivo apyrase and AMPCP treatments

B16 cells were mixed with apyrase, or left untreated (PBS), and immediately (<5 min) injected into the flanks of WT and CD73-deficient mice. Then, apyrase (1.5 units in 50 μL volume) or PBS (control) was injected into the peritumoral area using a 30G needle twice at 2-day intervals, and the animals were killed 3 days after the last injection. Pharmacological blockade of CD73 was achieved by peritumoral injections of AMPCP 56 (1 mM in 50 μL volume) using the same protocol (two injections at 2-day intervals, animals killed 3 days after the last injection).

Statistical analyses

The numerical data are presented as the mean±SEM. The difference between two groups was analyzed using Student's t-test (two-tailed). p-Values <0.05 were considered to be significant.


We thank Linda Thompson for providing the CD73-deficient mouse line, and Mikko Laukkanen for critical reading of the manuscript. This work was supported by the Finnish Academy and the Sigrid Juselius Foundation (to S. J. and M.S.).

Conflict of interest: The authors declare no financial or commercial conflict of interest.