Tetraspanin protein CD9 interacts with metalloprotease CD10 and enhances its release via exosomes



D. Mazurov, Laboratory of Immunochemistry, Institute of Immunology, 24-2 Kashirskoe shosse, Moscow 115472, Russia

Fax: +7 499 617 7765

Tel: +7 499 612 8854

E-mail: dvmazurov@yandex.ru


Tetraspanins interact with a wide variety of transmembrane and intracellular proteins called molecular partners, and modulate their function. In this article, we describe a new partner of tetraspanin web, membrane metalloprotease CD10, which is selectively associated with CD9. By constructing chimeras between tetraspanins CD9 and CD82 (the latter does not interact with CD10) or by using site-directed mutagenesis, we determined that a portion of the large extracellular loop from the CCG motif to transmembrane domain 4, as well as the C-terminal tail of CD9, are involved in the interaction with CD10. The stable expression of wild-type CD9 in K562 CD10-positive cells enhanced the level of CD10 released with exosomes five-fold. In contrast, the expression of chimeric CD9, which contained the cytoplasmic C-terminal domain from CD82, had little effect on CD10 release. Short hairpin RNA knockdown of CD9 expression in Nalm-6 pre-B cells resulted in a two-fold reduction in the amount of endogenous CD10 released with microvesicles. The peptidase activity of CD10 measured either on cells or on exosomes correlated with the level of CD10 expression, and was not significantly modulated by CD9 expression as such. Our data suggest that the interaction of CD10 with tetraspanin CD9 can play an important role in the redistribution of peptidase activity from the cell surface to outer microenvironments. In bone marrow, where CD10 presumably contributes to the maturation of pre-B cells and migration of B cells to the blood circulation, release of CD10 peptidase activity with exosomes may effectively regulate extracellular matrix microenvironments.

Structured digital abstract






endoplasmic reticulum


fluorescence-activated cell sorting


fluorescence immunoprecipitation analysis


large extracellular loop


multivesicular body


short hairpin RNA


transmembrane domain


western blot




Tetraspanins belong to a large family of proteins, which span the membrane four times and form small and large extracellular loops, one inner loop, and N-terminal and C-terminal ends located inside the cytoplasm. The conserved motifs CCG and PXSC in the large extracellular loop (LEL) stabilize the folding of this domain via the formation of two to four intramolecular disulfide bonds, and are hallmarks of the members of this family [1]. Tetraspanins predominantly interact laterally with each other and with a variety of transmembrane proteins and intracellular molecules known as molecular partners. Through these interactions, tetraspanins are organized into large molecular complexes called tetraspanin-enriched membrane microdomains or tetraspanin web. Tetraspanin-enriched membrane microdomains maintain the protein scaffold, and modulate cell adhesion, motility, migration, membrane fusion, tumor invasion, and intracellular signaling (reviewed in [2-4]). Tetraspanins are expressed on a broad variety of cells and tissues. Most of them are located and function on the plasma membrane, with the exception of CD63 (LAMP-3), which is found on lysomal membranes and limiting membranes of multivesicular bodies (MVBs). Exosomes, which are membrane microvesicles released from cells after fusion of MVBs with the plasma membrane, are also enriched in tetraspanins [5, 6].

CD9 is a 21–24-kDa tetraspanin that has one N-glycosylation site localated in the small extracellular loop. The tissue and cellular distribution of CD9 is wide, and includes many epithelial cells, endothelium, hematopoietic cells, smooth muscle cells, pre-B cells, and platelets. CD9 interacts directly with the immunoglobulin-like proteins EWI-2 [7] and EWI-F [8, 9], pregnancy-specific glycoprotein PSG17 [10], and the membrane-bound form of heparin-binding epidermal growth factor-like growth factor [11], as well as with CD81 and CD9, to form heterodimers and homodimers, respectively [12, 13]. Less strong associations of CD9 with other proteins, such as major histocompatibility complex, integrins, membrane proteases and conventional protein kinase C have been demonstrated [14-17]. CD9 has been implicated in egg–sperm fusion [18-20], diphtheria toxin binding [21, 22], monocyte [23] and myoblast [24] fusion, tumor growth suppression [25-27], and modulation of cell adhesion and migration [28, 29]. In bone marrow, CD9 contributes to stem cell differentiation [30], motility [31], and homing [32].

Although profiling of CD9 molecular partners has been performed on some epithelial cancer cells [15] and mature lymphoid cells, the spectrum of molecules associated with CD9 on the early progenitors of lymphocytes have not been comprehensively studied. In this study, we identified a number of proteins expressed on the pre-B cell line Nalm-6 that are associated with CD9. We found that CD10 metalloprotease is a new partner of CD9, and mapped the domains in CD9 that are responsible for the interaction with CD10. CD10 (neprilysin or neutral endopeptidase) is a zinc-dependent membrane protease that was initially described in leukemic B cells as a common acute lymphoblastic leukemia antigen. Later, CD10 was found in other tissues and organs. CD10 functions to cleave certain peptide mediators: enkephalins, neurotensin, substance P, angiotensin I, bradykinin, tachykinin, and oxytocin (reviewed in [33, 34]). Both CD9 and CD10 are transiently expressed on early B cells and pre-B cells [35], where they probably contribute to B-cell maturation and homing. Here, we demonstrate that association of CD10 with CD9 enhances the release of CD10 with exosomes. We assume that redistribution of CD10 activity from the plasma membrane to exosomes may play an important role in the regulation of extracellular microenviroments and the maturation of B lymphocytes.


