Binding of Netrin-4 to Laminin Short Arms Regulates Basement Membrane Assembly*

Netrins were first identified as neural guidance molecules, acting through receptors that are members of the DCC and UNC-5 family. All netrins share structural homology to the laminin N-terminal domains and the laminin epidermal growth factor-like domains of laminin short arms. Laminins use these domains to self-assemble into complex networks. Here we demonstrate that netrin-4 is a component of basement membranes and is integrated into the laminin polymer via interactions with the lamininγ1 andγ3 short arms. The binding is mediated through the laminin N-terminal domain of netrin-4. In contrast to netrin-4, other members of the netrin family do not bind to these laminin short arms. Moreover, a truncated form of netrin-4 completely inhibits laminin-111 self-assembly in vitro, and full-length netrin-4 can partially disrupt laminin self-interactions. When added to explant cultures, netrin-4 retards salivary gland branching morphogenesis.

Netrins were first isolated as long range guidance cues, acting in early embryogenesis by regulating the migration of neurons and the axonal growth cone (1). These proteins showed either chemoattractive or chemorepulsive effects upon distinct sets of cells, hence cells expressing netrin-1 can mimic the ability of the floor plate to repel the growth cones of trochlear motor neurons in vitro, while attracting the axons of spinal commissural neurons (2)(3)(4). Consequently, netrin-1 is considered a bifunctional guidance cue. This activity has been shown to relate to expression levels of specific receptors from either the DCC or UNC-5 families (5)(6)(7)(8)(9).
Recently netrins and their receptors have been shown to be significant in developmental and physiological events outside the nervous system, including their acting as angiogenic factors (17)(18)(19) and being involved in epithelial branching morphogenesis of the lung, intestine, pancreas, and mammary gland (20 -24). Furthermore, netrin-1 has been suggested as a ligand for the laminin receptors integrins ␣6␤4 and ␣3␤1 and hence has been described as acting as an adhesive cue (25).
We have previously shown that netrin-4 is also a component of basement membranes in a variety of tissues (13). Basement membranes are specialized extracellular matrices underlying all endothelia and epithelia and surrounding many forms of mesenchymal cells (26). They have both physical and signaling functions that alter with both tissue type and stage of development. In addition to forming surfaces to which cells attach, they also transmit force between cells and the surrounding extracellular matrix. Basement membrane components may also regulate many aspects of intracellular activity by signaling via the integrin, dystroglycan, and syndecan receptor families. Furthermore, basement membranes can act as reservoirs for cytokines and growth factors, e.g. members of the fibroblast growth factor family. Finally, signaling by other cytokines may be complemented by basement membrane-induced signals, as seen with neurotrophins. Hence basement membrane-induced signaling plays an important role in cell survival and differentiation as well as in cell migration and as an axonal guidance cue.
Laminins are the major noncollagenous proteins of basement membranes and are crucial in its formation. Laminin-111, the prototype laminin, is a cross-like shaped molecule formed as a multidomain heterotrimer assembled of one ␣1, one ␤1, and one ␥1 chain (27)(28)(29). The laminin trimer has one long and three short arms, the latter being formed from the three free N-terminal ends of the ␣1, ␤1, and ␥1 chains (30,31). These parts of the ␤1 and ␥1 chains each contain two globular domains, designated LN and L4/LF. The globules are interspersed by multiple LE modules, forming rods in domains LEa and LEb. Certain laminin chains have been described as maintaining the domain structure of those in the original laminin-111 (e.g. ␣2 and ␤2), whereas others have N-terminal truncations, lacking an LN domain. So far 15 laminin isoforms have been shown to occur (29). Laminins self-assemble into a network through Ca 2ϩ -dependent interactions between their N-terminal parts (32), and polymerization of laminin-111 may be inhibited by proteolytic laminin fragments that contain LN domains (33).
Netrin-4 is the most recently described soluble netrin family member, and its biological significance is still poorly understood. We have shown that netrin-4 is widely expressed with the protein being concentrated in certain basement membranes and having a spatial expression that broadens during later development. Here we demonstrate the integration of netrin-4 into the basement membrane via the binding to the N-terminal region of the laminin ␥1 chain. Furthermore, we show the significance of netrin-4 in basement membrane assembly and its effect upon branching morphogenesis.
