The tetraspanins CD151 and Tspan8 are essential exosome components for the crosstalk between cancer initiating cells and their surrounding.

Tspan8 and CD151 are metastasis-promoting tetraspanins and a knockdown (kd) of Tspan8 or CD151 and most pronounced of both tetraspanins affects the metastatic potential of the rat pancreatic adenocarcinoma line ASML. Approaching to elaborate the underlying mechanism, we compared ASMLwt, -CD151kd and/or Tspan8kd clones. We focused on tumor exosomes, as exosomes play a major role in tumor progression and tetraspanins are suggested to be engaged in exosome targeting. ASML-CD151/Tspan8kd cells poorly metastasize, but regain metastatic capacity, when rats are pretreated with ASMLwt, but not ASML-CD151kd and/or -Tspan8kd exosomes. Both exosomal CD151 and Tspan8 contribute to host matrix remodelling due to exosomal tetraspanin-integrin and tetraspanin-protease associations. ASMLwt exosomes also support stroma cell activation with upregulation of cytokines, cytokine receptors and proteases and promote inflammatory cytokine expression in hematopoietic cells. Finally, CD151-/Tspan8-competent exosomes support EMT gene expression in poorly-metastatic ASML-CD151/Tspan8kd cells. These effects are not seen or are weakened using ASML-CD151kd or -Tspan8kd exosomes, which is at least partly due to reduced binding/uptake of CD151- and/or Tspan8-deficient exosomes. Thus, CD151- and Tspan8-competent tumor exosomes support matrix degradation, reprogram stroma and hematopoietic cells and drive non-metastatic ASML-CD151/Tspan8kd cells towards a motile phenotype.


INTRODUCTION
Tumors are a heterogeneous mixture of cells, where a small population of so called cancer-initiating cells (CIC) [1,2] is supposed to account for tumor progression including epithelial-mesenchymal transition (EMT) and metastatic settlement and growth [3.4]. With the description of the phenomenon of a premetastatic niche [5] the question arose on the factors that modulate selected distant organs to facilitate disseminated tumor cell embedding [6]. Similar, in concern on the shift towards EMT, the question remains to be answered, why preferentially cells at the tumor rim become engaged and which trigger they receive [7,8]. Particularly for the establishment of a premetastatic niche strong evidence is accumulating that exosomes are the major actors [9][10][11].
Exosomes, 40-100nm membrane vesicles of endocytic origin [12], are secreted by many cell types, and abundantly by tumor cells [13]. Exosomes are suggested to be the most important intercellular communicators [14,15]. They are defined by size, buoyant density, lipid composition, and protein markers [16]. Their homogeneous size is a major criterion to differentiate from size-variable apoptotic blebs, microparticles and microvesicles [17]. Exosomes are also characterized by a panel of constitutive markers [18], which relate to their origin from endocytosis / early endosomes and include, beside others, components of endocytosis prone membrane domains and of the fission, scission and vesicular transport machineries [19]. One type of endocytosis www.impactjournals.com/oncotarget prone membrane microdomains, tetraspanin-enriched microdomains (TEM) [20], apparently play a major role, as tetraspanins are most widely used to characterize extracellular vesicles as exosomes [21,22]. Furthermore, the net which tetraspanins form between themselves and a multitude of associating molecules is maintained during internalization and intracellular traffic such that tetraspanin webs are recovered in exosomes [23,24]. This feature becomes of particular interest for tetraspanins that are engaged in metastasis formation, which is well known for CD151 and Tspan8 [25,26], the latter also being a CIC marker in pancreatic adenocarcinoma [27]. It is essential to mention that exosomes also contain mRNA and miRNA [16,28] and that all components of exosomes, lipids, proteins, mRNA and miRNA are function competent [29]. Furthermore, exosomes are found in all body fluids [30,31] and can bind to or be taken up by selected target cells, where exosomal tetraspanin webs are involved in target cell selection [32].
We recently uncovered by a CD151 knockdown (CD151 kd ) in a rat pancreatic adenocarcinoma that CD151 promotes adhesion, which is overcome by CD151associated MMPs degrading the extracellular matrix [51]. Tspan8 by associating with α6β4 [52,53] contributes to paxillin and FAK activation and supports motility [51]. Accordingly, both ASML-CD151 kd and ASML-Tspan8 kd cells displayed reduced metastatic capacity, the deficits suggesting supplementary activities of the two tetraspanins. To substantiate this assumption, we generated double ASML-CD151/Tspan8 kd clones, which poorly metastasize. Based on this finding, the known contribution of exosomes to the metastatic process [9][10][11] and the functional importance of exosomal tetraspanins [25], we controlled for the contribution of exosomal CD151 and Tspan8 in tumor progression. ASML-CD151/Tspan8 kd exosomes do not support metastatic settlement, they have severe defects in extracellular matrix (ECM) degradation, poorly stimulate chemokine and chemokine receptor expression and are not generating an inflammatory milieu. Finally, only ASML wt exosomes support EMT in poorly metastatic ASML-CD151/Tspan8 kd cells.

RESULTS
We recently described that CD151 and Tspan8 distinctly, but supplementarily support metastasis, where Tspan8 predominantly accounts for motility and CD151 for adhesion and matrix degradation [51]. We here evaluate the contribution of exosomal tetraspanins to tumor progression, using as additional control an ASML-CD151/Tspan8 kd line.