Interaction of CD9 with CD10

In order to identify molecules on pre-B lymphoid cells associated with CD9, we used a previously developed method of fluorescence immunoprecipitation analysis (FIPA) combined with MS identification of antigens [36, 37]. Briefly, Nalm-6 pre-B cells were labeled with rhodamine 6G and lysed in 1% Brij97 buffer; the proteins were precipitated with mAb against CD9 and resolved by 12% PAGE (Fig. 1A, left panel, line 2). All protein bands from the track were excised, in-gel trypsinized, and analyzed by LC-MS/MS. The list of identified proteins is shown in Table 1. Among the proteins coprecipitated with CD9 antigen, CD10 was detected with one of the highest scores. Then, using the same approach, we precipitated CD10 from Nalm-6 lysate to identify the antigens associated with CD10 (Fig. 1A, left panel, line 3; Table 1). CD9 was clearly identified after immunoprecipitation with mAb against CD10. To confirm the CD9–CD10 association, we generated expression plasmids encoding human CD9 and CD10 cDNAs, and transiently cotransfected them into 293T cells. Transfected 293T cells were labeled with rhodamine 6G, and lysed in Brij97 buffer; proteins were precipitated with mAb against CD9, mAb against CD10, or control mAb against CD47. As shown in Fig. 1A (first and third panels), CD9 and CD10 extracted from both cell lines were reciprocally coprecipitated. As expected for tetraspanins, CD9 was associated with numerous proteins. In contrast, the mAb against CD10 visibly coprecipitated only CD9. After transferring proteins from the gel to the blotting membrane, we stained immunoblots with mAbs against CD9 and CD10, and confirmed CD9–CD10 association by western blot (WB) (Fig. 1A, second and fourth panels).

Table 1. MS identification of proteins from Nalm-6 cells coprecipitated with mAb against CD9 or CD10. CALLA, common acute lymphoblastic leukemia antigen
 Protein nameSwiss-Prot accession no.Approximate molecular mass (kDa)Immunoprecipitation with
CD9 (TR12)CD10 (NA4)
mascot scoreNo. of peptidesmascot scoreNo. of peptides
1CD49f (VLA-6) P23229 1262588
2DAAM1 Q9Y4D1 1233529672
3CD315 (CD9P-1) Q9P2B2 98494
4CD29 P05556 88764
5CD10 (CALLA) P08473 854711888732
6CD71 P02786 851725
7CD156c (ADAM 10) O14672 841849
8Annexin A6 P08133 763931123910
9CD92 Q8WWI5 73611
10CD146 (MUC18) P43121 71863
11Protein C14orf21 A8MY76 69432
12CD316 (EWI-2) Q969P0 651135
13CD19 P15391 61933
14CD98 P08195 58562
15Protein disulfide isomerase A6 Q15084 481615
16Galactokinase P51570 42794
17HLA class I P01892 401204772
18Uncharacterized protein C3orf26 Q9BQ75 32462
19Tetraspanin-14 Q8NG11 30635
20Uncharacterized protein C14orf166B Q0VAA2 28624
21HLA class II P04229 282803503
22Uncharacterized protein C22orf13 Q96NT3 27382
23CD81 (TAPA-1) P60033 26841581
24CD9 P21926 261035824
25CD63 P08962 26361
26Tetraspanin-9 O75954 26761
27Protein FAM207A Q9NSI2 251304
28CD179b P15814 235711002
Figure 1.

Metalloprotease CD10 associates with tetraspanin CD9. (A) Reciprocal Co-IP between CD10 and CD9. 293T cells were cotransfected with CD9 and CD10 expression plasmids for 48 h. Transfected 293T or Nalm-6 cells were labeled with rhodamine 6G and lysed in 1% Brij97 buffer. After immunoprecipitation (IP) with corresponding mAbs, proteins were resolved by 12% PAGE and detected by laser scanner. Proteins from lines marked with asterisks were excised for MS analysis Alternatively, nonlabeled cell lysates were immunoprecipitated and then analyzed by WB. (B) CD9–CD10 Co-IP under mild and stringent lysis conditions. 293T cells cotransfected with plasmids encoding CD9 and CD10 were lysed in 1% Triton X-100 or Brij97 detergent in the presence of 1 mm Ca2+ and Mg2+ (left and right rows) or in 1% Brij97 with 2 mm EDTA (middle row). Straight WBs or WB after IP with mAb against CD10 were stained with indicated mAbs. (C) Co-IP of different tetraspanins with CD10. Flag-tagged tetraspanins were coexpressed with CD10 in 293T cells. Proteins from Brij97 cell lysates before and after precipitation with mAb against CD10 were resolved by 12% PAGE, transferred to the blotting membrane, and probed with mAb against Flag or mAb against CD10. The brackets indicate bands of tetraspanin proteins with expected molecular masses. The histogram shows the mean relative Co-IP values (see Experimental procedures for details) from the three independent experiments. All levels of relative Co-IP were statistically different from the control (CD9 Co-IP) (P < 0.01 by Student's t-test).

The interaction of a transmembrane protein with a tetraspanin can be specific and direct, or less strong and less selective for the members of the tetraspanin family. To address the question of how strong the association is between CD10 and CD9, we compared the levels of protein coimmunoprecipitation (Co-IP) in Triton X-100 and Brij97 detergents, with and without divalent cations. As shown in Fig. 1B, the association between CD10 and CD9 was completely disrupted in the stringent detergent Triton X-100, and was preserved in the milder detergent Brij97, even without Ca2+ and Mg2+ (bottom blot). These data were consistent with the observation that most of the protein–protein interactions in tetraspanin web are maintained in Brij96-97 detergent, but also suggested that association of CD9 with CD10 was weaker, than, for example, its association with EWI-F [9].