The expression vectors were transfected into 293-EBNA cells with FuGENE 6 transfection reagent (Roche Diagnostics), and selected clones with the highest protein expression were expanded for large scale production. The purification of the secreted proteins was performed as described previously (13).
Antibody Production-For antibody KR1, a rabbit was immunized with purified recombinant ⌬netrin-4. The antiserum was affinity-purified by applying it to a Sepharose column to which mouse ⌬netrin-4 protein had been coupled. Bound antibodies were eluted with triethylamine, pH 11.5, immediately neutralized, and dialyzed against PBS. For antibody KR 24, a rabbit was immunized with purified recombinant netrin-1, and the antiserum was affinity-purified as described for antibody KR1.
Immunofluorescence Microscopy-Newborn (P2) and adult mouse tissues were embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek Europe). 7-m-thick sections were cut with a Leica cryostat and stored at Ϫ20°C. For use, slides were returned to room temperature, fixed, washed in PBS, and blocked with 0.2% Tween 20 in PBS for 1 h at room temperature. Slides were fixed in 4% paraformaldehyde for 5 min, washed, blocked with 5% goat serum, and incubated with antibodies against netrin-4 (KR1, see under "Antibody Production") and the laminin ␥1 chain (rat anti-laminin ␤2 chain monoclonal antibody, Chemicon). After incubation with secondary antibodies (Cy3-labeled goat anti-rabbit IgG, Jackson ImmunoResearch; or Alexa 488-labeled goat anti-rat IgG, Molecular Probes), mounted sections were observed under a laser scanning confocal microscope (Leica) scanning the sections 16 times.
ELISA Style Ligand Binding Assay-Unless otherwise specified, all solutions used contained 2 mM CaCl 2 . For testing the divalent cation dependence of interactions, no CaCl 2 , 2 mM CaCl 2 , or 20 mM EDTA was added to the solutions. Purified proteins were diluted in TBS, pH 7.4, and coated at 10 g/ml (500 ng/well) overnight at room temperature onto 96-well plates (Nunc Maxisorb). After washing with TBS containing 0.05% Tween 20, plates were blocked for 2 h at room temperature with TBS containing either 5% milk powder or 1% bovine serum albumin. Ligands were diluted to concentrations between 0.001 and 50 nM and incubated in the wells for 1 h at room temperature. After extensive washing with TBS containing 0.05% Tween 20, bound ligands were detected with specific polyclonal rabbit antibodies directed against netrin-1 and netrin-4, respectively, followed by horseradish peroxidase-conjugated swine anti-rabbit immunoglobulins (DAKO Cytomation) and tetramethylbenzidine as substrate. Absorption was measured at 450 nm after stopping the reaction with 10% sulfuric acid.
Surface Plasmon Resonance Binding Assays-Assays were performed using a Biacore 2000 (BIAcore AB). Coupling of proteins to the CM5 chip was performed in 25 mM sodium acetate, pH 5.0, at a flow rate of 5 l/min. A 7-min pulse of 0.05 mM N-hydroxysuccinimide, 0.2 M N-ethyl-NЈ-dimethylaminopropyl carbodiimide was used to activate the surface. The protein was injected until the desired amount was coupled (500 -1000 RU), and excess reactive groups were deactivated by a 7-min pulse of 1 M ethanolamine HCl, pH 8.5. Measurements were carried out in HBS (20 mM Hepes, 150 mM NaCl, pH 7.4) containing 2 mM CaCl 2 and 0.005% P20 at a flow of 25 l/min. The injection of 100 l of protein solution (0.05-1 M) was followed by a 400-s dissociation. Each analysis was carried out at six different concentrations. The data were analyzed with BIAevaluation software 3.2 according to the Langmuir model for 1 to 1 binding. All binding curves were fitted with an accuracy of 2 Ͻ1% of R max (maximal RU). The mean k a , k d , and K D values for the six concentrations are given in Table 2.
