Laminin Isoforms 8 and 10 Are Primary Components of the Subendothelial Basement Membrane Promoting Interaction with Neoplastic Lymphocytes¹

The phenomenon of intravasation and extravasation of normal and neoplastic lymphocytes is known to involve intimate interactions with both the subendothelial basement membrane and its underlying connective tissue ECM(1).⁴ In particular, the transversing of the former acellular structure is considered to be a critical step in both the ingression and egression of the cells from the blood stream, which in turn is a prerequisite for the invasion of the surrounding connective tissues during inflammatory conditions or uncontrolled dissemination of neoplastic cells. The interaction of neoplastic human lymphocytes(mainly T-cell lymphomas and leukemia) with various basement membrane components has been examined previously (2, 3, 4, 5, 6, 7) in vitro in the effort to disclose the key molecule(s) responsible for the trans-endothelial passage of the cells. Originally,two primary basement membrane components that included EHS laminin-1 (LN-1), which has conventionally been adopted as a model LN-1 in the absence of alternative purified LN-1 molecules derived from higher mammals, and collagen type IV were identified as potential lymphocyte ligands. However, a significant involvement of these ECM components in the intravasation/extravasation process could largely be rejected on the basis of their inability to mediate lymphocyte attachment in the absence of artificial activation of the cells with phorbol esters, mitogens, cytokines, anti-CD3 antibodies, or anti-β1 integrin-activating antibodies (2, 3, 4, 5, 6, 7). On the other hand, an early study indicated that an epithelial basement membrane component denoted “epiligrin” and now known to correspond to a LN-5/LN-6/LN-7 complex could function as a putative attachment-promoting factor for skin-homing T-cell lymphomas(8). A certain preference for LN-8 and LN-10 has been proposed recently for normal hematopoietic progenitor cells of the murine bone marrow where these isoforms appear to prevail as cell-interactive ones (9).

Three plausible arguments may be put forward to explain the previously reported inability of both activated lymphocytes, e.g.,those engaged in inflammatory processes, and neoplastic lymphocytes that exhibit a diffuse tissue spreading to interact with LN-1. First,sporadic analyses of the single LN chains expressed in fetal and adult human blood vessels, murine bone marrow, developing murine endothelia,and cultured mammalian endothelial cells have identified LN-8, LN-9,LN-10, and LN-11 as the primary LNs produced by the endothelium(9, 10, 11, 12, 13). In contrast, LN-2 and LN-4 appear to be weakly expressed or to be deposited exclusively in the more distal layers of the blood vessel wall, such as within the pericyte and smooth muscle layer (media-adventitia) of larger arteries (14, 15). It has also been postulated that α1-containing isoforms may be scarcely represented during embryonic vasculogenesis (11, 14, 15)and in the adult vascular tissues (16), being prevalently expressed in capillaries and smaller vessels (17). Theα3-containing LNs, i.e., LN-5, LN-6, and LN-7, are thought to be largely absent from most adult vascular compartments and may only be found in fetal blood vessels and vascular structures of lymphoid tissues such as thymus and spleen (13, 18, 19, 20, 21, 22). Secondly,observations made on cultured cells suggest that EHS LN-1 could represent an “unusual LN-1” which because of its aberrant tumor-related glycosylation traits could be incapable of functioning as a universal cell-binding LN (23). Finally, the possibility has been raised that LNs may not be involved in the process of lymphocyte transmigration of the blood vessel wall, which,conversely, may be mediated primarily by other basement membrane components.

Thus, to define the potential role of LNs in the process of intravasation/extravasation of malignant lymphocytes, as well as identify the isoforms involved in this process, we have reexamined the distribution of LNs in vascular structures of adult human tissues and have probed the ability of purified LNs to promote adhesion of neoplastic B and T lymphocytes in vitro. Our data reveal a considerable redundancy in the LN isoforms deposited in the basement membranes of these structures and identify isoforms 8 and 10 as the candidate LNs responsible for the interaction of human neoplastic lymphocytes with the basal endothelial ECM of capillaries and smaller vessels. Thus, we propose that these LNs may represent key molecules involved in the ingress/egress of these cells from the blood and lymphatic circulation.