A CD151 / Tspan8 knockdown is accompanied by loss in metastatic tumor growth.
ASML-Tspan8 kd cells were transfected with CD151 shRNA, selecting clones 16 and 24 for further studies ( Figure 1A). After intrafootpad (ifp) application, the double knockdown clones hardly grew in draining nodes and only 1 of 5 rats developed lung metastasis. As described [51], the survival time of rats receiving ASML-CD151 kd or ASML-Tspan8 kd cells was prolonged and the tumor burden in LN and lung was reduced compared to ASML wt -bearing rats, but did not differ between ASML-CD151 kd -versus ASML-Tspan8 kd -bearing rats. Instead, the survival time and the metastatic burden of ASML-CD151/ Tspan8 kd bearing rats differed significantly from that of ASML-CD151 kd or -Tspan8 kd bearing animals ( Fig.1B-1E). Lymph nodes (LN) and lung of the one rat developing lung metastasis after receiving ASML-CD151/Tspan8 kd cells displayed weak CD151 expression that, however, did not exceed the level of normal lung tissue (Fig.1F). Immunohistology staining for CD44v6, which is highly expressed in ASML cells, confirmed the stability of the Tspan8 kd and/or CD151 kd in vivo (Suppl. Fig.1).
Exosomal CD151 and Tspan8 support metastatic settlement ASML wt exosomes are recovered in all lymphoid organs 48h after intravenous (iv) application. Recovery of ASML-CD151 kd exosomes is only reduced in LN. Recovery of ASML-Tspan8 kd and -CD151/Tspan8 kd (D) mean survival time, indicating significant differences compared to ASML wt -bearing rats and between ASML-CD151/Tspan8 kd versus -CD151 kd or Tpsan8 kd bearing rats: s; (E) No of rats with small or large LN metastasis and of rats with no, few or >1000 lung metastases; significant differences compared to ASML wt -bearing rats: *, and between ASML-CD151/Tspan8 kd versus -CD151 kd or Tpsan8 kd bearing rats: s; (F) recovery of CD151 and Tspan8 in lung and LN lysates of control and tumor-bearing rats. (G) Rats (3/group) received a single injected of dye-labeled exosomes, iv. Rats were sacrificed after 48h; (H) rats (3/group) received three injected of dye-labeled exosomes in 3d intervals, ifp, and were sacrificed 48h after the last injection; (G,H) lymphatic organs were excised and the recovery of dye-labeled cells (exosome uptake) was evaluated by flow cytometry. The mean±SD of dye-labeled cells is shown; significant differences to the uptake of ASML wt exosomes: *; (I) Rats (5/group) received 1x10 6 ASML-CD151/Tspan8 kd cells ifp and starting at day -6 in 3d intervals, 100µg exosomes, ifp. Rats were scarified after 21d. Recovery of tumor cells in draining LN, lung and BM was evaluated by flow cytometry after staining for the ASML marker C4.4A; the mean No±SD of tumor cells / 10 3 cells is shown; significant differences to ASML-CD151/ Tspan8 kd bearing rats: *. A CD151 kd or a Tspan8 kd retards tumor growth. ASML-CD151/Tspan8 kd cells rarely metastasize. ASML-CD151 kd and/or ASML-Tspan8 kd exosomes are poorly recovered in lymphoid organs, which is accompanied by ASML-Tspan8 kd exosome retention at the injection site. exosomes is reduced in bone marrow (BM), peritoneal exudate (PEC) and lung. Instead more exosomes are retained in the blood (Fig.1G), which could indicate a requirement for Tspan8 to leave the blood stream. After repeated ifp application, recovery in lymphoid organs, lung and liver was reduced in rats receiving ASML-CD151 kd and/or -Tspan8 kd exosomes. Recovery of ASML-Tspan8 kd and -CD151/Tspan8 kd exosomes being particularly poor in the blood (Fig.1H), confirms the Tspan8 engagement in crossing the blood barrier. Counterstaining with leukocyte markers revealed, as described [52], that all leukocyte subpopulations, but most pronounced Mφ and DC take up exosomes. The uptake of ASML-CD151 kd and -Tspan8 kd exosomes is slightly and that of -CD151/Tspan8 kd exosomes is more severely impaired, which accounts for all leukocyte subpopulations. Notably, all leukocytes that uptake ASML exosomes are CD53 + , which suggests a particular engagement of CD53 in exosome uptake by hematopoietic cells of the rat (Supp. Fig.3).
To obtain a hint, whether exosomal CD151 and Tspan8 affect premetastatic organ preparation, rats receiving poorly metastatic ASML-CD151/Tspan8 kd cells intrafoodpad (ifp) were pretreated with ASML wt , -CD151 kd or -Tspan8 kd exosomes. Exosome application (200µg/rat, ifp) was repeated every 3rd day. Rats were sacrificed 14 days after tumor cell application. The presence of ASML-CD151/Tspan8 kd cells was evaluated by flow cytometry in draining LNs, lung and BM. Except in rats receiving ASML wt exosomes, tumor cells were hardly recovered particularly in lung and BM, indicating that both exosomal Tspan8 and CD151 contribute to niche preparation (Fig.1I).
Nonetheless, CD151-and Tspan8-competent exosomes most efficiently restored metastatic settlement of poorly metastatic ASML-CD151/Tspan8 kd cells. Thus, we asked for the contribution of exosomal tetraspanins. Mean±SD (3 assays) of the % stained cells and exosomes; significant differences to ASML wt cells and exosomes: * and (B,D) WB. (E) Exosome lysates of ASML wt and -CD151 kd and/or Tspan8 kd were precipitated with the indicated anti-integrin antibodies and were blotted with anti-CD151 and anti-Tspan8. (F,G) Dye-labeled exosomes were seeded on matrix-coated plates. Where indicated exosomes were preincubated with antibodies. After 2h incubation, plates were washed and remaining fluorescence was evaluated in an ELISA reader. The mean percent±SD (triplicates) of bound exosomes is shown, (F) differences to ASML wt exosomes: *; (G) differences by antibody preincubation: *. Expression of additional tetraspanins and of integrins is not affected in ASML-CD151 kd and/or -Tspan8 kd cells. Instead, integrin expression in exosomes is reduced corresponding to the associating tetraspanin. Exosomal CD151 and Tspan8 slightly affect exosome adhesion to matrix proteins. www.impactjournals.com/oncotarget Exosomal CD151 and Tspan8 account for matrix remodelling ASML cells express the tetraspanins CD151, Tspan8, CD9 and CD81. These tetraspanins are recovered in exosomes and the deficit of exosomal CD151 and/or Tspan8 has no significant impact on expression of the remaining tetraspanins in cells and exosomes ( Fig.2A,2B). The major integrins in ASML are CD49c/(CD29), CD49f/ CD104 that is recognized by the B5.5 antibody and, though less pronounced CD11b. Integrin expression is not affected in ASML-CD151 kd , -Tspan8 kd and -CD151/Tspan8 kd cells, but CD11b and CD49c expression is reduced in ASML-CD151 kd and -CD151/Tspan8 kd exosomes and CD104 expression is reduced in ASML-Tspan8 kd and -CD151/ Tspan8 kd exosomes (Fig.2C,2D). Reduced CD11b and CD49c expression in CD151 kd exosomes correlates with the preferential association of CD151 with CD11b and CD49c; strongly reduced recovery of CD104 in ASML-Tspan8 kd and -CD151/Tspan8 kd exosomes correlates with co-immunoprecipitation of CD104 with Tspan8 ( Fig.2E), also described for ASML cells [53,54].