To determine the specificity of CD10 association with tetraspanins, we constructed expression plasmids encoding cDNAs of human tetraspanins commonly expressed on lymphocytes, such as CD53, CD63, CD81, CD82, and CD231. In all constructs, Flag tag epitope was appended to the N-termini of proteins. A CD9 expression plasmid with an N-terminal Flag tag was used as a control. The CD10 expression vector and one of the plasmids indicated above were cotransfected into 293T cells. Forty-eight hours later, proteins from the cells were extracted in 1% Brij97 lysis buffer, and precipitated with mAb against CD10. Proteins from lysates or after precipitation were resolved by 12% PAGE, and transferred to the blotting membrane; this was followed by staining with mAb against CD10 or Flag. To estimate correctly the levels of tetraspanins precipitated with CD10, we quantified relative Co-IP. For this, we normalized the levels of tetraspanin Co-IP to the levels of tetraspanin expression and CD10 precipitation, and compared the resulting values with the one obtained for CD9, which was set at 1. As shown in Fig. 1C, all tested tetraspanins except CD9 were not coprecipitated or weakly coprecipitated with CD10. These results suggested that CD10 interacts with CD9 with high selectivity.

Mapping domains of CD9 responsible for the association with CD10

As shown above, CD10 metalloprotease interacts with CD9, but not with CD82. Hence, to localize domains in CD9 that mediate interaction with CD10, we constructed chimeras between these two distantly related tetraspanins. Initially, we divided the CD9 sequence into three large parts: from the N-terminus to the second transmembrane domain (TM), TM2; the LEL from the beginning to the CCG motif; and the LEL from the CCG motif to the C-terminus. These regions were swapped with the corresponding parts from CD82. Chimeric molecules were tagged with Flag epitope at the C-terminus to enable detection of them by WB. The resulting chimeras, CD9×82, were designated N, M, and C. The amino acid sequences at the junctions are shown in Fig. S1A. The designed chimeras and CD10 were coexpressed in 293T cells, and the levels of relative Co-IP between CD10 and wild-type or chimeric CD9 were analyzed, as described above and in Fig. 1C. Unfortunately, chimera C and, partially, chimera M were poorly expressed in 293T cells (Fig. S1B), and were predominantly retained in the endoplasmic reticulum (ER) (see Fig. S1C with a confocal image analysis of HeLa cells transfected with plasmids encoding the indicated mutant proteins). The levels of colocalization between the ER and chimeras M and C estimated as Pearson coefficients of correlation (R) were high (0.58 ± 0.12 and 0.32 ± 0.16, respectively). In contrast, chimera N behaved like wild-type CD9; that is, it was associated with CD10, and efficiently expressed and delivered to the cell surface (R = −0.04 ± 0.06 for WT CD9 and R = −0.06 ± 0.03 for chimera N). Interestingly, although chimera M was only partially delivered to the plasma membrane, it remained associated with CD10 (Fig. S1B). Together, these data suggest that the N-terminal half of CD9 up to the CCG motif is dispensable for the association with CD10.

To further localize the CD9 domains mediating association with CD10, we split the C portion of CD9 designated above into the three smaller parts: from the CCG motif to the end of the LEL; TM4; and the C-terminal cytoplasmic tail. These parts were replaced with the similar regions from CD82, generating chimeras called CD9×82 cLEL, TM4, and C-term, respectively. The schematic chimera structures and amino acid sequences at the protein junctions are shown in Fig. 2A. These chimeras were tested by Co-IP with CD10, as described above. As shown in Fig. 2B, mutants cLEL and C-term did not precipitate with CD10. In contrast, the TM4 chimera was precipitated with mAb against CD10 even better than WT CD9. Microscopy analysis (Fig. 2C) revealed that the TM4 and C-term chimeras were efficiently delivered to the plasma membrane in HeLa cells (R = −0.08 ± 0.1 and R = −0.06 ± 0.04, respectively), whereas the cLEL chimera was largely retained in the ER (R = 0.39 ± 0.06). Thus, the results obtained with chimeras TM4 and C-term suggested that TM4 is dispensable and that the C-terminal tail of CD9 is important for the association with CD10. However, the role of the LEL was less evident.

Figure 2.

CD9×CD82 chimeric proteins cLEL and C-term do not interact with CD10. (A) Schematic presentation of CD9×CD82 chimeras. The CD82 sequence of chimera C (see Fig. S1) was split into three additional parts to produce new variants of chimeric tetraspanin. CD9 and CD82 portions of chimeras are colored in black and red, respectively. The amino acid sequences of protein junctions are provided below. (B) Effects of chimerization of CD9 on the association with CD10. The indicated chimeric proteins were coexpressed in 293T cells with CD10, and analyzed in a Co-IP assay, as described for Figs 1C and S1. Data are representative of three independent experiments. **Different from the control at P < 0.01 by Student's t-test (C) Confocal image analysis of CD9×CD82 chimera localization in HeLa cells. HeLa cells were transfected with plasmids encoding either wild type (WT) or chimeric protein. Cells were fixed and permeabilized 16 h after transfection. Tetraspanins (green) were probed with mAb against Flag, and this was followed by staining with anti-mouse Alexa 488. ER (red) was stained by concanavalin A (ConA) conjugated to rhodamine. Representative images are shown. The levels of colocalization between chimeric proteins and ER as the Pearson coefficients of correlation were quantified for individual cells from several slides, and are given in the Results. IP, immunoprecipitation.