Cross-linking Assays-Cross-linking assays were carried out using the lysine side chain-reactive cross-linker bis[sulfosuccinimidyl]suberate (Pierce). The reaction was carried out at a protein concentration of 2.8 M in a final volume of 50 l of PBS, pH 7.4. The cross-linker was used at concentrations from 0.5 to 2 mM. The reaction was allowed to continue for 1 h on ice and was stopped by the addition of 10 l of 1 M Tris-HCl, pH 8.0.
Organ Culture-Submandibular glands from embryonic day 13.5 BL/6 mice were placed on a Nucleopore filter (Whatman) and cultured at the air/medium interface. Cultures were carried out in serum-free Dulbecco's modified Eagle's medium/F-12 supplemented with 2 mM L-glutamine, 50 g/ml transferrin, and penicillin/streptomycin (Invitrogen). For testing the effects of full-length netrin-4 protein and ⌬netrin-4, lacking the netrin C domain and ⌬laminin ␥1, these were added to the culture medium at a final concentration of 50 g/ml at the onset of organ culture. For negative controls the same amount of PBS was added. Medium was changed every day. After 3 days the explants were photographed under a microscope and embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek Europe). Sections were cut and stained as described above.

Netrin-4 Expression in Newborn and Adult Mice-We
described previously the expression pattern of netrin-4 in the adult mouse and showed that it is present in basement membranes in a variety of tissues (13). To extend these observations, the netrin-4 expression in kidney was studied during development, initially by indirect immunofluorescence microscopy with a netrin-4-specific polyclonal antibody on kidney sections from day 14.5 (E14.5) and day 16.5 (E16.5) mouse embryos, newborn mice (P2), and 3-4-month-old adult mice (for antibody specificity see supplemental Fig. 1). Surprisingly, expression was not observed until birth (Fig. 1B), with no immunostaining of the embryonic kidney (not shown). Postnatally expression increases, so that while at P2 netrin-4 staining was mainly localized to the glomerular basement membrane and the vascular basement membranes (Fig. 1, B and C), and in the adult kidney the tubular basement membrane and that of the tissue capsule also stained for netrin-4 ( Fig. 1, E and F). A strong co-localization of netrin-4 and the laminin ␥1 subunit was observed, although netrin-4 was also observed at non-basement membrane sites (Fig. 1, C and F), for instance in the mesangium of the glomerulus. This suggests that netrin-4 expression occurs as epithelial structures mature, e.g. the kidney tubules.

Netrin-4 Interactions with Laminin LN Domains-
The basement membrane localization of netrin-4 suggests its involvement in protein-protein interactions. Its marked homology to the LN domains of laminin, which participate on laminin selfinteractions, raised the possibility that netrin-4, as well as other members of the netrin family, could be a laminin-binding protein. Hence, recombinant proteins were produced that corresponded to the N-terminal part of laminin chains or were full-length or modified forms of netrin-1 or netrin-4 (Fig. 2). Initially, surface plasmon resonance-based assays were carried out to identify any possible binding between short arms of laminins and netrin-1 and -4. This showed a strong binding of the ⌬laminin ␥1 chain-(LN ϩ LEa) and ␥3 chain (LN ϩ LEa ϩ L4 ϩ LEb)derived proteins to ⌬netrin-4 (LN ϩ EGF 1-3.5; Fig. 3) but not to ⌬netrin-1 (LN ϩ EGF 1-3; data not shown); the calculated K D was 2.35 ϫ 10 Ϫ8 M for the ␥1 chain fragment (Fig. 3A) and 1.96 ϫ 10 Ϫ8 M for the ␥3 chain short arm (Fig. 3B). In addition binding to the ⌬laminin ␤1 chain (LN ϩ LEa) could be observed (not shown). The following experiments focused on the ⌬laminin ␥1 chain, since this subunit is common to most laminin heterotrimers (29). We also tested the laminin ␥1 chain-netrin-4 interaction in an ELISA style ligandbinding assay. After coating the ⌬laminin ␥1 protein (LN ϩ LEa) onto wells, increasing concentrations of ⌬netrin-4 (LN ϩ EGF 1-3.5) were added, and binding with an apparent K D of 1.38 ϫ 10 Ϫ11 M was detected (not shown).