Despite reduced integrin expression, adhesion of exosomes to matrix proteins was not severely affected by a CD151 or Tspan8 deficit, but adhesion of ASML-CD151/Tspan8 kd exosomes was mitigated (Fig.2F). Blocking exosome adhesion by anti-CD49c or anti-CD104 confirmed a pronounced contribution of CD151 to CD49cpromoted adhesion and of Tspan8 to CD104-supported in ASML wt , -CD151 kd and/or Tspan8 kd cells and exosomes; mean±SD (3 assays) of the percent stained cells and exosome-coated beads; significant differences to ASML wt cells/exosomes: *; significant differences between cells and exosomes: *; (C) protease recovery in ASML wt and -CD151 kd and/or -Tspan8 kd cells and exosomes as revealed by WB; (D) Coimmunoprecipitation of exosomal CD151 and Tspan8 with proteases and (E) of MMP2 and MMP9 with CD151 and Tspan8; (F) recovery of proteases in light and heavy sucrose density fractions in ASML wt , -CD151 kd and/or -Tspan8 kd exosomes; (G) gelatin (zymography) degradation by ASML wt , -CD151 kd and/or Tspan8 kd exosomes and exosome-depleted conditioned medium; (H) native LnStr and LuFb matrix degradation (WB) by ASML wt , -CD151 kd and/or Tspan8 kd exosomes and (I) inhibition of exosome-mediated matrix protein degradation by TACE, MMP2 and MMP9/MMP13 inhibitors (WB). Recovery of proteases in exosomes is mostly dictated by the association with CD151 and/or Tspan8 such that in the absence of CD151 mostly MMP2 and in the absence of Tspan8 mostly MMP9 expression / activity are strongly reduced. Exosomal TACE activity apparently depends on both Tspan8 and CD151. Reduced exosomal protease recovery has consequences on matrix protein, matrigel and native matrix degradation. adhesion. Though differences did not reach statistical significance in all instances, the failure to block ASML-CD151/Tspan8 kd exosome adhesion by either anti-CD49c or anti-CD104 confirms the importance of the tetraspaninassociated integrins in exosome targeting (Fig.2G).
Exosome binding to matrix proteins is accompanied by matrix degradation, which is affected in ASML-CD151 kd , -Tspan8 kd and -CD151/Tspan8 kd exosomes. ASML-CD151 kd exosomes did not degrade coll I and FN and poorly degraded coll II and coll IV. Coll II, FN and more pronounced LN332 degradation was reduced in the presence of ASML-Tspan8 kd compared to ASML wt exosomes (Fig.3A). Reduced matrix degradation corresponds to reduced protease expression in exosomes. A CD151 kd mostly affects expression of MMP2 and MMP3, a Tspan8 kd affects MMP9, but promotes MMP14 and TACE expression. Furthermore, while exosomal MMP9 and MMP14 expression corresponds to cellular expression, MMP2, CD13 and TACE expression is reduced in ASML-Tspan8 kd exosomes (Fig.3B,3C). Notably, recovery in exosomes correlates with the tetraspanin association. Thus, after mild exosome lysis, CD13 (mostly), MMP9 and TACE co-immunoprecipitate with Tspan8, but not CD151. MMP2 and MMP14 co-immunoprecipitate with CD151, but not Tspan8 and exosomal MMP13 does not co-immunoprecipitate with CD151 or Tspan8, which was confirmed in the reverse setting in cell and exosome lysates for MMP2 and MMP9 (Fig.3D,3E). Furthermore, MMP2 was not recovered in light sucrose gradient fractions of ASML-CD151 kd exosomes, and MMP9, CD13 and TACE were not or poorly recovered in light sucrose gradient fractions of ASML-Tspan8 kd exosomes (Fig.3F). Thus, changes in the exosomal versus the cellular protease profile are due to the selective association of CD151 with MMP2 and MMP14 and of Tspan8 with CD13, MMP9 and TACE. Zymography confirmed the absence of active MMP2 in CD151 kd exosomes, strongly reduced MMP9 activity in ASML-Tspan8 kd exosomes and poor recovery of active MMP2 and MMP9 in CD151/Tspan8 kd exosomes (Fig.3G). The impact of CD151-and/or Tspan8-associated exosomal proteases also was apparent in degradation of the natural matrix of LnStr and LuFb. ASML-CD151 kd exosomes did not degrade coll I and LN111 and ASML-Tspan8 kd exosomes did not degrade LN332 (Fig.3H). Confirming the contribution of Tspan8-associated MMP9 and MMP13 to coll II, FN and LN332 modulation, degradation was abolished in the presence of an MMP9/ MMP13 inhibitor, while degradation of coll I, coll IV and FN, which require CD151-associated MMP2, was not seen in the presence of an MMP2 inhibitor. LN332, coll IV and FN degradation was also impaired in the presence of the TACE inhibitor TAPI (Fig.3I).
Exosomal tetraspanin-integrin and tetraspaninprotease complexes also modulate the host matrix. Thus, lymph node stroma (LnStr), lung fibroblasts (LuFb) and rat endothelial cells (RAEC) adhere more strongly to their own matrix, when modulated by ASML wt exosomes. Pronounced adhesion is not seen after pretreatment with ASML-CD151 kd or -CD151/Tspan8 kd exosomes (Fig.4A). Adhesion of stroma cells to the ASML wt exosomemodulated matrix is accompanied by actin reorganization [51], which is also seen, though weaker after treatment with ASML-CD151 kd exosomes (Fig.4B). Furthermore, ASML wt and -CD151 kd exosome-treated host matrices slightly promoted host cells motility, even though the effect was far weaker than on tumor cell motility (see Suppl. Fig.2B). Matrices treated with exosomes from Tspan8 kd or CD151/Tspan8 kd exosomes did not significantly affect host cell motility (Fig.4C). Finally, the deficit of protease activity in CD151 kd and/or Tspan8 kd exosomes had severe consequences on host cell invasiveness such that LnStr and RAEC poorly invaded matrigel, invasiveness being about 5-fold reduced in matrigel pretreated with ASML-CD151/Tspan8 kd exosomes compared to ASML wt exosomes (Fig.4D).