Both the LEL and the C-terminal tail of CD9 are involved in association with CD10

The problem with expression of chimeras M, C and cLEL indicates that swapping of a large portion of the LEL between CD9 and CD82 affects protein folding and therefore underestimates the role of the LEL in the association with CD10. To gain additional evidence of the importance of the LEL for the CD9–CD10 interaction, we chimerized CD82 with the full-length LEL and/or the C-terminal domain from CD9 to verify whether it restores the association with CD10. Unlike the CD9 LEL mutants described above, replacement of the intact LEL and/or the C-terminal end in CD82 with those from CD9 did not dramatically reduce the level of protein delivered to the cell surface: R for colocalization with the ER was < 0.2 for any chimeras (Fig. S2A). As shown in Fig. 3A, substitution of the LEL in CD82 with the LEL from CD9 was sufficient to establish the association of mutant CD82 with CD10. Interestingly, the C-terminal tail from CD9 increased the level of association with CD10, when combined with CD9 LEL substitution in CD82. These data are consistent with the results obtained for CD9 LEL chimeras, and demonstrate the important role of the LEL in mediating the interaction with CD10.

Figure 3.

The LEL and the C-terminal tail are important for association of CD9 with CD10. (A) The LEL from CD9 rescues the association of CD82 with CD10. The upper picture schematically shows generated CD82 chimeras. Two different domains of CD82 – the LEL and the C-terminal tail – were swapped with the same regions from CD9 (dashed line) individually or in combination. The sites of chimera junctions between CD82 and CD9 were the same as in Figs S1 and 2A. Chimeras with C-terminal Flag tag were coexpressed with CD10 in 293T cells and analyzed by immunoprecipitation (IP) and WB (bottom blots and graph), as described above. (B) Effects of mutation in the C-terminal tail of CD9 on association with CD10. Amino acids mutated or substituted from CD82 are shown in the upper picture in gray boxes. CD9 mutants nontagged with Flag were analyzed for expression and Co-IP with CD10 (middle blots and bottom graph), as described for other tetraspanin mutants. The graphs in (A) and (B) represent the mean of at least three independent experiments with the standard deviations. **The value is different from the control at P < 0.01 by Student's t-test.

In order to confirm the importance of C-terminal amino acids in CD9 for the association with CD10, we generated a number of nontagged mutants of CD9: substitution of the C-terminal tail from CD82 (82 C-term), mutations of two palmitoylated Cys residues [13] to Ser residues (SS), mutation of three positively charged Arg residues to Ala residues (R-A), and substitution of the three terminal residues Glu-Met-Val with Pro-Lys-Tyr from CD82 (PKY), as previously described [38] (Fig. 3B, upper part). The intracellular localization of generated mutants was studied in HeLa cells (Fig. S2B). Like WT CD9, all mutants were efficiently delivered to the plasma membrane (R for colocalization with the ER was close to zero). Then, we examined the levels of mutant expression and Co-IP with CD10 in 293T cells (Fig. 3B, lower part). The mutation of C-terminal Cys residues, the sites of palmitoylation, or changing the three terminal residues in CD9 did not influence the association of CD9 with CD10. As with the Flag-tagged version of CD9, swapping of the whole C-terminal domain between nontagged CD9 and CD82 dramatically reduced association of the mutant with CD10, suggesting that the C-terminal Flag epitope had no or little effect on CD10 association. Remarkably, the positively charged Arg residues at the C-terminal end of CD9 significantly contributed to this association.

Summarizing the results obtained with the different CD9×CD82 chimeras or mutants, we concluded that both the LEL and the C-terminal domain of CD9 are important for mediating the association of CD9 with CD10.

Expression of CD9 enhances the release of CD10 with exosomes

As most tetraspanins, including CD9, are actively sorted into exosomal membranes, we hypothesized that CD9 will bind and then recruit CD10 into exosomes. To test this hypothesis, we chose the erythromyeloid cell line K562, which expresses the lowest level of endogenous CD9 and is widely used in exosome studies. First, K562 cells were stably transduced with the lentiviral vector pUCHR CD10 IRES neo, and a pool of neomycin-resistant cells were grown. Then, the resulting cell line was transduced with either empty (no gene) or, WT CD9 or the mutant CD9×82 C-term IRES puro cassette, and this was followed by puromycin selection. The levels of CD9 and CD10 expression on the surface of the established cell lines were quantified by fluorescence-activated cell sorting (FACS) (Fig. 4B, upper graph). Proteins from cells and exosomes (see purification in Experimental procedures) were extracted and analyzed with mAbs against CD9, CD10 and tubulin-α (as a loading control) by direct WB or after immunoprecipitation with mAb against CD10. Using a band densitometry tool, we estimated the levels of CD10 expression on cells or exosomes. After setting values from the control samples (empty) at 1, the relative levels of CD10 expression were calculated (Fig. 4A). As shown in Fig. 4A,B, the expression of either WT or mutated CD9 in K562 cells did not alter the levels of CD10 expression. Likewise, mutation in the C-terminal end of CD9 had no influence on the level of its total or surface expression. Both WT and mutated CD9 were recruited into exosomes at similar levels. However, virtually only WT CD9 remained associated with CD10, which was consistent with the results obtained with transfected 293T cells. The coexpression of WT CD9 with CD10 in K562 cells increased the level of CD10 on exosomes five-fold relative to the control (without CD9), and 2.5-fold relative to the cells expressing the CD9×82 C-term mutant. These results indicated that CD10 was actively recruited into exosomes via the association with CD9.