A number of well-described anti-laminin polyclonal and mAbs were used in conjunction with two recently produced mAbs and one polyclonal antiserum (Table 1). If not otherwise stated, production and specificity of the mAbs can be obtained from the commercial source or from previous publications of the providers. Thus, mAbs 4C7 (anti-α5), 5H2 (anti-α2), 4E10(anti-β1), 2E8 (anti-γ1), P3H9–2 (anti-α3), and MAB1904(anti-β1) were purchased from Chemicon International (Temecula, CA). The murine hybridoma clone that produces mAb C4 (anti-β2) was obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA). Polyclonal antibodies 310 (anti-α2) and 317 (anti-α1) were obtained from Lydia Sorokin (Department for Experimental Medicine,University of Erlangen-Nurnberg, Erlangen, Germany) and are cross-reactive with mouse, rat, bovine, and human LN chains. mAbs BM165(anti-α3) and K140 (anti-β3) were provided by Patricia Rousselle(Institute de Biologie et Chimie des Proteines, Centre National de la Recherche Scientifique, Lyon, France) and Robert E. Burgeson (The Cutaneous Biology Research Center). mAbs 20H2 (anti-collagen type XII)and 15B8 (anti-collagen type XIV) were obtained from Claire Letthias(Institute de Biologie et Chimie des Proteines, Centre National de la Recherche Scientifique, Lyon, France). mAbs 652C4 and 652B2 against the human α4 LN chain were produced by conventional procedures by immunization of BALB/c mice with recombinant fragments corresponding to the COOH-terminal G1–G3 domains of the human α4 laminin chain. The selection and characterization of hybridoma clones were carried out by using various immunochemical assays that involve the diverse LN and collagen preparations as antigens. These mAbs do not react with the corresponding recombinant G1–G3 fragments of α1, α2, α3, andα5 human LN chains (data not shown). A rabbit polyclonal antiserum against the human laminin α5 chain was obtained from Marie-France Champliaud (The Cutaneous Biology Research Center, Harvard Medical School, Charlestown, MA). This antiserum does not cross-react with other α chains, and as found here, it does not recognize the β1,β2, and γ1 chains. Polyclonal antisera against the recombinant COOH-terminal G1–G5 domains of mouse α1 chain (As126.46) and against the EHS LN-1-nidogen complex were produced in rabbits according to standard protocols. ELISA and immunoblotting experiments show that the anti-α1 antiserum As126.46 does not react with human LN-2/(4),LN-5, LN-8, LN-9, LN-10, and LN-11 (Fig. 1; data not shown). The antimurine LN-1-nidogen complex antiserum recognizes human and bovine β1 and γ1 chains and nidogen but fails to react with LNs containing the α2, α3, α4, α5, β2, β3,and γ2 subunits (Refs. 24, 25, 26, 27; data not shown).

Antibodies to integrins were obtained as follows: anti-α3 clone F4 from Luciano Zardi (Istituto Tumori, Genova, Italy); anti-α1 (clone TS2/7) from Martin E. Hemler (Dana-Farber Institute, Boston, MA);anti-α1 (clone 1B3.1) from Ilam Bank (The Chaim Sheba Medical Center,Tel Aviv University, Israel); anti-α2 (clone 12F1) from Virgil Woods(Lombardi Cancer Research Center, Georgetown University, Washington,DC); anti-αv from Robert Pytela (Section for Cardiovascular Research,University of California, San Francisco, CA); anti-α6 (clone GoH3)from Arnoud Sonnenberg (Division of Cell Biology, the Netherlands Cancer Institute, Amsterdam, the Netherlands); anti-α3 (clone P1B5),anti-α₃β₁ complex(clone M-KD102), and anti-α₂β₁ complex(clone JBS2) from Chemicon International; and anti-β1 (clone 4B4)from Coulter Scientifics (San Francisco, CA).

