[CANCER RESEARCH 62, 867–874, February 1, 2002]

Differential Binding of Drugs Containing the NGR Motif to CD13 Isoforms inTumor Vessels, Epithelia, and Myeloid Cells1

Flavio Curnis, Gianluigi Arrigoni, Angelina Sacchi, Lucia Fischetti, Wadih Arap, Renata Pasqualini, andAngelo Corti2

Department of Pathology and Molecular Medicine, DIBIT-San Raffaele H Scientific Institute, 20132 Milan, Italy [F. C., G. A., A. S., L. F., A. C.], and M. D. Anderson CancerCenter, Houston, Texas 77030 [W. A., R. P.]

ABSTRACT

The NGR peptide motif is an aminopeptidase N (CD13) ligand thattargets angiogenic blood vessels. NGR-containing peptides have provenuseful for delivering cytotoxic drugs, proapoptotic peptides, and tumornecrosis factor-� (TNF) to tumor vasculature. Given that CD13 is not onlyexpressed in the angiogenic endothelium but also in other cell types, themechanism(s) for the tumor-homing properties of NGR-drug conjugatesremains elusive. We have examined the expression of CD13 in normal andneoplastic human tissues and cells by using two anti-CD13 monoclonalantibodies. The immunoreactivity patterns obtained with cultured cellsand tissue sections from kidney, breast, and prostate carcinomas suggestthat different CD13 forms are expressed in myeloid cells, epithelia, andtumor-associated blood vessels. Both, direct binding assays with aCNGRCG-TNF conjugate (NGR-TNF) and competitive inhibition exper-iments with anti-CD13 antibodies showed that a CD13 isoform expressedin tumor blood vessels could function as a vascular receptor for the NGRmotif. In contrast, CD13 expressed in normal kidney and in myeloid cellsfailed to bind to NGR-TNF. Consistently with these results, neithermurine125I-NGR-TNF nor 125I-TNF accumulated in normal organs con-taining CD13-expressing cells after administration to mice. These findingsmay explain the selectivity and the tumor-homing properties of NGR-drugconjugates and may have important implications in the development ofvascular-targeted therapies based on the NGR/CD13 system.

INTRODUCTION

In vivo screening of peptide-phage libraries has proven to be apowerful tool for the discovery of ligands that selectively home totumor vessels (1–3). Among the targeting probes identified thus far, apeptide containing the NGR motif has been coupled to differentanticancer compounds, such as doxorubicin, proapoptotic peptides,and TNF,3 and shown to enable targeted delivery of these drugs totumor vessels (4–6), e.g., the antitumor activity of NGR-TNF inanimal models is 10–30 times stronger than that of TNF, whereas itstoxicity is similar (5). Evidence suggests that aminopeptidase N(CD13) is an important receptor targeted by NGR-containing conju-gates (5, 7). However, in addition to endothelial cells of angiogenicvessels, most cells of myeloid origin, including monocytes, macro-phages, granulocytes, and their hematopoietic precursors, expressCD13 (8–10). This peptidase is also abundantly expressed in thebrush border of epithelial cells from renal proximal tubules and smallintestine, in prostatic epithelial cells, in bile duct canaliculi, in mastcells, and, in some cases, in fibroblasts and smooth muscle cells (9,11–14). In most of these cells, CD13 immunoreactivity localizes to

the cell membrane or in the apical part of epithelial cells (15).However, cytoplasmic staining (15–17) and detection of solubleCD13 in human plasma have also been reported (18, 19). In addition,stromal fibrillar components of certain connective tissues containCD13 immunoreactivity (20). Given the widespread distribution ofCD13 throughout the body, the mechanisms involved in the tumor-homing properties of NGR-drug conjugates still remain unclear.

To analyze this specificity, we have investigated the expression ofCD13 in various human tissues and tumors using two mAbs andstudied the CD13/NGR-TNF interaction in tissues expressing highlevels of CD13. In addition, we have studied the interaction of humanNGR-TNF with normal and neoplastic tissues in vitro, as well as thebiodistribution of radio-labeled murine NGR-TNF and TNF in amouse model. Our findings suggest that different CD13 isoforms areexpressed within tumor vessels, normal epithelia, and myeloid cells.Moreover, we show that NGR-TNF can bind to vessels within or closeto tumor nodules but not to vessels in normal tissues and to epithelialcells of CD13-rich organs. On the basis of these observations, wepropose that the mechanism involved in the tumor-homing propertiesof NGR-drug conjugates relies on recognition of a CD13 isoformselectively expressed within tumor associated vessels.

MATERIALS AND METHODS

Cell Lines and Reagents. Mouse RMA-T lymphoma cells of C57BL/6origin (21) were prepared and cultured as described previously (22). Molt-4cells (acute lymphoblastic leukemia) transfected with the human CD13 cDNA(CD13/Molt-4) were prepared as described (7, 23). THP-1 (acute monocyticleukemia) cells were obtained from Dr. F. Blasi (San Raffaele H ScientificInstitute, Milan Italy). Human macrophages were isolated from peripheralblood mononuclear cells by standard techniques. mAb 13C03 (antihumanCD13, IgG1) was from Neomarkers, LabVision Corp. (Fremont, CA); mAbWM15 (antihuman CD13, IgG1) was from PharMingen (San Diego, CA);mAb JC/70A (antihuman CD31, IgG1) was from DAKO (Copenhagen, Den-mark); mAb QBEND10 (antihuman CD34, IgG1) was from Serotec (Oxford,United Kingdom); mAb 78 (IgG1) was kindly supplied by Dr. E. Barbanti(Pharmacia-Upjohn, Milan, Italy). mAb 78 is an antihuman TNF antibody ableto form stable complexes with soluble TNF (Kd: 3.2 � 10�10 M) and neutralizeits interaction with membrane receptors (24).