Exosomal CD151 and Tspan8 and host cell modulation
Based on previous work demonstrating a major contribution of ASML exosomes to stroma cell protease and adhesion molecule as well as chemokine and chemokine receptor expression [63], we asked for a selective contribution of exosomal CD151 and Tspan8 in LnStr, LuFb and RAEC as well as LNC and BMC modulation. Exosomes were cocultured with LnStr, LuFb, RAEC and freshly harvested BMC and LNC for 48h.
We first controlled the impact of exosomal CD151 and Tspan8 on protease expression. Only ASML wt exosome uptake by LuFb and more pronounced by LnStr promoted TACE, MMP14, TIMP1 and TIMP2 expression. ADAMTS1, ADAMTS5 and uPA upregulation appeared largely CD151-dependent. Expression of MMP2 and MMP9 was not affected by short term in vitro coculture (Fig.5A,5B). However, distinct to the short term in vitro coculture, repeated ASML exosome application promoted TACE, but also MMP2 and MMP9 upregulation that was strongest in ASML wt exosome-treated rats. ASML-CD151/ Tspan8 kd exosomes did not induce MMP9 and TACE upregulation; MMP2 expression was hardly increased in ASML-CD151 kd exosome-and MMP9 and TACE expression in -Tspan8 kd exosome-treated rats (Fig.5C). These findings fitted to the appearance of LN metastases with poor recovery of MMP2 in ASML-CD151 kd tumors, poor recovery of MMP9 in both ASML-CD151 kd and ASML-Tspan8 kd tumors and very few ADAM17 + cells in ASML-Tspan8 kd tumors. The reduced protease recovery was accompanied by a dense coll IV and LN332 matrix in ASML-CD151/Tspan8 kd tumors (Suppl. Fig.4). Immunohistology of LN sections from ASML wt exosometreated rats additionally showed upregulated expression of CD49c and CD49d that was not seen in ASML-CD151/ Tspan8 kd exosome-treated rats. Upregulation of α6β4 was only seen in ASML wt and ASML-CD151 kd exosometreated rats (Fig.5D).
Thus, ASML exosomes exert a strong impact on host cell protease expression that partly depended on exosomal CD151 or Tspan8.
Premetastatic niches are frequently characterized by changes in chemokines and their receptors [64], where we controlled for a selective impact of exosomal CD151 and Tspan8 in LnStr, RAEC and LuFb after 48h coculture with ASML wt , -CD151 kd and/or -Tspan8 kd exosomes. LnStr, LuFb and RAEC did not uniformly respond to ASML exosome treatment. Only SDF1 and VEGFR1 were upregulated in all three cell populations after treatment with ASML wt exosomes. However, SDF1 depended more on CD151 than Tspan8, whereas VEGFR1 stimulation required CD151 and Tspan8. CXCR4 and VEGFR3 expression, stimulated in LnStr and LuFb, respectively in and RAEC were seeded on their corresponding untreated or exosome-treated matrix; mean±SD of adherent cells; significant differences compared to the native matrix: *; (B) confocal microscopy of actin cytoskeleton organization in LnStr and LuFb depending on exosome treated stroma (scale bar: 10µm); (C) videomicroscopy of LuFb and LnStr that were seeded on their untreated or ASML exosome-treated matrix. The relative migration of cells during 12h incubation (mean of 20 individual cells) and a representative example are shown; significant differences in migration due to ASML exosome treatment are indicated; (D) Matrigel penetration and invasion of LnStr and RAEC in dependence on CD151-and/or Tspan8-competent exosomes; the mean±SD (triplicates) of penetrating cells; significant differences to untreated matrigel: *; significant difference between ASML-CD151/Tspan8 kd versus -CD151 kd or -Tspan8 kd exosomes: s; and representative examples of invasion (scale bar: 250µm). Exosomal CD151 and Tspan8 slightly affect host cell adhesion to matrix proteins, but support stroma cell motility and invasiveness, which might be promoted by exosomal proteases. www.impactjournals.com/oncotarget Flow cytometry analysis of protease expression in LnStr and LuFb after coculture with ASML wt , -CD151 kd and/or -Tspan8 kd exosomes; mean percent±SD (3 assays) of stained cells; significant differences to untreated cells: *; (C,D) immunohistology of protease and adhesion molecule expression in draining LN after repeated ifp application of ASML wt or -CD151 kd and/or -Tspan8 kd exosomes (scale bar: 150µm). Short term in vitro coculture of stroma cells with ASML exosomes hardly affected protease and adhesion molecule (data not shown) expression. Instead, after repeated exosome application in vivo, ASML wt exosomes particularly promoted MMP2, MMP9 and TACE as well as CD49c, CD49d and CD104 expression. CD104 and TACE upregulation were weakened in ASML-CD151/Tspan8 kd exosome treated rats and CD49c, CD49d, MMP2 and MMP9 expression were not supported.
Taken together, a strong induction of cytokine/ chemokine and -receptor expression, most notably of VEGFR3 in RAEC as well as of CXCR4 both depend on Tspan8.
Tumor exosomes were reported to promote inflammation and immunosuppression [59][60][61]65]. Indeed, inflammatory TNFα and IL6 expression became upregulated in BMC and LNC. The latter also expressing C3 and HSP70 at an increased level. We also noted upregulated expression of the immunosuppressive cytokine IL10, but expression of the effector cytokines IL2 and IFNγ as well as of the activation markers CD44v6 and CD69 were unaltered or slightly reduced only in ASM wt exosome treated LNC. Upregulation of TGFβ was more pronounced in BMC than LNC cocultures, where it should be noted that ASML exosomes more strongly affected BMC than LNC. Confirming the exosome-induced shift towards inflammation / immunosuppression, TLR4, p-Stat3 and p-Stat4 expression were upregulated in BMC and LNC (data not shown) and FoxP3 expression was increased in LNC, but NFAT expression was not altered. Furthermore, while ASML-CD151/Tspan8 kd exosomes hardly exerted any effect, in most instances ASML wt exosomes had the strongest impact ( Fig.7A-7F).
In brief, there was a shift towards inflammatory and immunosuppressive cytokines without evidence for significantly impaired effector cytokines. The stronger impact on BMC than LNC and the missing effect of ASML-CD151/Tspan8 kd exosomes are in line with the more ready uptake of ASML exosomes by BMC and the poor uptake of ASML-CD151/Tspan8 kd exosomes (Suppl. Fig.3). Furthermore, though changes were most pronounced after coculture with ASML wt exosomes, they were not abolished after coculture with -CD151 kd or -Tspan8 kd exosomes, indicating that the major contribution of exosomal tetraspanins may rely on binding rather than distinct message transfer.