Healthy human tissues were obtained from various carcinoma, sarcoma,and cutaneous melanoma specimens removed by surgery from patients received for diagnosis and/or treatment at the Institute of Anatomy and Pathology, University of Catania, and at the National Cancer Institute,CRO-IRCCS, Aviano. Tissues surrounding the tumor lesion were dissected out, asserted by morphological criteria to be nontransformed, and frozen immediately in liquid nitrogen. Frozen specimens were embedded in OCT compound and cryosectioned. Immunohistochemistry was carried out by sequential incubation with primary and secondary antibodies and detection through the streptavidin-biotin complex kit (LSAB; Dakopatts,Copenhagen, Denmark), followed by counterstaining with H&E.

LN preparations (see below) were analyzed by SDS-PAGE under reducing and nonreducing conditions using 3–8% and 3–10% gradient gels and Coomassie Brilliant Blue or silver staining. For immunoblotting,resolved proteins were electrotransferred onto nitrocellulose membranes, which were subsequently saturated with PBS containing 1%α-casein for 2 h at room temperature and incubated with primary antibodies at 4°C overnight. After extensive washing in PBS-0.1%α-casein, the membranes were incubated with horseradish peroxidase-conjugated antirabbit, antirat, or polyvalent antimouse immunoglobulin antibodies and then revealed with the Enhanced ChemiLuminescence Plus chemiluminescence kit (Roche Molecular Biochemicals).

The LN complex from adult bovine kidney (see below) was resolved by SDS-PAGE on 3–8% gradient gels, and the separated polypeptides were electroeluted for further digestion with the lysine-specific protease from Achromobacter (Wako, Germany) for 6 h at 30°C in 50 mm Tris-HCl buffer (pH 9.0), using 4 μg of protease/200μg of protein. The peptides were separated by reversed phase chromatography on a Vydac C-18 column (0.2 × 20 cm)using a linear 0–70% gradient of aqueous acetonitrile in 0.1%trifluoroacetic acid, collected, and sequenced.

Native LN-1-nidogen complex from EHS mouse tumor and LN-2/4 complexes from human placenta and bovine heart were purified by a previously described EDTA-extraction procedure (24, 25, 26, 28). Comparative densitometric analysis of Western blotting-matched Coomassie Blue-stained bands SDS-PAGE gels, aimed at defining the relative ratios of α, β, and γ chain components in LN complexes,showed a significantly higher ratio of β2 in the bovine heart LN than in the placental one. This suggested a significantly higher relative amount of LN-4 in the former complex. We therefore refer to the vascular LN complexes from placenta and heart as LN-2/(4) (human placenta) and LN-2/4 (bovine heart), respectively. Another mouse heart LN-2/4 complex similar to that of bovine was kindly provided by Lydia Sorokin. LN-5 (also known as kalinin) isolated from cultured human keratinocytes through immunoaffinity chromatography on mAb BM165 (29) was kindly provided by Patricia Rousselle. Chick heart LN-2, LN-4, and LN-9, the latter two provided by Ralph Brandenberger and Matthias Chiquet(M. E. Muller Institute for Biomechanics, Bern, Switzerland), were purified from EDTA extracts by sequential lectin and immunoaffinity chromatographies on columns containing antichick β1 chain (mAb 11B2)antibodies, antichick α2 chain (mAb 8D3) antibodies, and the anti-β2 chain mAb C4 (30, 31, 32). Pepsin-extracted human placenta, affinity-purified on columns that contained mAb 4C7, was purchased from Chemicon International. Although a previous publication(33) indicated that this immunoaffinity purification protocol yields a mixture of LN-10 and LN-11, our data, based upon immunochemistry and amino acid sequencing of polypeptides resolved by SDS-PAGE, show that the preparation used in this study is composed exclusively of LN-10.⁵LN-11 was purified from cultured rat RN22 Schwannoma cells as previously described (34, 35).

A naturally occurring LN complex present in adult bovine kidney was received from Anders Lindblom, Mats Paulsson, and Ralph Broermann (Department of Biochemistry, University of Cologne, Germany)and purified in the laboratory according to minor modifications of a previously published protocol (27). Rat yolk sac LN was purchased from Chemicon International. Nidogen-free EHS LN-1, human plasma fibronectin, and rat tail collagen type I were purchased from Collaborative Research Biochemicals. Purified collagen type XII was obtained from Claire Letthias, and a mixture of collagen type XII and XIV was obtained from the differential elution of CM-cellulose columns during the purification of the bovine kidney LN complex(27). LN-8 was purified from cultured bovine aorta endothelium cells as previously described (10, 11). Consistent with previous observations, ELISA, SDS-PAGE, and Western blotting experiments with anti-α5 and anti-β2 chain antibodies confirmed here that endothelium-derived LN-8 was free of LN-9-LN-11(data not shown).