Human recombinant TNF and NGR-TNF (consisting of human TNF1–157

fused with the COOH terminus of CNGRCG) were prepared by recombinantDNA technology and purified from Escherichia coli cell extracts, as described(5). The protein concentration was measured with a commercial protein quan-tification assay kit (Pierce, Rockford, IL). The in vitro cytolytic activity ofNGR-TNF, estimated from a standard cytolytic assay with L-M mouse fibro-blasts (25), was (4.6 � 107) � 0.4 units/mg, whereas that of purified humanTNF was (5.45 � 107) � 3.1 units/mg. The hydrodynamic volume of NGR-TNF was similar to that of TNF, a homotrimeric protein (26), in gel filtrationchromatography on a Superdex 75 HR column (Pharmacia, Uppsala, Sweden).The molecular mass of NGR-TNF subunits was 17938.0 � 1.9 kDa (expectedfor NGR-TNF, 17939.4 kDa) by electrospray mass spectrometry (27).

Complexes of human NGR-TNF and anti-TNF mAb 78 (termed NGR-TNF/78) were prepared by incubating a mixture of 1 �g/ml NGR-TNF and 1 �g/mlmAb 78, both in PBS containing 2% BSA for 20 min (20°C). A mixture ofTNF and mAb 78 (termed TNF/78) was prepared in the same way using humanTNF instead of NGR-TNF.

Received 7/18/01; accepted 11/30/01.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

1 Supported by Associazione Italiana Ricerca sul Cancro and Ministero della Sanita ofItaly (Ricerca Finalizzata).

2 To whom requests for reprints should be addressed, at Department of Pathology andMolecular Medicine, DIBIT-San Raffaele H Scientific Institute, via Olgettina 58, 20132Milan, Italy. Phone: 39-02-26-43-48-02; Fax: 39-02-26-43-47-86; E-mail: corti.angelo@hsr.it.

3 The abbreviations used are: TNF, tumor necrosis factor �; mAb, monoclonal anti-body; NGR-TNF, CNGRCG-tumor necrosis factor; FACS, fluorescent-activated cellsorter.

867

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

Recombinant murine IFN� fused with CNGRCG (NGR-IFN�) was pre-pared by recombinant DNA technology, essentially as described for NGR-TNF, and purified by immunoaffinity chromatography using an antimouseIFN�. Reducing and nonreducing SDS-PAGE of the final product showed asingle band of 16 kDa. The synthetic peptide CgA (60–68), correspondingto the chromogranin A fragment 60–68, was prepared as described previously(28).

Immunohistochemical Studies. Surgical specimens of human tissues wereobtained from the Department of Histopathology, San Raffaele H ScientificInstitute. Sections (5–6 �m thick) of Bouin-fixed (4–6 h), paraffin-embeddedspecimens were prepared and adsorbed on polylysine-coated slides. Antigenswere detected using the avidin-biotin complex method as follows: tissue

sections were rehydrated using xylenes and a graded alcohol series, accordingto standard procedures. Tissue sections were placed in a vessel containing 1mM EDTA and boiled for 7 min using a microwave oven (1000 W). The vesselwas then refilled with 1 mM EDTA and boiled again for 5 min. The tissuesections were left to cool and incubated in PBS containing 0.3% hydrogenperoxide for 15 min to quench endogenous peroxidase. The samples were thenrinsed with PBS and incubated with 100–200 �l of PBS containing 2% BSA(PBS-BSA; 1 h at room temperature), followed by the primary antibody orNGR-TNF/78 complex in PBS-BSA (overnight at 4°C). The slides were thenwashed three times (3 min each) with PBS and incubated with PBS-BSAcontaining 2% normal horse serum (PBS-BSA-normal horse serum; VectorLaboratories, Burlingame, CA) for 5 min. The solution was then replaced with

Fig. 1. Expression of CD13-, CD31-, and NGR-TNF-bindingsites in normal (left panels) and neoplastic (right panels) renaltissues. Representative photomicrograph of tissues immunostainedwith mAb 13C03 (anti-CD13, diluted 1:2; A and B), 0.5 �g/mlWM15 (anti-CD13; C and D), JC/70A (anti-CD31, diluted 1:50; Eand F), 1 �g/ml NGR-TNF/78 (G and H), 1 �g/ml TNF/78, (I andL), and 1 �g/ml mAb 78 (anti-TNF; M and N). g, kidney glomeruli;t, tumor cells; arrowheads, vessels. A–N, �400.

868

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

3 �g/ml biotinylated horse antimouse IgG (H�L; Vector Laboratories) inPBS-BSA-normal horse serum and further incubated for 1 h at room temper-ature. The slides were washed again and incubated for 30 min with VectastainElite Reagent (Vector Laboratories) diluted 1:100 in PBS. A tablet of 3,3�-diamino-benzidine-tetrahydrochloride (Merck, Darmstadt, Germany) was thendissolved in 10 ml of deionized water containing 0.03% hydrogen peroxide,filtered through a 0.2-�m membrane, and overlaid on tissue sections for 5–10min. The slides were washed as above and counterstained with Harris’ hema-toxylin.