Exosomal CD151 and Tspan8 and the feedback towards non-metastasizing tumor cells
As exosomes itinerate [66], tumor exosomes could well affect neighboring tumor cells, such that tumor stem cell exosomes modulate non-CIC.
In concern of EMT-related factors including EMTassociated transcription factors, only FN, vimentin, Notch and, borderline, Snail were downregulated in ASML-CD151 kd and/or ASML-Tspan8 kd cells. In exosomes, only expression of Slug and, slightly of Notch was reduced by the CD151 kd and/or the Tspan8 kd (Fig.8A,8B). Instead, after coculture of ASML-CD151/Tspan8 kd cells with exosomes, expression of vimentin, Slug, Twist and, most strikingly, Notch became upregulated. While vimentin, Snail and Slug appeared to depend mostly on Tspan8, the strong upregulation of Notch required CD151-and Tspan8-competent exosomes (Fig.8C-8E).
Coculture of poorly metastatic ASML-CD151/ Tspan8 kd cells with ASML wt exosomes promoting upregulation of several EMT-related proteins and transcription factors supports the hypothesis that exosomes from metastasizing tumor cells can drive neighboring nonmetastatic tumor cells into EMT.
In brief, the metastasis-supporting activity of the tetraspanins CD151 and Tspan8 most likely is promoted by exosomes, which strongly modulate the host matrix, affect hematopoietic and stroma cell activation including (lymph)angiogenesis. In addition, the transfer of exosomes into neighboring non-metastatic tumor cells supports expression of EMT-related genes.