In addition, we investigated whether ⌬netrin-4 also binds to native laminin-111 isolated from mouse EHS tumor. ELISA style assays were carried out with laminin-111 as the immobilized ligand, and increasing concentrations of ⌬netrin-1 and ⌬netrin-4 were added to the solution (Fig. 4). ⌬Netrin-1 failed to bind laminin-111, whereas ⌬netrin-4 bound to native laminin-111 with an apparent K D of 3.70 ϫ 10 Ϫ11 M, similar to the one observed in the experiment with recombinant ⌬laminin ␥1 (LN ϩ LEa) chains in the same kind of assay. These results suggest that the netrin-4 binding activity for laminin-111 resides in the LN and LE domains of the laminin ␥1 chain.
The laminin ␥1 chain-netrin-4 interaction was also studied using the water-soluble covalent cross-linking agent BS 3 . As shown previously and in this cross-linking experiment, the N-terminal ⌬laminin ␥1 chain fragments do not self-interact (Fig. 5A) (37). ⌬Netrin-4 on the other hand shows self-interaction as well as an interaction with ⌬laminin ␥1 chain fragments ( Fig. 5, A and B). Remarkably, ⌬netrin-4 binding to an N-terminal ⌬laminin ␥1 chain fragment (LN ϩ LEa) results only in the formation of dimers, whereas the ⌬netrin-4 self-interaction results in multimers. As expected no interaction between netrin-1 and the ⌬laminin ␥1 chain was detected (Fig. 5C).  Fig. 2A).

The Netrin-4 Interaction with Laminin LN and LE Domains Is Independent of Divalent Cations-To
Some laminin LN domains undergo a conformational change in the presence of Ca 2ϩ (37). To determine whether the conformation of ⌬netrin-4 is also influenced by Ca 2ϩ , circular dichro-  ELISA style ligand binding assays were performed using native laminin-111 (trimer consisting of the laminin chains ␣1, ␤1, and ␥1 and isolated from mouse EHS tumor). Different concentrations of ⌬netrin-1 and ⌬netrin-4 were incubated in the presence of 2 mM CaCl 2 with laminin-111 coated onto the well. Binding was detected in an ELISA style manner using specific antibodies directed against netrin-1 and netrin-4, respectively. The apparent K D of the ⌬netrin-4-laminin-111 interaction was 3.70 ϫ 10 Ϫ11 M. The assays showed a strong interaction between ⌬netrin-4 and native laminin-111, whereas ⌬netrin-1 did not bind to laminin-111. Abs., absorbance. Conc., concentration. After SDS-PAGE separation on a 6% gel, ⌬netrin-4 and ⌬laminin ␥1 protein were identified by immunoblotting using antibodies specific for either netrin-4 or the laminin ␥1 chain (A). A self-interaction of ⌬netrin-4 was detected, whereas the ⌬laminin ␥1 chain is not able to self-interact. Incubation with increasing amounts of BS 3 cross-linker (0 -2 mM) of both proteins at a final concentration of 2.8 M resulted in an interaction between ⌬netrin-4 and the ⌬laminin ␥1 chain (B). Using the same conditions, ⌬netrin-1 is not able to interact with the ⌬laminin ␥1 chain (C).
ism spectra of ⌬netrin-4 were recorded in the absence of CaCl 2 , in the presence of 2 mM CaCl 2 , and after the addition of excess EDTA (Table 1; supplemental Fig. 2B). In contrast to the laminin LN domains, no significant conformational changes in ⌬netrin-4 could be observed upon addition of CaCl 2 or EDTA. Apparently neither the conformation of the netrin-4 LN domain nor its interaction with the ⌬laminin ␥1 subunit requires the presence of divalent cations, suggesting a different mechanism of interaction than between laminins.