FACS Analysis. Expression of CD13- and NGR-TNF-binding sites bycultured human macrophages (stimulated with 1 �g/ml lipopolysaccharide for16 h), THP-1, MOLT-4, and CD13/MOLT-4 cells was measured by FACS.Before analysis, the cells were preincubated with Dulbecco’s PBS (BioWhit-

taker) containing 30% human serum and 2% fetal bovine serum for 5 min,followed by 1 �g/ml WM15, 1 �g/ml NGR-TNF/78, or mAb 13C03 (diluted1:2) in the same buffer (14 min, 4°C). After washing, the cells were incubatedwith a goat antimouse-FITC secondary antibody (Sigma Chemical Co.) dilutedin the same buffer (14 min, 4°C) and fixed with 2% formaldehyde in PBS.

In Vivo Biodistribution Studies. Murine 125I-NGR-TNF (1.41 �Ci/�g)and 125I-TNF (1.8 �Ci/�g) were prepared using Iodo-Gen precoated tubes(Pierce) according to the manufacturer’s instructions. 125I-NGR-TNF or 125I-TNF was injected i.p. into C57BL/6 mice weighing 19–20 grams (CharlesRiver Laboratories, Calco, Italy; 1 �g in 125 �l of 0.9% sodium chloridecontaining 0.1 mg/ml endotoxin-free human serum albumin). After 6 or 24 h,the animals were sacrificed and surgically dissected. The radioactivity intissues was quantified by �-scintillation counting of 125I.

RESULTS

Different CD13 Forms Are Expressed within Tumor Vesselsand Epithelia. The expression of CD13 in normal and neoplastictissues was evaluated by indirect immunohistochemistry using twoanti-CD13 mAbs (13C03 and WM15). The immunoreactivity of thesemAbs with a renal cell carcinoma and non-neoplastic kidney was firstinvestigated. The 13C03 epitope, but not the WM15 epitope, wasexpressed in the brush border of the renal proximal tubule epithelialcells (Fig. 1, A and C). The weak cytoplasmic staining of kidneytubules by WM15 is not likely to be specific, as it was observed alsowhen the antibody was omitted (data not shown). Although none ofthese epitopes were expressed by neoplastic cells (Fig. 1, B and D),both epitopes were expressed by stromal cells within and around the

Fig. 2. Expression of CD34- and NGR-TNF-binding sites in breastneoplastic tissues (left panels) and normal adipose tissue around theneoplastic nodules (right panels). Representative photomicrograph oftissues immunostained with QBEND10 (anti-CD34, diluted 1:100; A andB), 1 �g/ml NGR-TNF/78 (C and D), 1 �g/ml TNF/78 (E and F), mAb78 (G), and a mixture of 1 �g/ml NGR-TNF/78 plus 50 �g/ml NGR-IFN� (H). t, tumor nodules; arrowheads, vessels. A, B, and D–H, �400;C, �630.

Table 1 Binding of antibodies to renal cell carcinoma vessels (mAb WM15) andproximal tubules of normal kidney (mAb 13C03), by immunohistochemistry, in the

presence of various competitors

Competitora

Antibody binding

WM15 13C03

None � �NGR-TNF (25 �g/ml) � �NGR-IFN� (50 �g/ml) � NDb

CNGRC (100 �g/ml) � NDTNF (25 �g/ml) � �Human serum albumin (25 �g/ml) � �Synthetic CgA (60–68; 100 �g/ml) � ND

a The competitor in PBS containing 2% BSA was added in the blocking step and mixedwith the primary antibody.

b ND, not determined.

869

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

neoplastic lesion. The majority of stained cells in tumor stromacorresponds to endothelial cells, as suggested by similar stainingpatterns obtained with an anti-CD31 mAb, a well-known marker ofendothelial cells (Fig. 1F). Of note, both epitopes were either notexpressed or expressed very weakly in vessels of normal kidney,either within renal glomeruli or in connective tissue (Fig. 1, A and C).These results suggest that: (a) different antigenic forms of CD13 areexpressed within normal epithelial cells and endothelial cells oftumor-associated vessels; (b) the 13C03 epitope is expressed on bothisoforms; and (c) the WM15 epitope is differentially expressed onthese isoforms.

A CD13 Isoform Associated with Tumor Vessels Is a Receptorfor NGR-TNF in Renal Cell Carcinoma. Next, we compared theexpression of CD13 with that of NGR-binding sites by immunohisto-chemistry. Sections of normal kidney and renal cell carcinoma wereincubated with human NGR-TNF precomplexed with the antihumanTNF mAb 78 (NGR-TNF/78). Controls with TNF/78 complexes ormAb 78 alone were also included. These complexes offer the advan-tage that they can be used as a single reagent in parallel with otherantibodies. The staining patterns obtained with NGR-TNF/78 werevery similar to those of WM15 and distinct from those of 13C03. LikeWM15, NGR-TNF/78 interacted with tumor-associated vessels butnot with the brush border of renal proximal tubule epithelial cells(Fig. 1, G and H). No binding was observed with controls, such asTNF/78 or mAb 78 alone (Fig. 1, I–N). These results indicate that: (a)the binding of mAb 78 to endogenous TNF is negligible; (b) thebinding of TNF/78 to TNF receptors is undetectable; and (c) the

binding of NGR-TNF/78 depends on the interaction of the NGRdomain with an NGR receptor. Accordingly, the binding of NGR-TNF/78 was totally competed by coincubation with an NGR-IFN�conjugate (data not shown). These and the above results suggest thatthe NGR receptor and WM15 antigen colocalize in tumor-associatedvessels.