Figure 4.

Association of CD10 with CD9 in K562 cells enhances the release of CD10 with exosomes. (A) WB analysis of CD10 expression in cells and exosomes. K562 cells were stably transduced with CD10 and then with one of the IRES puro cassettes: empty, WT CD9, or mutant CD9. Cells and exosomes from K562 sublines were prepared for WB as described in Experimental procedures. Blots were probed directly with mAb against CD9, CD10 or tubulin-α (loading control), or mAb against CD10 after IP. The levels of CD10 expression relative to the cellular or exosomal controls were semiquantified with a band density tool (histogram below) (B) Cell surface expression (upper graph) and enzymatic activity (low graph) of CD10 on cells and exosomes. The levels of C10 and CD9 cell surface expression were quantified by FACS analysis, and expressed as mean fluorescence intensity (MFI). The values of CD10 peptidase activity on cells and exosomes in relative fluorescence units (RFU) were estimated after incubation with the fluorogenic substrate DAGNPG for 3 h. The averages from four independent experiments with standard deviations are presented in (A) and (B). Data are significantly different at P < 0.01 (**) or at P < 0.05 (*) by Student's t-test. exo, exosomes; IP, immunoprecipitation.

It has been reported that CD9 associates with ADAM10 and ADAM17, members of the ADAM family of metalloproteinases, and regulates their sheddase activity [39, 40]. To verify whether the association of CD9 with CD10 modulates the enzymatic activity of CD10, we quantified the ability of CD10 to cleave the peptide substrate N-dansyl-d-Ala-Gly-p-nitro-Phe-Gly (DAGNPG) in cellular and exosomal fractions from K562 sublines. As shown in Fig. 4B (lower graph), the presence of WT or mutant CD9 did not alter cell-associated CD10 enzymatic activity. In contrast, the level of CD10 activity on exosomes released from CD9-positive cells was 1.6-fold higher than that in the empty sample, and 1.25-fold higher than that of the CD9×82 C-term mutant. These data correlate with the levels of CD10 expression, and suggest that CD9 enhances the enzymatic activity of CD10 indirectly by recruiting more protease molecules into exosomes.

Although the expression of CD9 and CD10 in K562 cells showed a role of CD9 in the segregation of CD10 into exosomes, it is not clear whether this effect occurs in pre-B cells, where the CD9–CD10 interaction is proposed to be functionally relevant. To examine the effect of endogenous CD9 on the release of CD10 from pre-B cells, we stably silenced CD9 in Nalm-6 cells by using two different short hairpin RNA (shRNA) constructs. shRNAs targeting protein P of vesicular stomatitis virus and human CD82 (two variants) were used as controls in knockdown experiments. The efficacy of silencing was confirmed by WB (Fig. 5A) and FACS analysis (Fig. 5B, upper graph) of puromycin-resistant cells. The levels of expression (Fig. 5A) and enzymatic activity (Fig. 5B, bottom graph) of CD10 on cells and exosomes were estimated, as described for K562 cells. As shown in Fig. 5A,B, transduction of Nalm-6 cells with sh9-105 resulted in a five-fold reduction in CD9 expression, a two-fold decrease in CD10 release with microvesicles (Fig. 5A), and a 1.3-fold reduction in CD10 activity on exosomes relative to the shRNA control. CD9 knockdown with sh9-133 was less efficient than with sh9-105. It was also accompanied by decreases in the expression and enzymatic activity of CD10 on exosomes, but the differences from control were not statistically significant. Transduction of cells with either control shRNA or shRNAs against CD82 did not affect the expression or peptidase activity of CD10.

Figure 5.

shRNA knockdown of CD9 in Nalm-6 cells decreases the levels of expression and activity of CD10 on exosomes. (A) Effect of CD9 silencing on CD10 release with exosomes. Nalm-6 cells were stably transduced with the indicated shRNAs. Cells and exosomes were prepared for WB analysis, as described for K562 cells. (B) The levels of cell surface expression (upper histogram) and peptidase activity of CD10 on cells and exosomes (bottom histogram) were examined by FACS and fluorometric analysis, respectively, and as described for Fig. 4. All graphs represent the mean of at least three independent experiments with standard deviations. The differences relative to control are statistically significant at P < 0.01 (**) or at P < 0.05 (*) by Student's t-test. exo, exosomes; MFI, mean fluorescence intensity; RFU, relative fluorescence units.

In summary, using two cell systems, K562 and Nalm-6, we demonstrated the role of CD9 in the enrichment of exosomal vesicles with CD10. The peptidase activity of CD10 correlated with the level of CD10 expression. However, the effect of CD9 on CD10 enzymatic activity was smaller than that on CD10 expression. This difference could reflect a possible nonlinear signal response in the protease assay, as is often observed in assays with a high level of background (see dashed lines in Figs 4B and 5B).


Tetraspanins predominantly interact with each other and with different transmembrane proteins in cis, i.e. laterally. There are few exceptions to this rule. Thus, hepatitis C virus env protein hijacks CD81 to enter the target cells [41, 42], and CD9 serves as a receptor for the soluble pregnancy-specific glycoprotein PSG17 [10, 43]. The majority of specific cis and trans interactions of tetraspanins with their partners are mediated by the LEL, which is the most divergent domain in proteins of this family [44]. Here, we have described a new tetraspanin partner, CD10, that associates with CD9, and have determined that the LEL and the short C-terminal tail of CD9 are involved in this interaction. The LEL of tetraspanins contains constant A, B and E helices, and a variable region comprising helices C and D. We have demonstrated that the LEL from CD9 placed in the context of tetraspanin CD82, which did not bind CD10, is sufficient to maintain the interaction with CD10. The portion of the LEL that includes helices A and B from CD82 (see results for mutant M) did not abolish the interaction of CD9 with CD10. Together, these data suggest that the variable region of the LEL comprising helices C and D are most likely involved in the association with CD10. Moreover, the constant helix E is unlikely to be involved in this association, as the other members of the tetraspanin family tested in the present work weakly interacted with CD10.