Localization of the Laminin-binding Site on Netrin-4-To identify which domain of netrin-4 interacts with the ⌬laminin ␥1 subunit, surface plasmon resonance binding studies were carried out with the N-terminal fragment of the ⌬laminin ␥1 subunit coupled to a Biacore CM5 chip and different ⌬netrin-4 deletion proteins and ⌬netrin-4-netrin-G1 hybrid proteins as analytes in solution ( Fig. 2B; Table 2). In the ⌬netrin-4 deletion proteins either the second EGF repeat, the third EGF repeat, or the LN domain were deleted, whereas in the hybrid proteins parts of ⌬netrin-4 were exchanged for the corresponding domains of netrin-G1 (Fig. 2B). The strongest interactions were detected using ⌬netrin-4 proteins containing the LN domain and all 3.5 EGF repeats (K D , 1.69 ϫ 10 Ϫ8 M) or the first two EGF domains (K D , 1.39 ϫ 10 Ϫ8 M). Furthermore, the deletion of the second EGF repeat did not appear to alter binding (K D , 5.66 ϫ 10 Ϫ8 M). Although a slightly weaker interaction occurred upon injecting a fusion protein of the netrin-4 LN domain and netrin-G1 EGF-repeats (K D , 5.82 ϫ 10 Ϫ7 M), no interaction was detected when injecting netrin-G1, a combination of netrin-G1 LN domain and netrin-4 EGF-repeats, or the isolated netrin-4 EGF repeats, suggesting that the presence of the LN domain is a prerequisite for the interaction.
Influence of Netrin-4 on the Aggregation of Laminin-111-We then asked whether the interaction of netrin-4 with laminin ␥1 short arms could have an effect upon the ability of laminin to polymerize. Samples of laminin-111 were preincubated at 37°C for 30 min to allow temperature equilibration. Addition of CaCl 2 to a final concentration of 1 mM gave rise to an immediate increase in turbidity reaching a plateau after 300 min, whereas addition of ⌬netrin-4 and 1 mM CaCl 2 significantly decreased the turbidity development to a level similar to that of samples treated with EDTA (Fig. 6A). Interestingly, addition of fulllength netrin-4 and 1 mM CaCl 2 inhibited the turbidity development to a lesser extent than ⌬netrin-4 (Fig. 6B). Addition of a molar excess of EDTA to the polymerized samples led to depolymerization and a rapid decrease in turbidity to a level approaching that of the samples that had been kept in the presence of EDTA or ⌬netrin-4 (Fig. 6B). The influence of netrins on laminin polymerization was also analyzed with a second assay. Purified laminin-111 was allowed to polymerize in the presence and absence of 10 mM EDTA or various concentrations of ⌬netrin-4 and ⌬netrin-1 at 37°C for 3 h. In the absence of EDTA, laminin-111 formed a polymer, which could after centrifugation be detected in the pellet fraction, whereas the addition of 10 mM EDTA prevented the polymerization (Fig.  6C). In the presence of ⌬netrin-4 the laminin-111 polymerization was almost completely abolished (Fig. 6D). Addition of full length netrin-4 also resulted in a decreased polymerization (results not shown). In contrast, the addition of ⌬netrin-1 had no effect on the ability of laminin-111 to polymerize (Fig. 6E).
Influence of Netrin-4 on Epithelial Branching Morphogenesis-To analyze the effect of netrin-4 on epithelial branching, the salivary gland was studied. First the expression of netrin-4 was determined. In newborn mice a weak, disorganized netrin-4 expression was observed in the sublingual part of the salivary gland, with little co-localization of netrin-4 and the laminin ␥1 subunit (supplemental Fig. 3, A-D). With increasing age netrin-4 expression takes on a network-like pattern that colocalizes with staining for the laminin ␥1 subunit (supplemental   AUGUST 17, 2007 • VOLUME 282 • NUMBER 33 Fig. 2, E-H). Strikingly, in both perinatal and adult salivary gland netrin-4 is only expressed in the sublingual and not in the submandibular part.