To establish that the CD13 isoform recognized by the WM15antibody is the one that is recognized by the NGR peptide, weperformed competitive binding assays. Various reagents were used inthese experiments. The binding of WM15 to tumor-associated vesselswas inhibited by NGR-TNF, NGR-IFN�, and CNGRC, but not byother control reagents lacking the NGR motif (Table 1). In contrast,the binding of 13C03 to the brush border of renal proximal tubuleepithelial cells was not competed by NGR-TNF.

In summary, NGR-TNF interacts with a receptor expressed in renalcell carcinoma-associated vessels that corresponds to a CD13 isoformrecognized by mAb WM15.

CD13 Expression in Breast and Prostate Cancer Metastasis andin Normal Tissues. To test whether other tumors express the CD13isoform recognized by NGR-TNF and mAb WM15, we analyzedhuman breast cancer tissue sections. In this case, an anti-CD34 mAbwas used to identify vessels around tumor nodules (Fig. 2A) and innormal adipose tissue in the same tissue section (Fig. 2B). NGR-TNF/78 recognized an antigen on the endothelial lining of vesselsclose to tumor nodules but did not stain normal vessels present inadipose tissue distant from the tumor (Fig. 2, C and D). Cytoplasmicstaining of tumor cells with NGR-TNF/78 but not with TNF/78 or

Fig. 3. Expression of CD13- and NGR-TNF-bindingsites in bone metastasis of breast cancer. Representativephotomicrograph of tissues immunostained with 1 �g/mlmAb 78 (B), 1 �g/ml TNF/78 (C), 1 �g/ml NGR-TNF/78(D), 13C03 (diluted 1:2) (E), and 1 �g/ml WM15 (F).Negative control without primary antibody (A). Arrows,vessels. A–C, E, and F, �400; D, �600.

870

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

mAb 78 alone was also observed (Fig. 2, E–G). Binding to vessels andtumor cells was competed with NGR-IFN� (Fig. 2H), suggesting thatin both cases, binding depended on the NGR domain.

To investigate whether NGR-TNF can also bind to vessels associ-ated with metastatic tumors, we performed immunohistochemicalanalyses of bone metastasis of human breast and prostate cancer.Analysis of breast cancer bone metastasis showed staining of vesselsin a manner reminiscent of that of anti-CD13 mAbs (Fig. 3). Thestaining was not limited to endothelial cells, as other cells aroundlarge vessels, presumably pericytes and smooth muscle cells, werepositive. Of note, staining of fibrillar components of fibrotic tissuesoccurred with both 13C03 and WM15, possibly related to a shed formof the membrane antigen. Interestingly, NGR-TNF/78 bound veryweakly or not at all to these forms.

Analysis of prostate cancer bone metastasis showed vessel stainingafter incubation with NGR-TNF/78 within and around the tumornodules (Fig. 4). Additionally in this case, staining was not limited tothe endothelial cells, as weak staining of neoplastic and stromal cellsoccurred with both NGR-TNF/78 and WM15.

Normal vessels in kidney, liver, small intestine, and placenta tissuesections were not stained by NGR-TNF/78 (data not shown).

In conclusion, the results suggest that human NGR-TNF can bind toa receptor expressed in vessels within or close to human primary andmetastatic tumors but not to normal tissues.

Characterization of CD13 in Myeloid and Lymphoid Cells. Tofurther characterize the CD13 isoform recognized by 13C03, WM15,and NGR-TNF/78, we investigated the binding of these reagents to

myeloid and lymphoid cells. FACS analysis showed that WM15recognizes THP-1 cells (acute monocytic leukemia) and lipopolysac-charide-stimulated human macrophages but not Molt-4 cells (acutelymphoblastic leukemia; Fig. 5). In contrast, neither 13C03 nor NGR-TNF/78 immunoreacted with these cell lines. Notably, these reagentsfailed to react with Molt-4 cells transfected with the human CD13cDNA. In addition, NGR-TNF (50 �g/ml) did not inhibit the bindingof WM15 to THP-1 cells (data not shown). The lack of detectablebinding of NGR-TNF/78 to CD13/Molt-4 cells apparently contrastswith the results of previous studies showing that these cells bindphages expressing the CNGRC peptide on their surface (7). Anexplanation for this discrepancy is that these cells could express a verylow fraction of CD13 molecules able to bind both NGR-phages andNGR-TNF/78. Whereas bound phage particles could be easily de-tected, even when present in low numbers, the binding of low amountsof NGR-TNF is undetectable by FACS. In addition, we have foundthat immunodetection of phage particles is at least two to three ordersof magnitude less sensitive than colony counting in bacterial infectionassay.4 Thus, although a number of phages can be detected as binderscompared with controls, the amount of NGR-binding CD13 isoform isquite low. In any case, the high level of CD13 expression and the lackof detectable binding of NGR-TNF strongly suggest that most CD13molecules on these cells are unable to bind NGR-TNF.

Granulocytes, monocytes, and lymphocytes present in the vessel

4 R. Pasqualini and W. Arap, unpublished observations.

Fig. 4. Expression of CD34- and NGR-TNF-bindingsites in bone metastasis of prostate cancer. Representa-tive photomicrograph of tissues immunostained with 1�g/ml mAb 78 (A), QBEND10 (diluted 1:100; B and C),1 �g/ml WM15 (D), 1 �g/ml TNF/78 (E), and 1 �g/mlNGR-TNF/78 (F). t, tumor cells; arrowheads, vessels. Aand B, �200; C–F, �400.