Currently, little is known about the functional role of the C-terminal cytoplasmic domain. The presence of tyrosine-based internalization and PDZ-binding motifs at the C-termini of some tetraspanins has led to an assumption of a role in protein trafficking. However, the functional role of these motifs has been clearly demonstrated only for CD63 [45-47]. Although the C-terminal residues REMV in CD9 match with the class III PDZ-binding motif X(D/E)XΦ [48], the function of this motif remains uncertain. Recently, Wang et al. [38] showed that replacing the C-terminal residues EMV in CD9 with PKY from CD82 profoundly affected many functions of CD9, particularly cell adhesion, aggregation, and microvillus formation. In our study, PKY mutation and mutation of C-terminal palmitoylated Cys residues did not reduce the association of CD9 with CD10. However, the finding that mutation of Arg residues or replacement of the C-terminal domain from CD82 affected this interaction suggested that the C-terminal tail is involved in the CD9–CD10 interaction. The real structure of tetraspanins is proposed to be compact, with transmembrane and cytoplasmic domains packed closely [2]. Taking in consideration this point of view, the C-terminal domain of CD9 may influence the highly ordered structure of the LEL, thereby affecting the association of CD9 with CD10. Thus, the LEL can mediate the primary interaction of CD9 with CD10. However, a second site of interaction between cytoplasmic domains of CD9 and CD10 may also exist. Although the relatively selective interaction of CD9 with CD10, which, in addition, involves the variable LEL domain, suggests that this association is probably specific and direct, we cannot rule out the possibility that CD9 interacts with CD10 via other molecular intermediates.

One of the major roles of tetraspanins is to modulate the function of molecules that interact with them. It has been reported that CD9 associates with membrane proteases of the ADAM, γ-secretase and matrix metalloprotease families and regulates their enzymatic activities (reviewed in [17]). In our study, we did not observe CD9-dependent changes in the expression or peptidase activity of CD10 on the cell surface. In contrast, both the expression and activity of CD10 on exosomes were elevated when CD9 was ectopically expressed in K562 CD10-positive cells, or decreased when CD9 was silenced with shRNA in Nalm-6 cells. Although the two parameters of CD10 correlated with each other, in both cell models the effect of CD9 on CD10 activity was smaller than that on CD10 expression. We explain this disparity by the low signal-to-noise ratio of the protease assay, but do not exclude the possibility that CD9 slightly inhibits CD10 peptidase activity on exosomes. Relative to WT CD9, the CD9×82 C-term mutant was approximately two-fold less efficient in the enhancement of CD10 expression and activity on exosomes, but was nevertheless more active than the negative control. This intermediate effect may result from the residual association of this mutant with CD10, which can be seen on Co-IP blots (Figs 3B and 4A) after overexposure (data not shown). Thus, CD9 as such had minimal if any effects on the peptidase activity of CD10, whereas it significantly contributed to the sorting of CD10 into exosomes. Our data are consistent with the many observations that tetraspanins, including CD9, are enriched in limiting membranes of MVBs and exosomes, where they participate in intracellular protein sorting and protein release with exosomes [49].

As a transmembrane protease, CD10 cleaves soluble peptide mediators at the cell surface. The function of CD10 outside cells is poorly understood. Both negative and positive roles of CD10 expression in metastatic invasion of different malignant cells have been ascribed (reviewed in [34]), but the mechanisms, such as CD10-mediated cleavage of extracellular matrix proteins or matrix-bound peptides, have not been provided. Both CD10 and CD9 were initially described as markers of B-cell maturation, as they were expressed at high levels on early lymphocytes and pre-B lymphocytes, but not on peripheral blood B lymphocytes [35]. However, the physiology of this phenotype conversion is obscure. In light of our findings, we hypothesize that expression of CD9 on pre-B cells helps to release CD10 with exosomes to exert its enzymatic activity outside of plasma membrane area. This remote function of CD10 can modify extracellular matrix microenviroments more efficiently than plasma membrane CD10 facilitating migration of immature B cells to the different hematopoetic zones or even release of cells into the blood circulation. Overall, the regulation of pre-B-cell migration and maturation could be a complex process, and differ from metastatic invasion of malignant cells. It would be very interesting to expand our research to the area of CD10 physiology, and determine the mechanisms by which CD10 is involved in B-cell maturation.

In summary, we have described a new interaction between the tetraspanin CD9 and the membrane endopeptidase CD10. Using mutagenic analysis, we determined that both the LEL and the C-terminal tail of CD9 are involved in CD10 association. Functionally, this association manifested in the enhanced release of CD10 peptidase activity with cellular exosomes. It remains unknown which portion of CD10 and which residues in the hypervariable region of CD9 are critical for the interaction of these two molecules. Further in-depth mutagenic analysis will provide new insights into the precise mechanism of this molecular interaction.