Netrin-4 Interacts with the Laminin ␥1 Short Arm
Since the basement membrane localization of netrin-4 is seen first after birth in the sublingual gland, embryonal glands can be used to study the effect of exogenously added netrin-4 on gland morphogenesis. Submandibular glands of day 13.5 mouse embryos were incubated with either full-length netrin-4 ( Fig. 7, C-F), ⌬netrin-4 (not shown), or ⌬laminin ␥1 for 3 days. Both kinds of netrin-4 protein caused a decrease of epithelial branching as compared with control glands not treated with protein (Fig. 7, A and B). In contrast, addition of ⌬laminin ␥1, a molecule with the same domain structure as ⌬netrin-4, which had been prepared in an identical manner, had no effect on branch-ing of submandibular glands (Fig. 7, I and J). Double immunofluorescence labeling using antibodies against netrin-4 and the laminin ␥1 subunit showed little endogenous netrin-4 expression in the control submandibular glands (Fig. 8A). The laminin ␥1 chain is located in the basement membranes around the ducts and does not colocalize with netrin-4 in untreated E13.5 submandibular glands (Fig.  8A). In the netrin-4-treated submandibular glands (Fig. 8, B and C) netrin-4 co-localizes with the laminin ␥1 chain in the basement membranes around the ducts.

DISCUSSION
Netrin-4 is structurally related to the laminin ␤ chains. It is widely expressed outside the nervous system, most abundantly in vasculature, kidney, ovary, and heart (13), and is often deposited in basement membranes. Co-localization studies in mouse kidney and salivary gland revealed overlapping spatial expression patterns of the laminin ␥1 chain and netrin-4 in basement membranes of post-natal and adult animals. However, the temporal regulation is independent, as early embryonic expression does not show extensive co-localization; indeed co-distribution is not achieved until birth. Netrin-4 expression increases from the newborn to the adult animal, and in parallel, the co-localization with the laminin ␥1 chain becomes more pronounced.
Several binding assays based on different principles revealed strong interactions between netrin-4 and both the laminin ␥1 chain and the whole trimeric laminin-111. In contrast, netrin-1 does not bind to either protein. Ca 2ϩ ions are not required for the interaction, and circular dichroism spectroscopy showed no detectable conformational change in netrin-4 induced by adding Ca 2ϩ or removing divalent cations with EDTA. Since the laminin short arm self-interaction is Ca 2ϩdependent (37), it appears that netrin-4 binding to laminin-111 does not occur through the same interaction site. Biacore binding studies showed that the netrin-4 LN domain is crucial for interaction with the ⌬laminin ␥1 subunit. We speculate that netrin-4 binds to laminin-111 via the netrin-4 LN domain, which might be folded back onto the first and second EGF repeat, since exchanging the netrin-4 EGF FIGURE 6. Self-aggregation of laminin-111. Purified laminin-111 was pre-equilibrated for 30 min at 37°C in the cuvettes of a spectrophotometer (0.67 mg/ml (A) and 0.5 mg/ml (B)). Laminin-111 (F) was incubated in the presence of 1 mM CaCl 2 and shows the typical time course of turbidity formation, which was monitored as absorbance at 360 nm. After 300 min, EDTA was added to a final concentration of 5 mM. An identical sample (E) was kept in the presence of 5 mM EDTA for the duration of the experiment. In additional samples, ⌬netrin-4 (gray triangle) (A, 0.21 mg/ml) or full-length netrin-4 (gray box) (B, 0.19 mg/ml) and CaCl 2 was added at the beginning of the measurements. In a second experiment, aliquots of laminin-111 (0.6 mg/ml) were incubated in reaction vials. After addition of 1 mM CaCl 2 , the samples were kept at 37°C for 3 h and centrifuged. Supernatant (S) and pellet (P) were analyzed by reducing SDS-PAGE and Coomassie staining. Purified laminin-111 was incubated in the presence or absence of 10 mM EDTA (C). Laminin-111 was incubated under the same conditions with increasing concentrations (1:125 to 1:1 netrin-4:laminin-111 molar ratio) of ⌬netrin-4 (D). At a 1:1 ratio of ⌬netrin-4 to laminin-111, the polymerization was completely inhibited (D, asterisk). In contrast, ⌬netrin-1 had no influence on the laminin-111 polymerization (E).
repeats for netrin-G1 EGF repeats leads to a slightly reduced binding strength.