The lymphocyte lines Karpas 299 (T-cell lymphoma), HUT78, HUT102(cutaneous T-cell lymphoma), CEM (T-cell lymphoma), Molt-4, Jurkat(T-cell leukemia), Raji, Ramos and Sc-1 (Burkitt type lymphoma), Ri-1(large nonconvoluted B-cell lymphoma), BV-173 (B-lymphoid blast crisis of a chronic myeloid leukemia), and Ci-1 (B-cell lymphoblastic lymphoma) were grown in RPMI 1640 medium supplemented with 10–15%FCS. The sarcoma cell lines HT1080 (fibrosarcoma), RD-kD(rhabdomyosarcoma), SK-LMS-1, SK-UT-1, and MES-SA (leiomyosarcoma)were grown in DMEM supplemented with 10% FCS. Peripheral B lymphocytes were isolated after informed consent from patients affected by CLL or ALL and referred to the Department of Hematology and Oncology of the National Cancer Institute, CRO-IRCCS, Aviano. For flow cytometric analysis of integrin expression by indirect immunofluorescence, cells were washed in PBS with 5 mm EDTA and incubated with primary antibodies in PBS with 5 mm EDTA and 0.01%NaN₃ for 45 min on ice, followed by washing and incubation with secondary immunoglobulin-species specific antibodies for 30 min on ice. Antibody-labeled cells were fixed and examined for surface fluorescence in a FACScan flow cytometer, and the data were analyzed with the Lysis II program (Becton Dickinson).

Despite of the previously documented cell-binding activity of the single LN isoforms and LN complexes used in this study, we initially ascertained the ability of these LNs to support adhesion of well-established LN-binding cells. Details of the cell adhesion assay CAFCA used in this study have been provided in previous publications(36, 37, 38). A series of tests were initially carried out to determine the minimal reverse centrifugal force required to detach cells adhering nonspecifically to BSA yet sufficiently low to detect specific low-avidity interactions with LNs. This minimal force was established to be 12 × g for all cell lines when applied for 5 min at 37°C, whereas the optimal force for ex vivo CLL and ALL cells was found to be 42 × g. Another series of experiments, in which the CAFCA miniplate assemblies were centrifuged at 12, 45, 100, 177, 399, and 740 × g, were carried out with the aim of determining the relative strength of cell adhesion to the various LN substrates (see below). Fibronectin was consistently used as a reference adhesion-promoting molecule. The relative number of cells bound to the substrate (i.e., remaining bound to the bottom miniplates) and cells that fail to bind to the substrate(i.e., remaining in the wells of the top miniplates) were estimated by top/bottom fluorescence detection in a computer-interfaced SPECTRAFluor Plus microplate fluorometer (TECAN, AG, Ropperswit,Switzerland). Fluorescence values were elaborated by the CAFCA software (TECAN) to determine the percentage of adherent cells out of the total cell population analyzed, according to a previously published formula (36, 38). Relative strengths of cell adhesion in dynes/cell (A_fd) were calculated as previously described according to the formula: A_fd = (D_c − D_m) × V_c × F_c, where D_c is the specific density of the cell, previously established to be 1.07 g/cm³(36); D_m is the specific density of the medium = 1.00 g/cm³; V_c is the volume of the cell; and F_c is the centrifugal force exerted on the bound cell. The average cell diameter was estimated to be 12 μm. Statistical significance was determined by Student’s t test with the significance limit set at P < 0.001.