Experimental procedures

Cell cultures and reagents

The human cell lines Nalm-6, K562, 293T and HeLa were obtained from the ATCC. Nalm-6 and K562 cells were maintained in RPMI-1640 medium containing 10% fetal bovine serum, 2 mm glutamine, 100 U·mL−1 penicillin, and 0.1 mg·mL−1 streptomycin. 293T and HeLa cells were grown in DMEM containing 10% fetal bovine serum, 50 μm β-mercaptoethanol, 2 mm glutamine, 100 U mL−1 penicillin, and 0.1 mg mL−1 streptomycin. Human CD9 clone TR12, CD10 clone NA4, CD47 clone He3 and CD82 clone MC8D12 mAbs were from Sorbent (Moscow, Russia). mAb clone M2 against Flag epitope was purchased from Sigma (Saint Louis, MO, USA). mAb Tu-01 against tubulin-α was obtained from Exbio (Praha, Czech Republic). Secondary antibodies were goat anti-mouse Alexa Fluor 488 (Molecular Probes, Eugen, OR, USA), and mouse TrueBlot horseradish peroxidase-conjugated anti-mouse IgG (eBioscience, San Diego, CA, USA). Concanavalin A tetramethylrhodamine conjugate was purchased from Molecular Probes. The fluorogenic substrate DAGNPG was obtained from Sigma.

Plasmids and transfections

The human CD81 and CD82 pCMVpA expression plasmids have been described previously [50]. The human CD9 and CD10 cDNAs were amplified from the total RNA from the pre-B-cell line Nalm-6 by RT-PCR. The pOTB7 and pCMV-SPORT6 plasmids encoding human CD53, CD63 and CD231 cDNAs were purchased from Open Biosystems (Huntsville, AL, USA). The cDNAs of tetraspanin genes were cloned either into the mammalian expression plasmid pCMVpA or into the lentiviral expression vector pUCHR IRES puro, which was derived from pUCHR IRES neo [51] by replacing the puromycin resistance gene with the neomycin resistance gene at the 5′-BstXI and 3′-PspXI cloning sites. The CD10 cDNA was cloned into the lentiviral expression plasmid pUCHR IRES neo, using NheI/XmaI restriction sites. Gene fusions and mutations of CD9 and CD82 were performed by standard PCR. All clones were verified by nucleotide sequencing. The C-terminal Flag tag was appended to the WT or chimeric CD9 via a Gly-Ala-Gly linker. The N-terminal Flag epitope was fused with the second (for all tetraspanins except CD9) or third (for CD9) residues through a Gly-Ala-Gly-Ala-Ser linker. For the stable knockdown of CD9 and CD82 expression, we used the pGIPZ mir30-based lentiviral vector from Open Biosystems. Two pGIPZ plasmids with target sequences for the CD9 ORF at positions 105 (CCGATTCGACTCTCAGACCAA) (sh9-105) and 133 (TTCGAGCAAGAAACTAATAAT) (sh9-133) and two pGIPZ vectors targeting the CD82 ORF at positions 11 (CCTGTATCAAAGTCACCAAAT) (sh82-11) and 42 (CCTCTTCAACTTGATCTTCTT) (sh82-42) were used to knock down the expression of CD9 and CD82, respectively. pGIPZ plasmid with the shRNA target sequence CGATAATATAACTGCAAGATT for the phosphoprotein of human respiratory syncytial virus was used as a control in knockdown experiments. For the transient gene expression, 293T and HeLa cells were transfected with pCMVpA or pUCHR expression plasmid by use of Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) or FuGENE 6 (Roche Applied Science, Mannheim, Germany) transfection reagent, respectively, and according to the manufacturer's instructions. To generate stable cell lines, K562 or Nalm-6 cells were infected with virus-like particles derived from 293T cells, which were cotransfected with the following three plasmids: transfer vector pGIPZ or pUCHR IRES neo (puro), HIV-1 packaging plasmid pCMVΔ8.2R, and pCMV-VSVG plasmid expressing Env G from vesicular stomatitis virus, as described previously [51]. Three days after infection, a pool of positive clones were selected by growing cells for 2 weeks in the media with geneticin (Gibco, Grand Island, NY, USA) 800 μg·mL−1 and/or puromycin (Sigma) 1 μg·mL−1. The efficiency of gene expression or knockdown in the selected cells was controlled by FACS analysis.