Most studies on netrins have concentrated on their role in neuronal guidance within the central nervous system. However, they are widely expressed outside the nervous system, and their functions in these tissues have only recently begun to be addressed (for a review see Ref. 38). We demonstrated that netrin-4 is a basement membrane-associated protein (13), and it is therefore possible that netrin-4 and laminins regulate each others' activities. Recently the role of netrins in lung branching morphogenesis was explored (20). It turned out that netrin-4, as well as netrin-1, is able to suppress lung budding. For netrin-4, inhibition of lung budding was seen already at concentrations of 10 -50 g/ml, and for netrin-1 a higher concentration (50 g/ml) was needed to achieve the same level of inhibition as with netrin-4.
To study the in vivo function of netrin-4, we now focused on a different organ, the salivary gland, which also depends upon branching morphogenesis for its development. Since netrin-4 is expressed first at late stages of salivary gland development, this is an excellent model to study the effect of exogenous netrin-4. Treatment of submandibular gland explants with netrin-4 (50 g/ml) leads to a drastic suppression of epithelial branching. Significantly, this effect was achieved using the full-length netrin-4 protein and ⌬netrin-4, the truncated form of netrin-4 lacking the C domain. To ensure the specificity of netrin-4 effect, ⌬laminin ␥1 chain (50 g/ml) was added to salivary gland in an identical manner. This did not alter the epithelial branching, indicating that similarly isolated proteins do not contain inhibitory substances. Immunofluorescence staining of the treated explants revealed an accumulation of netrin-4 at the basement membrane, co-localizing with the laminin ␥1 chain. In the extracellular matrix, where levels of free Ca 2ϩ are about 1.5-2 mM, laminin forms large insoluble complexes. Remarkably, aggregation assays with laminin-111 showed that adding ⌬netrin-4 leads to a complete inhibition of laminin polymerization, whereas addition of full-length netrin-4, which forms dimers over its C domain, inhibits laminin polymerization moderately. Thus full-length netrin-4 might partially, and ⌬netrin-4 completely, destabilize the basement membrane and thereby suppresses budding of the glands. Since in embryonic lung culture netrin-1 also inhibits branching morphogenesis, even though to a much lesser extent than netrin-4, it is clear that netrin-4 has only a modulating effect on laminin polymerization. Several additional explanations for this strong inhibiting effect of netrin 4 are possible. Branching morphogenesis involves interactions between different cell types, but also the extracellular matrix, proteases, and growth factors play a role (for review see Ref. 39). Netrin-4 may act through an alteration of the basement membrane, FIGURE 7. Morphology of submandibular gland explants treated with ⌬netrin-4 or ⌬laminin ␥1 for 3 days. A, B, G, and H, controls incubated with PBS. C-F, explants treated with 50 g/ml full-length netrin-4. Incubation of the submandibular glands with netrin-4 led to a significant decrease of epithelial branching. I and J, explants treated in an independent experiment with 50 g/ml ⌬laminin ␥1 chain. Incubation of the submandibular glands with ⌬laminin ␥1 had no influence on epithelial branching. Scale bar, 0.5 mm in F and H (also applies to A-J). FIGURE 8. Distribution of the laminin ␥1 subunit and netrin-4 in submandibular gland explants treated with netrin-4. Double immunofluorescence staining using antibodies against netrin-4 (red) and the laminin ␥1 chain (green). The control submandibular glands show hardly any netrin-4 expression (A). The laminin ␥1 chain is located in the basement membrane around the ducts and does not co-localize with netrin-4 in native embryonic (E13.5) submandibular glands. In the netrin-4-treated submandibular glands (B and C), netrin-4 diffuses into the tissue and co-localizes with the laminin ␥1 subunit in the basement membrane around the ducts. Scale bar, 100 m. which can then influence the laminin signaling to the epithelial cells. On the other hand, the integration of netrin-4 into the laminin network may lead to a high local netrin-4 concentration and may therefore enhance the presentation of netrin-4 to a putative receptor. The fact that netrin-4 has an inhibitory function on branching morphogenesis and is increasingly expressed in postnatal development suggests a role for netrin-4 in late stages of embryonic development, in maturation, and in tissue homeostasis. Netrin-4 may also play a role in regulating tissue regeneration and contribute to tissue stability by preventing hyperplasia.
Further studies will aim at determining which effects of netrin-4 are mediated by modulation of basement membrane structure and which effects by its potential interactions with cell surface receptors.