A panel of anti-LN antibodies with previously documented specificity and/or with reactive patterns established herein (Table 1) was used immunohistochemically to determine the in situ distribution of various LN isoforms in vascular structures of 11 representative adult tissues, which included skin, thyroid, placenta, kidney, skeletal muscle, spleen, endometrium, mammary gland, lung, colon, and stomach. With the exception of larger veins and arteries and irrespective of the type of vascular structure, i.e., smaller capillaries,arterioles, or postcapillary venules, subunits that compose 9 of the 12 currently known LN isoforms (i.e., LN-1–5 and LN-8–11)were detected with the chain-specific antibodies in coincident patterns in the majority of the studied tissues. Representative stainings obtained for α1-α5, β1, β3, and γ1 chains are shown in Fig. 2. Small- and medium-sized arteries of lung and skin failed to expressα3 and β3, as indicated by the lack of reactivity with mAb K140 and BM165, respectively (Fig. 2). Similarly, α2 expression was not observed in capillaries of the kidney cortex glomerular capsules and in vascular structures of the lung (data not shown). This suggests that,at least in humans, adult and fetal blood vessels may differ and may exhibit a LN isoform composition divergent from that of the vasculature of lower mammals. Our data on the relative widespread distribution of the LN-1 in vascular basement membranes differ somewhat from that recently reported (17) using α1 subunit-specific mAbs and showing a more restricted expression of this chain to smaller capillaries. The reason for this discrepancy is presently unclear but may be associated with differential masking effects that impair full recognition of certain partially buried α1 chain epitopes.

The heterogeneous LN complex isolated from adult bovine kidney(27, 39, 40) was characterized further by biochemical and immunochemical means. Amino acid sequences that were derived from 14 different high-performance liquid chromatography-separated tryptic peptides generated from the putative α chain bands of this LN complex and running on SDS-PAGE under reducing conditions in the range of M_r 350,000–400,000 indicated the presence of the α1 and α5 LN subunits, as well as Col XII as a primary non-LN contaminant (Fig. 3). Immunoblotting of this complex with chain-specific anti-LN antibodies both asserted the presence of LN α subunits and, in accordance with previously published observations (39, 40), identified theα5 subunit as the molecular species migrating at about M_r 370,000–380,000 (Fig. 3). In contrast, no α3 or α4 LN subunit could be detected, whereas Col XII but not Col XIV was confirmed immunochemically to be associated with this LN complex (Fig. 3). The relative ratio of the two identifiableα chain components, i.e., α1 and α5, was technically impossible to establish because the isoforms that contained these subunits were not completely separable one from another. However, data obtained by immunoblotting and densitometric analysis of theα1/M_r 400,000 andα5/M_r 370,000–380,000 Coomassie blue-stained bands suggested that the α1 prevailed over the α5 subunit. Thus, collectively, our observations provide a further characterization of this LN complex (27, 40) and indicate that it is composed of a mixture of LN-1, LN-10, and LN-11. Immunoblotting of this complex with the anti-α5 chain antiserum also disclosed a M_r 200,000 polypeptide (Fig. 3). Bands that migrate below M_r 300,000 have been identified in other tissue-derived putative LN-10/LN-11 complexes (32, 41) but not in immunopurified LN-10 preparations (39, 42), suggesting that the M_r ∼200,000 polypeptide may correspond to a naturally occurring proteolytic fragment of the tissue form of the α5 LN subunit. This fragment may also be found in LN-11-containing preparations, as suggested by the fact that LN-11 isolated from RN22 Schwannoma cells and α5-containing LNs from astrocytoma cells both exhibit “Y-shaped” configurations(33, 34, 43, 44), which resemble those observed for the bovine kidney LN complex (27).

Our immunochemical data on the previously described rat yolk sac LN (45, 46, 47) suggest that it corresponds to a mixture of LN-1 and LN-10, as shown by positive immunoreaction of antisera against the α1 and α5 subunits with the M_r400,000 and M_r 370,000–380,000 bands(Fig. 3). In analogy with the LN complex from the kidney, a putative M_r 200,000 α5 fragment could be detected in this preparation (Fig. 3). Because Western blotting with anti-α1 and anti-α5 antisera suggested a prevalence of LN-1 (Fig. 3), we refer to this nidogen-free complex as LN-1/(10).