DISCUSSION
Exosomes are important intercellular communicators [14,15] and it was suggested that tumor exosomes, particularly cancer initiating cell exosomes account for premetastatic niche preparation [9][10][11]. Although the mechanisms are not yet elucidated, it is evident that exosomal targeting structures as well as transferred proteins and most prominently transferred miRNA account for the reprogramming of host cells towards supporting settlement and growth of disseminated tumor cells [9]. As tetraspanins are constitutive exosome components and known to be engaged in exosome binding and uptake [32], we speculated that the metastasis-promoting tetraspanins CD151 and Tspan8 [25,26,[37][38][39][40] could play a central role in the process of tumor exosome-mediated host modulation. We demonstrate that ASML-CD151/ Tspan8 kd cells largely lost the capacity to metastasize, but regain metastatic capacity by the support of ASML wt exosomes. Deciphering the role of exosomal CD151 and Tspan8 revealed that both tetraspanins contribute to matrix degradation that supports tumor and host cell motility, affect stroma and hematopoietic cells and are engaged in a feedback towards non-metastatic tumor cells, where they promote EMT. www.impactjournals.com/oncotarget

Loss of metastasis formation by a double knockdown of CD151 and Tspan8
In advance of evaluating the impact of exosomal tetraspanins on the crosstalk with the host, we reassured that ASML-CD151/Tspan8 kd cells lost metastatic potential. The in vivo and in vitro growth characteristics of ASML-CD151/Tspan8 kd cells resembled the sum of that reported for ASML-CD151 kd and ASML-Tspan8 kd cells [51]. Similar to ASML-Tspan8 kd cells, anchorage independent growth of ASML-CD151/Tspan8 kd cells was strongly impaired and apoptosis resistance was more severely affected than in ASML-CD151 kd or ASML-Tspan8 kd cells (data not shown). Migration was similar to that of ASML wt cells, i.e. the striking gain in motility seen in ASML-CD151 kd cells was annulated. ASML-CD151/Tspan8 kd cells were poorly invasive and scarce transendothelial cell migration resembled that of ASML-Tspan8 kd cells without a measurable contribution of CD151. These findings, which confirmed our results on ASML-CD151 kd and ASML-Tspan8 kd cells, provided a solid basis to answer the question on the role of exosomal CD151 and Tspan8 in metastasis.

Exosomal CD151 and Tspan8 and the host matrix
Exosomes maintain the tetraspanin webs of their cells of origin [24], the presented data confirming the general validity of TEM complex recovery in exosomes. Thus, the tetraspanin -integrin associations are maintained in exosomes. Reduced expression of either CD151 or Tspan8 apparently is not sufficient to significantly alter exosome-binding to matrix proteins. This includes the adhesion promoting, but migration inhibitory effect of CD151 that was visible, but less pronounced in exosomes than cells [51]. Instead, the exosomal tetraspanin-protease complexes displayed full activity with a strong impact on invasion, where matrix degradation became strongly impaired or abolished by a deficit of the tetraspanin partner for proteases. Thus, a deficit in CD151 was accompanied by a failure to degrade coll I or FN and degradation of coll IV was reduced, which correlated with poor recovery of CD151-linked MMP2 in ASML-CD151 kd cells. Due to the preferential association of MMP9 with Tspan8, ASML-Tspan8 kd exosomes poorly degraded coll I, coll IV and LN332 and degradation was abolished in the presence of an MMP9 inhibitor. The TACE inhibitor TAPI impaired LN332 and partly coll II, coll IV and FN degradation, which could be due to the preferential, though not exclusive association with Tspan8. Importantly, exosomal tetraspanin-associated proteases modulate stroma matrices in vivo such that invasion is facilitated. This became most obvious for Tspan8 upon iv and ifp application of dyelabeled exosomes, where Tspan8 kd and CD151/Tspan8 kd exosomes were retained in the blood after iv and poorly recovered in the blood after ifp application, which fitted to the deficit of ASML-Tspan8 kd and -CD151/Tspan8 kd cells to pass an endothelial cell monolayer and previous reports on Tspan8 preferentially targeting endothelial cells [32,67].
In brief, the crosstalk between tumor exosomes and the matrix is strongly influenced by exosomal tetraspanins. Due to their associations with integrins and proteases, the tetraspanin web facilitates binding, motility and matrix degradation.