871

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

lumen of kidney, breast, and prostate tissue sections, as well astumor-infiltrating lymphocytes, were not stained by NGR-TNF/78 inimmunohistochemistry assays. Of note, NGR-TNF/78 and WM15 butnot TNF/78 immunoreacted with scattered granular cells, presumablymastocytes, within the neoplastic tissues (data not shown).

Biodistribution of 125I-NGR-TNF and 125I-TNF in Mice. Torule out the possibility that CD13 isoforms expressed by epithelialcells in kidney or other CD13-rich organs bind NGR-TNF in vivo, weexamined the biodistribution of radiolabeled murine NGR-TNF andTNF in a mouse model. 125I-NGR-TNF and 125I-TNF (1 �g each)were administered to C57BL/6 mice. The kidney:blood ratios of125I-NGR-TNF and 125I-TNF 6 or 24 h after administration weresimilar, suggesting that NGR-TNF does not accumulate in the kidney(Fig. 6). This finding agrees with the results of the above in vitrostudies showing a lack of binding of NGR-TNF to CD13 isoformsexpressed in normal kidney. Similarly, the tissue:blood ratios of otherorgans, including small intestine, liver, bone, lung, heart, muscle, andspleen, were similar, indicating that NGR-TNF does not accumulatein these organs.

DISCUSSION

NGR-containing peptides, identified by in vivo screening of phagelibraries, have proven useful for preparing anticancer drugs that targettumor vessels. Previous studies showed that aminopeptidase N

(CD13) is an important receptor for NGR-containing conjugates (7).This protein is an abundant myeloid differentiation antigen that is alsoexpressed by many nonhematopoietic cell types. For example, in theintestinal brush border, aminopeptidase N constitutes 6–8% of thetotal proteins and is thought to play an important role in proteindigestion (9). Other studies have shown a strong expression of CD13in the brush border of renal proximal tubules (14). Accordingly, usingmAb 13C03, we observed marked expression of CD13 antigen in theapical part of epithelial cells of proximal tubules in normal kidney.Nevertheless, the results of immunohistochemical and biodistributionstudies indicate that NGR-TNF does not bind to these cells in vitroand does not accumulate in the kidneys or other CD13-rich organs invivo. These data suggest that NGR-drug conjugates bind receptorswith a more restricted expression pattern than CD13. Immunohisto-chemical analysis of normal and neoplastic tissues with anti-CD13antibodies that recognize different epitopes (13C03 and WM15)showed that different CD13 isoforms are expressed in normal epithe-lial cells, tumor-associated vessels, and myeloid cells. Moreover, theresults of direct and competitive binding experiments with NGR-TNFand anti-CD13 mAbs suggest that a tumor vessel-related CD13 iso-form but not those associated with epithelial and myeloid cells func-tions as an NGR receptor.

The binding of WM15 to tissues but not 13C03 was competed withNGR-TNF or CNGRC peptide. This suggests that the NGR bindingsite sterically overlaps, at least partially, with the WM15 epitope onhuman CD13. The steric overlapping and the similar staining patternobserved with WM15 and NGR-TNF in tissue sections do not nec-essarily indicate that the structural determinants of WM15 epitope andNGR binding site are identical or have the same accessibility in all

Fig. 6. Biodistribution of murine 125I-NGR-TNF and 125I-TNF in C57BL/6 mice.Shown is the amount of radioactivity in various organs 6 h after the injection of 1 �g ofprotein/animal (5 mice/group in two separate experiments; A and B) and 24 h afterinjection (12 mice/group; C).

Fig. 5. Expression of CD13- and NGR-TNF-binding sites by cultured human macro-phages (stimulated with 1 �g/ml lipopolysaccharide for 16 h), THP-1, MOLT-4, andCD13/MOLT-4 cells, as measured by FACS.

872

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

tissues. Indeed, we found that NGR-TNF does not recognize WM15-positive myeloid cells, by FACS analysis, suggesting that the struc-tures of NGR binding site and WM15 epitope are different or differ-entially accessible in these cells. Thus, the CD13 isoform that bindsNGR-TNF is likely a subpopulation of WM15-antigenic forms.

The selectivity of NGR-TNF for tumor-associated endothelial cellswas not absolute. Other cells present in the tumor stroma and peri-nodular connective tissue, likely fibroblasts and mastocytes, were alsostained. In addition, we observed a weak reactivity of NGR-TNF withbreast and prostate cancer cells. Because it is known that breast cancercells usually express CD13 (14), it is possible that the staining oftumor cells was related to expression of CD13. We do not knowwhether these cells are accessible to systemically administered NGR-TNF. If they are, they may represent an additional homing site fortargeted delivery of NGR-drug conjugates to tumors. In any case,expression of CD13 by tumor cells is not a prerequisite for tumortargeting, because we have shown previously that even the growth ofCD13-negative tumor cells in vivo is efficiently inhibited by NGR-TNF when the tumors are well established and vascularized (5).

