|Home | About | Journals | Submit | Contact Us | Français|
Neuropilin-1 (NRP1) is a coreceptor to a tyrosine kinase receptor for both the vascular endothelial growth factor (VEGF) family and semaphorin (Sema) family members. NRP1 plays versatile roles in angiogenesis, axon guidance, cell survival, migration, and invasion. NRP1 contains three distinct extracellular domains, a1a2, b1b2, and c. We report here the identification of two novel soluble human NRP1 isoforms, which we named sIIINRP1 and sIVNRP1. These soluble NRP1 isoforms were generated by alternative splicing of the NRP1 gene, a common regulatory mechanism occurring in cell surface receptor families. Both sIIINRP1 and sIVNRP1 contain a1a2 and b1b2 domains, but no c domain, and the rest of the NRP1 sequence. Additionally, sIIINRP1 is missing 48 amino acids within the C-terminus of the b2 domain. Both sIIINRP1 and sIVNRP1 are expressed in human cancerous and normal tissues. These molecules are capable of binding to VEGF165 and Sema3A. Furthermore, recombinant sIIINRP1 and sIVNRP1 proteins inhibit NRP1-mediated MDA-MB-231 breast cancer cell migration. These results indicate the multiple levels of regulation in NRP1 function and suggest that these two novel NRP1 isoforms are useful antagonists for NRP1-mediated cellular activities.
Neuropilin-1 (NRP1),1 first characterized in Xenopus, is a cell surface receptor that is important in both growth of blood vessels and guidance of neuronal axons [1–3]. The role of NRP1 in these processes is mediated by its interaction with numerous ligands, most notably, vascular endothelial growth factor (VEGF) in angiogenesis and semaphorin (Sema) in neuronal guidance [4,5]. NRP1 has no known enzymatic activity and therefore participates in signal transduction events through the complex formation with tyrosine kinase receptors. To carry out its axon guidance function, NRP1 binds to Sema3A or other class III semaphorins and forms a complex with a coreceptor of the plexin family to form a functional Sema3 receptor [6,7]. To promote vessel growth, NRP1 interacts with several members of the VEGF family, including the 165-amino-acid isoform of VEGF-A (VEGF165), VEGF-B, VEGF-E, and placental growth factor 2 but not diffuse VEGF-A isoform, VEGF121 [5,8–10]. The biological effects of VEGF are mediated by a family of VEGF tyrosine kinase receptors, Flt-1 (VEGFR-1), Flk-1/KDR (VEGFR-2), and Flt-4 (VEGFR-3) [11,12]. NRP1 has been shown to enhance the binding of VEGF165 to VEGFR-2 and increase endothelial cell chemotaxis . The interaction of NRP1 with its various ligands and coreceptors has been shown to be important for the development of the nervous and vascular systems. NRP1-deficient mice die because of severe defects in the development of the cardiovascular system, vascularization, and peripheral nervous system [13,14]. NRP1 also plays important roles in tumor angiogenesis, growth, survival, and invasiveness . For example, overexpression of NRP1 by tumor cells increased angiogenesis and tumor growth in a rat prostate carcinoma model . Autocrine VEGF165 was shown to increase cell survival in MDA-MB-231 breast carcinoma cells . Treatment of MDA-MB-231 cells by an anti-NRP1 antibody or VEGF antisense oligonucleotides inhibited cell migration and invasion induced by the conditioned medium of fibroblast (NIH 3T3) cells .