To determine the capability of neoplastic lymphocytes to interact with vascular LNs and the possible isoform preference of this interaction,we compared the ability of the various unstimulated neoplastic lymphocyte types to adhere to the panel of purified LNs (using fibronectin as reference adhesion-promoting molecule). Lymphocytes largely failed to bind to LN-1, LN-4, LN-9, and LN-11, bound weakly to moderately well to LN-2, LN-5, and LN-8, and attached to LN-10 to a similar or higher extent as to fibronectin (Table 2). A similar high-avidity binding was observed to the bovine LN-1/10/11 complex (Fig. 4), which suggested that the apparently lower amounts of the largely inactive LN-1 and LN-11 did not counteract the adhesion-promoting activity of LN-10. In contrast, LN-1/(10) from rat yolk sac and LN-2/4 complexes from human placenta and bovine heart were significantly less effective than the purified isoforms, i.e., LN-2 and LN-10,and than complexes in which the active LNs predominated, i.e., LN-2/(4) and LN-1/10/11 (Fig. 4). Apart from LN-4 and LN-11, proposed to play solely an important role in the promotion of neurite outgrowth, all of the LNs that showed a weak ability to support lymphocyte adhesion actively sustained attachment of human sarcoma cells. This finding indicates that LNs unable to mediate significant adhesion of neoplastic lymphocytes did not lack a generalized capability to bind cells.

Some differences were noted in the cell-LN interaction between the different lymphocyte types. Leukemic and lymphoma T cells pronouncedly attached to LN-2 and LN-5, whereas leukemic B cells and Burkitt-type lymphomas were ineffective in binding to these isoforms. The superior ability of LN-10 to promote neoplastic lymphocyte adhesion was also demonstrated by the relative higher cell-binding avidity displayed by both T and B neoplastic lymphocytes to this LN isoform. In fact, when compared, the avidity of lymphocyte adhesion to LN-10 and LN-1/10/11 was 2–4-fold higher than to LN-2/(4) and LN-5, and the average relative force necessary to detach cells from the LNs was estimated to be A_fd of ≤7 × 10⁻⁹ versusA_fd of ≤2 × 10⁻⁷ dynes/cell for LN-10/LN-1/10/11 and LN-2/(4)/LN-5, respectively (Fig. 5). Lymphocyte binding to LNs was entirely cation-dependent and was promoted by physiological concentrations of Mg²⁺(0.3–3 mm) and Mn²⁺(0.1–30 μm) but not Ca²⁺(data not shown).

To establish whether the observed LN isoform preferences exhibited by the neoplastic cell lines were representative also for lymphocytes derived from patients with hematological malignancies, we examined the ability of peripheral lymphocytes isolated from CLL and ALL patients to interact with the various LNs. Although showing some variability, B lymphocytes from CLL patients attached well to LN-8, LN-10, and the LN-1/10/11 complex, with the adhesion capability of cells from the different patients being closely related to the overall ability of the cells to recognize LNs (Table 3). On the other hand, ALL lymphocytes that completely lacked LN-binding integrins or that showed weak expression of these receptors bound significantly more poorly to the isolated LN-8 and LN-10 but still adhered to LN-1/(10) and LN-1/10/11 complexes (Table 3).

Lymphocyte adhesion to all LNs tested herein was completely blocked by anti-β1 integrin subunit antibodies but was unaffected by antibodies to β3 and β5 subunits (not shown). A similar strong reduction in lymphocyte attachment to LNs was seen after the addition of antibodies to either the α3 or α6 integrin subunit and/or against theα ₃β₁ complex but not by antibodies to α1, αv, α2, andα ₂β₁ integrin subunit/complex (Fig. 6). Affinity chromatography of iodinated Karpas 299 and Jurkat cell membranes on immobilized LN-1, LN-2/(4), and LN-1/10/11 confirmed the high-affinity binding of theα ₃β₁ (on Karpas 299)and α₆β₁ (on Jurkat)integrin receptors, because these were specifically eluted with EDTA from columns containing LN-2/(4) and LN-1/10/11 but not LN-1 (not shown).