It is not known what the structural determinants responsible fordifferential binding of anti-CD13 mAbs and NGR-TNF to CD13isoforms are. CD13 is synthesized as a 130-kDa intracellular precur-sor of 967 amino acids and post-translationally modified in the Golgito produce a 150–240 kDa mature cell surface molecule (13, 29, 30).In the mature protein, 25–30% of the molecular weight is carbo-hydrate. Previous investigations have shown that differential or in-complete utilization of O-glycosylation sites results in at least fiveisoforms that are differentially recognized by antibodies, likely attrib-utable to variable masking of protein epitopes (31). We have foundthat enzymatic deglycosylation of soluble CD13, isolated from humanplacenta, with O-glycanase and neuraminidase does not increase thebinding of NGR-TNF in various ELISA and ligand blotting assays(data not shown). Moreover, treatment of normal kidney tissue sec-tions with these enzymes did not increase to binding of NGR-TNF tothe brush border of the epithelial cells (data not shown). These resultsargue against the hypothesis that carbohydrates mask the NGR bind-ing site in nonfunctional CD13 isoforms. As alternative hypotheses, itis possible that the differential reactivity of NGR-TNF with differentcells is related to different conformations of CD13, e.g., previousstudies showed that the binding of CD13 to natural peptide substratesor antibodies induces conformational changes and exposure of crypticepitopes (32). Alternatively, association with other proteins or ele-ments present in the tumor microenvironment or in tissues can causedifferential reactivity or availability to NGR-TNF. Additional work isnecessary to clarify this point.

It has been reported that the biological function of CD13 variesdepending on tissue microenvironment. CD13 exists as a membranebound or a soluble form, both of which catalyze the removal ofNH2-terminal residues, preferentially neutral, from small peptides (9).It has been suggested that CD13 plays a role in antigen processing(33), in neuropeptide and cytokine degradation (34, 35), in cell cyclecontrol and differentiation (13, 36, 37), and in tumor invasion andextracellular matrix degradation (38, 39). Two recent reports showedthat CD13 is also important in angiogenesis and is activated byangiogenic signals (7, 40). The presence of the NGR binding site intumor vessels could be required for the selective interaction of this sitewith other compounds potentially mimicked by NGR during angio-genesis and tumor invasion. NGR-TNF could be a valuable tool forthe identification of this CD13 isoform, as well as for the character-ization of its biological function.

The observation that human NGR-TNF can bind vessels in humanneoplastic tissue section could also have important clinical implica-tions. Previously, we showed that systemically administered murine

NGR-TNF is 10–30 times more efficient than murine TNF in de-creasing the tumor burden of animals bearing well-established lym-phomas, but both TNF and NGR-TNF were equally effective inanimals with freshly implanted, and hence avascular, tumors (5).Coadministration of murine NGR-TNF with an antimurine CD13antibody or a CNGRC peptide markedly decreased its antitumoreffects, suggesting that the NGR domain of NGR-TNF can interactwith murine CD13. The results of the present work suggest thatNGR-TNF can also bind human CD13. Human NGR-TNF mighthave, therefore, more potent antitumor properties than human TNF inpatients with vascularized tumors, as observed previously with murineTNF in mice. Indeed, several lines of evidence suggest that theantitumor activity of TNF largely depends on selective obstructionand damage to tumor-associated vessels (41–45). Because the clinicaluse of TNF as an anticancer drug is limited to locoregional treatmentsattributable to dose-limiting systemic toxicity, it is possible that tar-geted delivery of TNF to vessels via the NGR domain may allowsystemic and/or locoregional treatment of patients with lower toxicityand stronger antitumor effects.

ACKNOWLEDGMENTS

We thank Antonella Monno and Corazon Bucana for excellent technicaladvice and Maurilio Ponzoni for helpful discussions.

REFERENCES

1. Pasqualini, R., Koivunen, E., and Ruoslahti, E. �v Integrins as receptors for tumortargeting by circulating ligands. Nat. Biotechnol., 15: 542–546, 1997.

2. Koivunen, E., Arap, W., Rajotte, D., Lahdenranta, J., and Pasqualini, R. Identificationof receptor ligands with phage display peptide libraries. J. Nucl. Med., 40: 883–888,1999.

3. Koivunen, E., Arap, W., Valtanen, H., Rainisalo, A., Medina, O. P., Heikkila, P.,Kantor, C., Gahmberg, C. G., Salo, T., Konttinen, Y. T., Sorsa, T., Ruoslahti, E., andPasqualini, R. Tumor targeting with a selective gelatinase inhibitor. Nat. Biotechnol.,17: 768–774, 1999.

4. Arap, W., Pasqualini, R., and Ruoslahti, E. Cancer treatment by targeted drug deliveryto tumor vasculature in a mouse model. Science (Wash. DC), 279: 377–380, 1998.

5. Curnis, F., Sacchi, A., Borgna, L., Magni, F., Gasparri, A., and Corti, A. Enhance-ment of tumor necrosis factor � antitumor immunotherapeutic properties by targeteddelivery to aminopeptidase N (CD13). Nat. Biotechnol., 18: 1185–1190, 2000.

6. Ellerby, H. M., Arap, W., Ellerby, L. M., Kain, R., Andrusiak, R., Rio, G. D.,Krajewski, S., Lombardo, C. R., Rao, R., Ruoslahti, E., Bredesen, D. E., andPasqualini, R. Anticancer activity of targeted pro-apoptotic peptides. Nat. Med., 5:1032–1038, 1999.

7. Pasqualini, R., Koivunen, E., Kain, R., Lahdenranta, J., Sakamoto, M., Stryhn, A.,Ashmun, R. A., Shapiro, L. H., Arap, W., and Ruoslahti, E. Aminopeptidase N is areceptor for tumor-homing peptides and a target for inhibiting angiogenesis. CancerRes., 60: 722–727, 2000.