NRP1 is known to exist as a membrane-bound form (full-length) as well as two soluble isoforms. The NRP1 gene consists of 17 exons and 16 introns. The full-length form of the protein contains all 17 exons, whereas two previously reported soluble NRP1 isoforms, s12NRP1 and s11NRP1, are created by reading through into introns [19,20]. The extracellular domain of full-length NRP1 is composed of CUB motifs homologous to complement components C1r/C1s (also called a1 and a2 domains), two domains that are homologous to coagulation factors V and VIII (also called b1 and b2 domains), and a Meprin A5 μ-phosphatase domain (also called the c domain) [21,22]. The 40-amino-acid positively charged carboxyl-terminal basic domain does not display homology to any known proteins . The mRNA for s12NRP1 contains the first 12 exons of the gene followed by a sequence derived from intron 12 and a polyadenosine (poly(A)) tail. The protein encoded by this mRNA contains the a1a2 and b1b2 domains and most of the b/c linker followed by 3 novel amino acids . The mRNA for s11NRP1, another soluble isoform, contains the first 11 exons of the gene followed by a sequence derived from intron 11 and a poly(A) tail. The resulting protein contains the a1a2 and b1b2 domains, followed by the portion of the b/c linker encoded by exon 11 and 83 novel amino acids . Full-length NRP1 is expressed in endothelial cells and tumor cells [5,20], whereas s12NRP1 and s11NRP1 are expressed in a variety of tissues but not in endothelial cells [19,20]. s12NRP1 and s11NRP1 are thought to function as VEGF165 antagonists . s12NRP1 has been shown to inhibit VEGF165 binding to cells expressing NRP1 and also inhibit VEGF165-induced tyrosine phosphorylation of VEGFR-2 in vitro. When overexpressed in a rat prostate carcinoma model, s12NRP1 caused a higher percentage of apoptotic cells, hemorrhaging within the tumors, and a lower number of blood vessels . The antagonistic effects of s12NRP1 can be explained by the domains of the protein responsible for the competitive binding of various NRP1 ligands. Recent studies have shown that the a1a2 and b1 domains are required for Sema3A binding . The VEGF165 binding site is located in the b1b2 domains . The s12NRP1 and s11NRP1 isoforms contain the entire a1a2 and b1b2 domains but not the c domain. Therefore, both soluble NRP1 proteins have the protein structure necessary to bind and sequester VEGF165. Although the physiological effects of s11NRP1 have not been studied, the protein would be expected to have biochemical and physiological properties similar to those of s12NRP1 because of the domains present .
It seems possible that there could exist other soluble NRP1 isoforms that have not been characterized. Indeed, neuropilin-2 (NRP2), the homologue of NRP1, has a soluble isoform that is different in cDNA and protein structure from both s12NRP1 and s11NRP1 . In this study, we report the discovery and characterization of two novel soluble NRP1 isoforms that we named sIIINRP1 and sIVNRP1. These two new NRP1 isoforms are naturally occurring and are generated by alternative splicing. The sIIINRP1 and sIVNRP1 molecules displayed binding capacities to Sema3A and VEGF165 similar to that of s12NRP1. Furthermore, the soluble NRP1 isoforms including s12NRP1 are able to inhibit the full-length NRP1-mediated MDA-MB-231 breast carcinoma cell migration.
As depicted in Fig. 1A, previous studies report that human NRP1 exists as a full-length membrane receptor and two naturally occurring soluble receptors, s12NRP1 and s11NRP1, both generated by reading through into the intron of the NRP1 gene [19,20]. To determine whether there are other naturally occurring NRP1 isoforms, we searched the human EST clone database at the NCBI Web site and discovered two classes of sequences that encode two types of novel soluble NRP1 isoforms apparently generated by alternative splicing. One class of alternatively spliced soluble NRP1 EST clones contain full-length sequences upstream of the beginning of exon 10 (1614 of the full-length sequence). Exons 10 and 11 are deleted (position 1614 to 1864 of the full-length sequence), while exon 12 (position 1864 to 1924) is present. After exon 12, a sequence of 28 bp identical to the beginning of NRP1 intron 12 is present. This class of EST clones will be referred to as sIIINRP1 (Fig. 1B, top). The protein produced from this class of EST clones would contain all amino acids derived from exons 1–9 of the gene. It is missing the last 48 amino acids of the b2 domain and all domains between the b2 domain and the C terminus but contains 13 novel amino acids derived from a different reading frame of exon 12. The stop codon TAG was discovered within the exon 12 sequence (position 1904 of the full-length NRP1 sequence). Four different EST clones of this class were identified in the human EST clone database, Accession Nos. AA976114 derived from lung neuroendocrine carcinoid, AA987327 and AA973607 derived from lung carcinoid, and AI953301 derived from ovary normal epithelial tissue. Clones AA976114 and AI953301 were purchased from Incyte Genomics, Inc. (St. Louis, MO, USA) and completely sequenced. AA976114 was over 1 kb in length. Homology with the full-length NRP1 began at position 699 of the full-length NRP1 open reading frame (ORF) sequence. Thirty-two base pairs identical to NRP1 intron 12 were present before the poly(A) signal. In AI953301, 27 bp were present before the poly(A) signal. Thus, sequencing confirmed that a soluble NRP1 sequence missing exons 10 and 11 was present in EST clones derived from two different tissue types (lung neuroendocrine carcinoid and ovary normal epithelium).