Immunoprecipitation, WB and MS analysis

FIPA for the detection of proteins in the gel has been described previously [36]. Briefly, 5 × 107 Nalm-6 cells or 1 × 107 transfected 293T cells were pelleted, washed twice with NaCl/Pi, and labeled with 0.3 mg of rhodamine 6G (Syntol, Moscow, Russia) in 1 mL of buffer containing 150 mm NaCl and 10 mm NaHCO3 (pH 8.0) for 20 min. The reaction was stopped by adding 10 mm glycine in NaCl/Pi. Cells were washed with NaCl/Pi again, and extracted in 1 mL of ice-cold lysis buffer containing 1% Brij97 (Sigma), 1 mm CaCl2, 1 mm MgCl2, 150 mm NaCl, 10 mm Tris (pH 7.5), and protein inhibitor cocktail (Complete Mini EDTA-free; Roche Applied Science). If indicated, 1% Brij97 and divalent cations were replaced with 1% Triton X-100 and 2 mm EDTA respectively. After 1 h of extraction at 4 °C, insoluble material was removed by centrifugation at 14 000 g for 10 min. Protein A/G PLUS-agarose beads (Santa Cruz Biotechnology) were preloaded with isotype control IgG mAbs (to preclear lysates at 4 °C for 1–2 h), and with mAbs against CD10 or CD9 (for specific immunoprecipitation). Cell extracts were immunoprecipitated at 4 °C overnight. After rinsing four times with ice-cold lysis buffer, immune complexes were eluted by heating at 80 °C in SDS sample loading buffer, resolved by 12% SDS/PAGE, and analyzed with a laser scanner-based Molecular Imager FX (Bio-Rad, Hercules, CA, USA). Alternatively, unlabeled proteins resolved on PAGE directly or after immunoprecipitation were transferred to Immobilon membranes (GE Healthcare, Piscataway, NJ, USA). Membranes were blocked with 5% nonfat milk in NaCl/Pi with 0.02% Tween (NaCl/Pi-T), probed with primary antibodies, washed with NaCl/Pi-T, and developed with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling, Danvers, MA, USA) for direct immunoblots or with TrueBlot horseradish peroxidase-conjugated antibodies (eBioscience) for immunoprecipitates. Blots were washed again, and immunoreactive bands were detected with Immobilon Western reagent (Millipore, Billerica, MA, USA) on a Molecular Imager ChemiDoc XRS (Bio-Rad). The Volume Analysis tool in quantity one software was used to measure the chemiluminescence intensity of bands on immunoblots. As mutations in CD9 often affected not only coprecipitation with CD10, but also its expression and sometimes the expression of CD10, we calculated the relative Co-IP value. For that, the amount of mutant protein recovered after immunoprecipitation with CD10 was normalized to the level of its expression and to the amount of CD10 immunoprecipitation from the same sample. The resulting value was divided by that obtained for WT tetraspanin. The following formula describes the calculation of the relative Co-IP value (R): R = (DXcoIP/DXexpr/DCD10IP)/(DWcoIP/DWexpr/DCD10IP), where DX is a band density measured for the mutant protein, and DW is the band density of WT CD9 (Figs 2B and S1B) or CD82 (Fig. 3A). R-values for different tetraspanins (Fig. 1C) relative to CD9 were calculated, as indicated for mutants. For identification of proteins by MS, all visible protein bands in the lane were excised, in-gel digested with trypsin, and analyzed by reverse-phase nano-LC-MS/MS with an Agilent 1100 nanoflow LC system coupled to an Agilent 1100 SL Series MSDTrap (Agilent Technologies, Alpharetta, GA, USA). Protein identification was performed with mascot software, and all tandem mass spectra were searched against the NCBI nonredundant database.

Exosome purification and estimation of endopeptidase activity

Exosomes were isolated by differential centrifugation, as previously described [52]. Briefly, 5 × 105 K562 cells, or 2 × 106 Nalm-6 cells were cultured in 3 mL of exosome-free RPMI growth medium for 48 h. Cells were pelleted at 300 g for 10 min, and washed with NaCl/Pi once. The supernatants were centrifuged at 10 000 g for 20 min to remove cellular debris. The remaining exosomes were pelleted at 100 000 g for 2 h. To measure the endopeptidase activity of CD10 [53], cellular and exosomal pellets were resuspended in 200 μL of 50 mm Tris buffer (pH 7.5) containing 150 mm NaCl and 100 μm of fluorogenic substrate DAGNPG (Sigma). Reaction was performed for 3 h at 37 °C, and stopped by adding 800 μL of ice-cold NaCl/Tris buffer supplemented with 5 mm EDTA. The samples were centrifuged at 300 g for 5 min, and the levels of substrate conversion in supernatants were quantified immediately by using a Hitachi F-2000 spectrofluorometer and the following settings: excitation wavelength, 342 nm; and emission wavelength, 562 nm. The ratios of CD10 activity in exosomes to those in cells were calculated and compared between different sublines of K562 cells. In parallel, cells and exosomes were lysed in 1% Brij97 lysis buffer, loaded proportionally on the gel, and resolved by 12% PAGE. The levels of CD10 expression in cells and exosomes were estimated by WB as described in the previous section.

Immunofluorescence microscopy and FACS analysis

One day before transfection, HeLa cells were seeded on coverslips in 24-well plates in complete medium. Cells were transfected overnight, washed with NaCl/Pi, and fixed with 4% paraformaldehyde (Sigma) in NaCl/Pi for 15 min. After being washed with NaCl/Pi, cells were permeabilized in NaCl/Pi solution containing 0.1% saponin (Sigma) and 1% fetal bovine serum for 30 min. WT or chimeric tetraspanin proteins were stained sequentially with mAb against Flag and secondary antibody (Alexa 488 anti-mouse). The ER was stained with rhodamine-labeled concanavalin A in permeabilization solution without serum. After washing, coverslips were transferred to the slides and maintained in Dako Cytomation Fluorescent Mounting Medium. Slides were analyzed with a Nikon Eclipse TE2000 confocal microscope. Image data were processed with fiji image analysis software. Colocalization of tetraspanins with the ER was estimated through Z-stacks of individual cells and from different areas of the samples as the Pearson coefficient of correlation (1.0 is full colocalization, 0 is no colocalization, and −1.0 is full exclusion). Immunofluorescent staining of cells with primary mAb and secondary phycoerythrin-labeled antibody (SantaCruz Biotechnology) was performed on live suspension cells according to a standard protocol. Samples were analyzed with a FACScan flow cytometry instrument (Becton Dickinson, San Diego, CA, USA) and cell quest software.


This work was partially supported by the Russian Foundation for Basic Research (RFBR grants No. 09-04-00426-a and No. 09-04-00318-a). We thank A. Sapozhnikov and E. Kovalenko (Shemyakin and Ovchinnikov Institute of Bioorganic Chemistry, Moscow, Russia) for assistance with confocal image analysis.