As shown by antibody blockade experiments and FACS analyses,utilization of α₃β₁ orα ₆β₁ integrins for lymphocyte interaction with LN-2, LN-5, LN-8, and LN-10 was strictly dependent upon which of the integrins prevailed on the cell surface. However, the α₆β₁integrin consistently dominated over theα ₃β₁ in mediating LN binding, even when present in apparently lower relative amounts. Thus,Karpas 299, HUT78, HUT102, Ri-1, Sc-1, and Ci-1 showed high expression of α₃β₁ and no detectable surface expression ofα ₆β₁, and these cells used the former integrin receptor for binding to all interactive LN isoforms. On the other hand, Jurkat, Molt-4, Raji, Ramos, and BV-173 cells expressed superior levels ofα ₆β₁ thanα ₃β₁, and in these cases the LN interaction was dictated by the former integrin. An interesting correlate was that only cells that expressed substantial amounts of α₆β₁exhibited a weak binding to LN-1 (e.g., Jurkat and Raji cells; Table 2), indicating thatα ₆β₁ may be a universal and genuine LN-1-binding integrin.

To elucidate the role of subendothelial LNs of capillaries and smaller vessels in the neoplastic lymphocyte intravasation/extravasation phenomena, we have: (a) mapped the distribution of the subunits composing 9 of the 12 putative LNs known to date in a number of representative adult healthy human tissues/organs; (b) characterized biochemically and immunochemically isolated LNs and LN complexes derived from highly vascularized tissues; and (c) comparatively examined the ability of these LNs to promote attachment of lymphocyte lines and lymphocytes isolated from leukemic patients. We find that, despite the fact that virtually all of the subunits examined here codistributed in the smaller vessels corresponding to sites of lymphocyte intravasation/extravasation, LN-10 and, to a lesser extent, LN-8 prevailed in their ability to interact with neoplastic lymphocytes. This finding strongly suggests that these LN isoforms may represent primary lymphocyte ligands of the subendothelial zone. The reason for the excessive redundancy in the tissue compartmentalization of LNs in the adult human body is presently unclear, but the observation strongly suggests a well-articulated functional diversity of the various isoforms in these tissue compartments. In light of previous observation in embryonic and adult mouse tissues, this observation also emphasizes marked differences in LN distributions in vascular structures of lower mammals and man.

Assignment of LN-8 and LN-10 as the candidate components that govern neoplastic lymphocyte interaction with the subendothelium was made possible through the combined immunolocalization of these isoforms in the corresponding basement membrane of adult human tissues and via the utilization of a number of purified LNs and LN complexes with asserted composition. In particular, the isolation and characterization ofβ1-containing isoforms (i.e., LN-1, LN-2, LN-8, and LN-10)and β2-containing isoforms (i.e., LN-4, LN-9, and LN-11)were instrumental in the discrimination of the lymphocyte responses to the major vascular LNs. Different B and T neoplastic lymphocytes bound avidly to LN-2, LN-5, LN-8, and LN-10 in a cation-regulated manner but largely failed to attach to LN-1, LN-4, LN-9, and LN-11. Similar observations had been made earlier when examining the LN-binding ability of ex vivo lymphocytes derived from patients affected by CLL-B (36), and these findings were extended here by examining lymphocytes derived from an analogous hematological malignancy, i.e., ALL. In these cases, some variation in the binding capability of the cells to LN-8 and LN-10 was noted and was associated with an overall weak capability of cells from certain patients to attach to both LNs and fibronectin. However, inefficient recognition of LNs in these cells appeared to be attributed to a poor cell surface-expression of LN-binding integrins, whereas a certain preferential binding of ALL cells to LN-1/10 complexes indicated a potential cooperative effect of these isoforms in mediating attachment of these specific cell types. The molecular basis for this different behavior of ALL cells was not further investigated here. Taken together, the findings highlight a clear bias of transformed lymphocytes for LNs that contain the β1/γ1 and α3-α5 subunits,which was unlikely to be determined by a loss of biological activity during purification of the less active or completely inactive LN isoforms. This is because LN isoforms tested in this study were ascertained to have retained cell adhesion-promoting activity for selected anchorage-dependent tumor cells, as well as the ability to stimulate neurite outgrowth in vitro.