Stroma and hematopoietic cell modulation by exosomal CD151 and Tspan8
Coculture of LnStr, LuFb, RAEC and freshly harvested BMC or LNC with ASML wt , -CD151 kd , -Tspan8 kd and -CD151/Tspan8 kd exosomes allowed appointing changes in target cells to be CD151-and/or Tspan8-dependent or -independent. However, it should be remembered that there is strong evidence that exosomes modulate target cells mostly via transferred miRNA [63]. As we have not yet explored the impact of CD151 and Tspan8 on miRNA recruitment into exosomes, we presently cannot differentiate between exosomal CD151and Tspan8-mediated effects due to tetraspanin binding versus a potential impact due to transferred messages.
Independent of this open question, exosomal CD151 and Tspan8 significantly contribute to target cell modulation, where we concentrated on adhesion molecules, proteases, chemokines and their receptors and, particularly in BMC and LNC on changes associated with immunosuppression and inflammation.
Exosome uptake significantly affected adhesion molecule and protease expression in lymph nodes / lymph node stroma cells and lung / lung fibroblasts. With few exceptions, ASML wt exosomes exerted the strongest effect, which we interpret in the sense that a major contribution of CD151 and Tspan8 might rely on facilitating binding and uptake. Notably, effects on protease and adhesion molecule expression were mostly stronger after in vivo application than after in vitro coculture. We consider in vivo cooperation between exosome target cells and the exosome modulated matrix as a likely explanation.
Expression of several chemokines and angiogenesisrelated factors and -receptors were significantly affected by ASML wt exosomes. Differences in the response to ASML wt versus -CD151 kd , -Tspan8 kd and -CD151/Tspan8 kd exosomes provided additional evidence for a selective engagement of exosomal CD151 in SDF1 upregulation and of exosomal Tspan8 in VEGFR2, VEGFR3 and CD31 upregulation. In cooperation with exosomal CD49d Tspan8 initiates overshooting angiogenesis [67][68][69]. Instead, ASML wt cells and exosomes, which do not express CD49d, selectively support lymphatic metastasis [70], fitting the upregulation of VEGFR3 [71]. Though many carcinoma metastasize via blood and lymphatic vessels, lymph nodes frequently are the first metastatic station [72]. Thus, the selective and efficient contribution of exosomal Tspan8 towards VEGFR3 induction deserves further studies to unravel the underlying mechanism, which may provide a hint towards therapeutic interference.
Tumor exosomes can be immunosuppressive [59], which frequently is accompanied by a shift towards immunosuppressive subpopulations like myeloid-derived suppressor cells and T reg [11,58,59,73,74]. We noted, an increase in the inflammatory cytokines TNFα and IL6 and the immunosuppressive cytokine IL10 [75][76][77], however without evidence for significantly impaired effector cytokine expression. This is surprising as FoxP3 became strongly upregulated in LNC cocultured with ASML wt exosomesand [78]. In LNC and, more pronounced in BMC TFGβ, a central regulator of immune response [78,79], was upregulated. In line with this finding, ASML exosomes promoted JAK/Stat activation [76][77][78]. The impact of exosomes on inflammatory cytokines and TGFβ being more pronounced in BMC, which very efficiently take up tumor exosomes, and being strongest after coculture with ASML wt , but hardly seen with poorly binding ASML-CD151/Tspan8 kd exosomes points towards exosomal CD151 and Tspan8 particularly contributing to exosome binding/uptake.
Taken together, tumor exosomes support induction of an inflammatory response and stroma cell activation. The engagement of exosomal CD151 and Tspan8 in inflammation mostly relies on facilitating binding and uptake. Instead, exosomal CD151 and/or Tspan8 have bearing on SDF1, CXCR4, VEGFR1 and, most relevant for lymphatic metastasis, VEGFR3 expression, such that a blockade of tetraspanins could well provide a therapeutic option. This consideration gains weight by the feedback of exosomes from metastasizing on non-metastatic tumor cells.

The contribution of exosomal CD151 and Tspan8 to epithelial mesenchymal transition
Metastasizing tumor cells are supposed to present a subpopulation of CIC [80] and it has been argued that CIC transfer the potential to metastasize to non-metastatic tumor cells possibly via exosomes [81][82][83]. Thus, we finally controlled the impact of ASML wt exosomes on ASML-CD151/Tspan8 kd cells. RTK expression remained unaltered except for VEGFR1 and VEGFR3. Expression being very low in ASML cells and exosomes excludes changes by coculture with ASML wt exosomes to be due to direct protein transfer. The functional relevance of VEGFR expression in tumor cells as well as the pathway, whereby exosomes account for VEGFR expression, also seen in non-transformed host cells, remains to be explored. Lymphangiogenesis regulation is a complex process, which involves Ras, the MAPK cascade, PI3K/ Akt signaling, TGFβ and Notch [71], where our data point towards a special engagment of Notch.
In view of the partial overlap of (lymph)angiogenesis and EMT [84] and being aware that EMT is one of the first steps towards metastasis [85,86], we evaluated EMT gene expression in ASML-CD151/Tspan8 kd cells after coculture with ASML wt and for comparison -CD151 kd and -Tspan8 kd exosomes. ASML wt exosomes most strongly upregulated Notch. The intracellular domain of Notch binds to RBP-Jκ (recombination binding protein-Jκ), converting the repressor complex into an activator of Notch target gene transcription [87]. Thus, it is tempting to speculate that exosomal proteins and/or miRNA affect a central, upstream component in Notch signaling, as vimentin, Snail, Slug and Twist expression also became upregulated, though to a minor degree. Similar to the regulation of VEGFR3, the impact of ASML-Tspan8 kd exosomes was weaker than that of ASML wt exosomes, which we interpret as indicating a contribution of exosomal tetraspanins, but also excluding the tetraspanins as exclusive mediators.
Thus, ASML wt exosomes promote EMT in nonmetastatic cells. The striking impact of exosomal CD151 and Tspan8 demands for elucidating the underlying mechanism, where we presently cannot exclude a major contribution by tetraspanin-facilitated exosome binding and uptake. Despite this open question, induction of EMT by exosomes of a metastasizing tumor line in a nonmetastasizing tumor line strongly supports our hypothesis on the central importance of exosomal tetraspanins in promoting metastasis.