8. Chen, H., Kinzer, C. A., and Paul, W. E. p161, a murine membrane protein expressedon mast cells and some macrophages, is mouse CD13/aminopeptidase N. J. Immunol.,157: 2593–2600, 1996.

9. Riemann, D., Kehlen, A., and Langner, J. CD13: not just a marker in leukemia typing.Immunol. Today, 20: 83–88, 1999.

10. Drexler, H. G. Classification of acute myeloid leukemias: a comparison of FAB andimmunophenotyping. Leukemia, 1: 697–705, 1987.

11. Taylor, A. Aminopeptidases: structure and function. FASEB J., 7: 290–298, 1993.12. Shipp, M. A., and Look, A. T. Hematopoietic differentiation antigens that are

membrane-associated enzymes: cutting is the key. Blood, 82: 1052–1070, 1993.13. Razak, K., and Newland, A. C. The significance of aminopeptidases and haemato-

poietic cell differentiation. Blood Rev., 6: 243–250, 1992.14. Dixon, J., Kaklamanis, L., Turley, H., Hickson, I. D., Leek, R. D., Harris, A. L., and

Gatter, K. C. Expression of aminopeptidase-n (CD 13) in normal tissues and malig-nant neoplasms of epithelial and lymphoid origin. J. Clin. Pathol. (Lond.), 47: 43–47,1994.

15. Atherton, A. J., Monaghan, P., Warburton, M. J., and Gusterson, B. A. Immunocy-tochemical localization of the ectoenzyme aminopeptidase N in the human breast.J. Histochem. Cytochem., 40: 705–710, 1992.

16. Mechtersheimer, G., and Moller, P. Expression of aminopeptidase N (CD13) inmesenchymal tumors. Am. J. Pathol., 137: 1215–1222, 1990.

17. Bogenrieder, T., Finstad, C. L., Freeman, R. H., Papandreou, C. N., Scher, H. I.,Albino, A. P., Reuter, V. E., and Nanus, D. M. Expression and localization ofaminopeptidase A, aminopeptidase N, and dipeptidyl peptidase IV in benign andmalignant human prostate tissue. Prostate, 33: 225–232, 1997.

18. Tokioka-Terao, M., Hiwada, K., and Kokubu, T. Purification and characterization ofaminopeptidase N from human plasma. Enzyme (Basel), 32: 65–75, 1984.

873

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

19. Favaloro, E. J., Browning, T., and Facey, D. CD13 (GP150; aminopeptidase-N):predominant functional activity in blood is localized to plasma and is not cell-surfaceassociated. Exp. Hematol. (Charlottesville, Va), 21: 1695–1701, 1993.

20. Bordessoule, D., Jones, M., Gatter, K. C., and Mason, D. Y. Immunohistologicalpatterns of myeloid antigens: tissue distribution of CD13, CD14, CD16, CD31, CD36,CD65, CD66, and CD67. Br. J. Haematol., 83: 370–383, 1993.

21. Ljunggren, H. G., and Karre, K. Host resistance directed selectively against H-2-deficient lymphoma variants. Analysis of the mechanism. J. Exp. Med., 162: 1745–1759, 1985.

22. Moro, M., Pelagi, M., Fulci, G., Paganelli, G., Dellabona, P., Casorati, G., Siccardi,A. G., and Corti, A. Tumor cell targeting with antibody-avidin complexes andbiotinylated tumor necrosis factor �. Cancer Res., 57: 1922–1928, 1997.

23. Ashmun, R. A., and Look, A. T. Metalloprotease activity of CD13/aminopeptidase Non the surface of human myeloid cells. Blood, 75: 462–469, 1990.

24. Barbanti, E., Corti, A., Ghislieri, M., Casero, D., Rifaldi, B., Portello, C., Breme, U.,Trizio, D., and Marcucci, F. Mode of interaction between tumor necrosis factor � anda monoclonal antibody expressing a recurrent idiotype. Hybridoma, 12: 1–13, 1993.

25. Corti, A., Poiesi, C., Merli, S., and Cassani, G. Tumor necrosis factor (TNF) �quantification by ELISA and bioassay: effects of TNF �-soluble TNF receptor (p55)complex dissociation during assay incubations. J. Immunol. Methods, 177: 191–198,1994.

26. Smith, R. A., and Baglioni, C. The active form of tumor necrosis factor is a trimer.J. Biol. Chem., 262: 6951–6954, 1987.

27. Corti, A., Gasparri, A., Sacchi, A., Curnis, F., Sangregorio, R., Colombo, B., Siccardi,A. G., and Magni, F. Tumor targeting with biotinylated tumor necrosis factor �:structure-activity relationships and mechanism of action on avidin pretargeted tumorcells. Cancer Res., 58: 3866–3872, 1998.

28. Ratti, S., Curnis, F., Longhi, R., Colombo, B., Gasparri, A., Magni, F., Manera, E.,Metz-Boutigue, M. H., and Corti, A. Structure-activity relationships of chromograninA in cell adhesion. Identification and characterization of an adhesion site for fibro-blasts and smooth muscle cells. J. Biol. Chem., 275: 29257–29263, 2000.

29. Look, A. T., Ashmun, R. A., Shapiro, L. H., and Peiper, S. C. Human myeloid plasmamembrane glycoprotein CD13 (gp150) is identical to aminopeptidase N. J. Clin.Investig., 83: 1299–1307, 1989.