The marked to complete failure of neoplastic lymphocytes to attach to isolated LN-1 and LN-1-nidogen complexes seemed, on one hand, to be associated with the apparent inability of theα ₃β₁ integrin (when present as the primary LN-binding integrin on the cell surface) to recognize this specific isoform and, on the other hand, to be because of an apparent inability of the nonactivatedα ₆β₁ integrin to mediate the LN interaction with the same efficiency as that observed for mesenchymal/epithelial cells. Despite being widely expressed in vascular tissues, β2-containing LNs were virtually inactive in supporting neoplastic lymphocyte adhesion. This observation is highly consistent with the preferential ability of LN-4/LN-9/LN-11 to exert a guiding function during embryonic neurite growth in vitroand in vivo (30) but to fail to promote embryonic cell movement in vitro and in vivo(30, 31). The use of CAFCA demonstrated a significantly higher avidity binding to the isolated “vascular” LN-10 (and the LN-1/10/11 complex comprised of this isoform) than to the“peri-vascular” LN-5 and LN-2/(4). This observation raises an interesting point in that, although cells use the same integrin receptors (see below) for recognizing the different LNs, the strengthening of the adhesive responses transduced to cells by these ligands markedly differs.

Static interaction of T and B neoplastic lymphocytes was mediated by both the α₃β₁ andα ₆β₁ integrin receptors, depending upon which of these integrins was the prevailing putative LN receptor expressed on the cell surface, with a predominant binding activity exerted by theα ₆β₁ integrin. This finding is in disagreement with a previously postulated propensity for an α₃β₁-LN-10(40) versusα ₆β₁-LN-1/LN-8(48) interaction, but it is in accordance with a recent study on carcinoma cells showing interchangeable utilization of both integrins for the interaction with LN-10 (49). The observation also affirms that theα ₆β₁ may be the primary, universal β1 LN integrin receptor and that LN-1 may not be a preferred α₃β₁ ligand. The cation requirement for optimal cell attachment to LN-8 and LN-10 largely paralleled that reported for optimal activity of theα ₃β₁ integrin when transfected into K562 cells (40). However, in contrast to this previous report, we do not find evidence that Ca²⁺ would function in an inhibitory manner forα ₃β₁-mediated lymphocyte adhesion to LN. Similarly, we do not find evidence for a high-affinity binding of theα ₃β₁ integrin to EHS LN-1 that would be confined to a single linear sequence in the COOH-terminal globules of LNs (50). First,α ₃β₁⁺/α₆β₁⁻B and T lymphocytes failed to interact with EHS LN-1 and poorly reacted with the rat yolk sac LN-1/(10) complex. Secondly, manipulation of the tertiary structure of the LN-10 heterotrimer strongly reduced the levels of theα ₃β₁-mediated cell adhesion (data not shown). A corollary observation was that refolding(after reversible unfolding) of the LN-1/10/11 complex largely reconstituted the ability of Karpas 299 cells to recognize the complex through their α₃β₁integrin, which suggests that the nature of its binding site within the LN-10 may be different than that previously identified for the fibrosarcoma and gliomaα ₆β₁ integrin in EHS LN-1 (51).

In conclusion, our observations reiterate in a different paradigm the previously postulated nonpermissive effects of LN-1 and β2-containing LN isoforms on adhesion of certain cell types that still express the adequate LN-binding integrin receptors and emphasize de novo a function for LN-8 and LN-10 in promoting lymphocyte-subendothelium interactions. Thus, the observations suggest the existence of positive and negative controls operated by different codistributed LN isoforms on different cell adhesion phenomena.

We are indebted to Drs. Ralph Brandenberger, Ralph Broermann,Matthias Chiquet, Anders Lindblom, Tracey Marsh, Mats Paulsson, and Patricia Rousselle for contribution with various purified LNs and antibodies, and to Claire Letthias and Lydia Sorokin for anti-collagen and anti-LN antibodies. Drs. Ilam Bank, Martin Hemler, Arnoud Sonnerberg, Virgil Woods, and Luciano Zardi are thanked for providing various anti-integrin antibodies. We are grateful to Antonella Corsaro,Daniela Messina, Maria Teresa Mucignat, and Michela Zambon for their technical assistance.