CONCLUSION
We here provide strong evidence for a central role of exosomal tetraspanins in metastasis. Exosomes facilitate motility and invasiveness by modulating the extracellular matrix, a process wherein exosomal CD151 and Tspan8 are directly engaged due to their associations with integrins and proteases. Exosomes also drive hematopoietic cells towards an inflammatory phenotype and initiate protease and chemokine / chemokine receptor expression in stroma cells. Most importantly, exosomes stimulate EMT in non-metastatic cells. CD151 and Tspan8 are essential for exosome-initiated target cell and non-CIC activation and reprogramming. It remains to be explores, whether the tetraspanin contribution is a sequel of the engagement in exosome binding and uptake or whether tetraspanins additionally contribute to the recruitment of signaling molecules and/or miRNA. Irrespective of this open question, a blockade of exosomal Tspan8 and CD151 appears most promising for interfering with metastasis.

Cell and tissue preparation
The local tumor, draining lymph nodes, lung and bone marrow were collected. Single cell suspensions were obtained by pressing through fine gauze. Alternatively, organs were shock frozen or lysates were prepared by tissue disruption (UltraTurrax, 3-times, 30sec on ice).
Flow cytometry followed routine procedures. For intracellular staining, cells were fixed with 1% formalin and permeabilized (0.5% Tween in PBS) in advance. Apoptosis was determined by AnnV/PI staining. Exosomes were incubated with 4µm aldehyde-sulfate latex-beads (Invitrogen, Karlsruhe, Germany), blocking free aldehyde groups (PBS/100mM glycine, 20min, 20 o C) before staining. Staining was evaluated using a FACS-Calibur and the Cell Quest program (BD, Heidelberg, Germany).

Histology and immunofluorescence
Snap frozen sections (8µm) were fixed with chloroform/aceton (1:1), 4min, dried and incubated with antibodies, washed, exposed to biotinylated secondary antibodies and alkaline phosphatase conjugated avidinbiotin solution, counter-staining sections with H&E. Endogenous tissue alkaline phosphatase activity was ablated with levamisole solution and non-specific binding was blocked using an avidin-biotin blocking kit (Vector Laboratories, Burlingame, CA, USA). Immunofluorescence staining followed routine procedures. For intracellular staining cells were permeabilized with 0.1% Triton after fixation with 4% Formaldehyde. Slides were mounted in Elvanol to generate digitized images (Carl Zeiss LSM710 confocal microscope; software Carl Zeiss Axioview Rel. 4.6).

Adhesion
Cells were seeded on matrix protein-coated 96well F-bottom plates. After washing, adherent cells were stained with crystal-violet and lysed, evaluating OD595 photometrically. Adhesion is presented as the percentage of input cells.

Migration
Cells, in the upper part of a Boyden chamber (RPMI/0.1%BSA), were separated from the lower part (RPMI/20%FCS) by 8µm pore size polycarbonatemembranes. After 16h the lower membrane side was stained (crystal-violet), measuring OD595 after lysis. For in vitro wound healing, a subconfluent monolayer was scratched with a pipette tip, following wound closure by light-microscopy. For video-microscopy, Hoechst 33342 stained cells (5x10 4 ) were seeded on matrix-coated 24-well plates. Plates were placed under an Olympus IX81 inverse microscope with a Hg/Xe lamp, an incubation chamber (37 o C, 5%CO 2 ), a CCD camera (Hamamatsu) and a ScanR acquisition soft ware (Olympus, Hamburg, Germany). Two pictures (20-fold magnification) / chamber (2ms exposure) were taken every 20min for 12h. Migration was quantified according to Manual_tracking plugin running in the open-source software Image J.

cells in 200µl
RPMI/1% BSA were placed on the gels. The lower chamber contained RPMI with 20% FCS. After 48h at 37 o C, 5%CO 2 , the medium in the insert was removed and cells not invading the gels were washed off. Matrix invasion and recovery on the lower membrane side was evaluated microscopically and photometrically after crystal-violet staining and lysis.
Cell cycle progression was evaluated using PI staining and standard procedures.

Soft agar assay
Tumor cells in 0.3% agar were seeded in 6-well plates on a preformed 1% agar layer counting colonies after 3wk.

In vivo assays
BDX rats received 1x10 6 tumor cells, ifp. Where indicated, rats received concomitantly and repeated in 3 day intervals 200µg exosomes ifp. The local growth, growth in the popliteal and the inguinal lymph node were evaluated weekly with sliding calipers. Animals were sacrificed at the indicated time points or when draining nodes reached 2cm diameter, upon >10% weight loss, short breathing and a stiff thorax due to lung metastases or latest after 120d. For the evaluation of exosome distribution, rats received 200µg dye-labeled exosomes, iv or ifp. Animals were sacrificed at the indicated time points. They were bled by heart puncture and all lymphatic / hematopoietic organs were excised and analyzed by flow cytometry for the presence of labeled cells. Animal experiments were Government-approved (Baden-Wuerttemberg, Germany).

Statistical analysis
Assays were repeated at least 3 times. P-values <0.05 (two-tailed Student's t-test, Anova) were considered significant.
Authorship SY, WM and UE performed and analyzed experiments, MZ planned and analyzed experiments and wrote the manuscript.