30. Look, A. T., Peiper, S. C., Rebentisch, M. B., Ashmun, R. A., Roussel, M. F.,Rettenmier, C. W., and Sherr, C. J. Transfer and expression of the gene encoding ahuman myeloid membrane antigen (gp150). J. Clin. Investig., 75: 569–579, 1985.

31. O’Connell, P. J., Gerkis, V., and d’Apice, A. J. Variable O-glycosylation of CD13(aminopeptidase N). J. Biol. Chem., 266: 4593–4597, 1991.

32. Xu, Y., Wellner, D., and Scheinberg, D. A. Cryptic and regulatory epitopes inCD13/aminopeptidase N. Exp. Hematol. (Charlottesville, Va), 25: 521–529, 1997.

33. Stryhn-Hansen, A., Noren, O., Sjostrom, H., and Werdelin, O. A mouse aminopep-tidase N is a marker for antigen-presenting cells and appears to be co-expressed withmajor histocompatibility complex class II molecules. Eur. J. Immunol., 23: 2358–2364, 1993.

34. Matsas, R., Stephenson, S. L., Hryszko, J., Kenny, A. J., and Turner, A. J. Themetabolism of neuropeptides. Phase separation of synaptic membrane preparationswith Triton X-114 reveals the presence of aminopeptidase N. Biochem. J., 231:445–449, 1985.

35. Hoffmann, T., Faust, J., Neubert, K., and Ansorge, S. Dipeptidyl peptidase IV (CD26) and aminopeptidase N (CD 13) catalyzed hydrolysis of cytokines and peptideswith N-terminal cytokine sequences. FEBS Lett., 336: 61–64, 1993.

36. Kenny, A. J., O’Hare, M. J., and Gusterson, B. A. Cell-surface peptidases asmodulators of growth and differentiation. Lancet, 2: 785–787, 1989.

37. Lendeckel, U., Arndt, M., Frank, K., Wex, T., and Ansorge, S. Role of alanylaminopeptidase in growth and function of human T cells (Review). Int. J. Mol. Med.,4: 17–27, 1999.

38. Saiki, I., Fujii, H., Yoneda, J., Abe, F., Nakajima, M., Tsuruo, T., and Azuma, I. Roleof aminopeptidase N (CD13) in tumor-cell invasion and extracellular matrix degra-dation. Int. J. Cancer, 54: 137–143, 1993.

39. Menrad, A., Speicher, D., Wacker, J., and Herlyn, M. Biochemical and functionalcharacterization of aminopeptidase N expressed by human melanoma cells. CancerRes., 53: 1450–1455, 1993.

40. Bhagwat, S. V., Lahdenranta, J., Giordano, R., Arap, W., Pasqualini, R., and Shapiro,L. H. CD13/APN is activated by angiogenic signals and is essential for capillary tubeformation. Blood, 97: 652–659, 2001.

41. Corti, A., and Marcucci, F. Tumour necrosis factor: strategies for improving thetherapeutic index. J. Drug Target., 5: 403–413, 1998.

42. Gasparri, A., Moro, M., Curnis, F., Sacchi, A., Pagano, S., Veglia, F., Casorati, G.,Siccardi, A. G., Dellabona, P., and Corti, A. Tumor pretargeting with avidin improvesthe therapeutic index of biotinylated tumor necrosis factor � in mouse models. CancerRes., 59: 2917–2923, 1999.

43. Nawroth, P. P., and Stern, D. M. Modulation of endothelial cell hemostatic propertiesby tumor necrosis factor. J. Exp. Med., 163: 740–745, 1986.

44. Nawroth, P., Handley, D., Matsueda, G., De Waal, R., Gerlach, H., Blohm, D., andStern, D. Tumor necrosis factor/cachectin-induced intravascular fibrin formation inmeth A fibrosarcomas. J. Exp. Med., 168: 637–647, 1988.

45. Eggermont, A. M., Schraffordt-Koops, H., Lienard, D., Kroon, B. B., van Geel, A. N.,Hoekstra, H. J., and Lejeune, F. J. Isolated limb perfusion with high-dose tumornecrosis factor-� in combination with interferon-� and melphalan for nonresectableextremity soft tissue sarcomas: a multicenter trial. J. Clin. Oncol., 14: 2653–2665,1996.

874

DIFFERENTIAL EXPRESSION OF CD13 ISOFORMS

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from

2002;62:867-874. Cancer Res   Flavio Curnis, Gianluigi Arrigoni, Angelina Sacchi, et al.   Isoforms in Tumor Vessels, Epithelia, and Myeloid CellsDifferential Binding of Drugs Containing the NGR Motif to CD13

  Updated version

  http://cancerres.aacrjournals.org/content/62/3/867

Access the most recent version of this article at:

   

   

  Cited articles

  http://cancerres.aacrjournals.org/content/62/3/867.full#ref-list-1

This article cites 43 articles, 21 of which you can access for free at:

  Citing articles

  http://cancerres.aacrjournals.org/content/62/3/867.full#related-urls

This article has been cited by 34 HighWire-hosted articles. Access the articles at:

   

  E-mail alerts related to this article or journal.Sign up to receive free email-alerts

  Subscriptions

Reprints and

  .pubs@aacr.orgDepartment at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/62/3/867To request permission to re-use all or part of this article, use this link

on April 19, 2018. © 2002 American Association for Cancer Research.cancerres.aacrjournals.org Downloaded from