Another class of alternatively spliced soluble NRP1 EST clones, which will be referred to as sIVNRP1, was also found (Fig. 1B, bottom). In this class, NRP1 sequence was present before exon 11 (ORF position 1759) but not as far back as the stop codon. Exon 11 (position 1759 to 1864) was missing. Exon 12 was present (position 1864 to 1924). No exons after exon 12 were present. After exon 12, 28 bp identical to the beginning of intron 12 were present. The protein produced by this EST clone would be identical to full-length NRP1 for the first 586 residues (encoded by exons 1–10), would be missing the 35 amino acids encoded by exon 11, but would include the wild-type amino acids encoded by exon 12, followed by the sequence G I K derived from intron 12. The splicing out of exon 11 does not change the reading frame. The last 23 amino acids are the same as in the previously reported soluble NRP1 isoform, s12NRP1 . Two EST clones of this class were found: Accession No. BG120342 derived from a liver adenocarcinoma cell line and Accession No. BF939225 from adult prostate. An agar stab of bacterial transformed BG120342 was purchased. This clone was sequenced completely from the 5′ end, and its authenticity was thus verified.
To determine whether the two novel soluble NRP1 isoforms are expressed in human normal or cancerous tissues, we first performed RT-PCR analyses using PCR primers to amplify sIIINRP1, sIVNRP1, or s12NRP1 cDNA in the same reaction. The expression of all three isoforms was observed in normal human brain tissue, human glioma cell lines (Fig. 2A, top), and a human lung adenocarcinoma cell line (data not shown). The intensity of the bands for sIIINRP1 and sIVNRP1 was consistently less than the band corresponding to s12NRP1 (3- to 10-fold less than that of s12NRP1 quantified using the expression level of β-actin for normalization), indicating low abundance of these two NRP1 isoforms in various tissues and cell lines. Next, the expression of sIIINRP1 and sIVNRP1 was further examined using primers specific for each isoform. sIIINRP1 mRNA was consistently expressed in human glioma cell lines and normal brain (Fig. 2A, bottom), human glioma tissues at various WHO grades (Fig. 2B), human lung adenocarcinoma tissue, 784T (a human lung adenocarcinoma cell line), and 128-88T (a human squamous carcinoma cell line) (data not shown). The expression of sIVNRP1 mRNA was found in most normal and cancerous human tissues that express s12NRP1 and sIIINRP1 isoforms (Fig. 2C and data not shown). In addition to those tissues shown, sIVNRP1 expression was also observed in normal human spleen and pancreas. The authenticity of the PCR product for each NRP1 isoform was completely verified by automated sequencing after subcloning the PCR products into a TA cloning vector (TOPO 2.1).
We performed Western blot analyses to determine the protein expression of s12NRP1, sIIINRP1, and sIVNRP-1 in total cell lysates prepared from various glioma and lung cancer cell lines shown in Fig. 2A using an anti-NRP1 antibody (N-18; Santa Cruz Biotechnology) that was raised against a peptide epitope at the N-terminus of NRP1 protein and recognizes both full-length and various soluble forms of NRP1 proteins. However, we could not detect the expression of the three NRP1 isoform proteins or full-length NRP1 protein in various human tumor cell lines. We also examined the expression of full-length NRP1 protein in these cells by Western blot analyses using several other anti-NRP1 antibodies that recognize full-length NRP1 protein (A-12 and C-19 from Santa Cruz Biotechnology, Ab-1 from Oncogene Science, or NP1ECD1A ). No NRP1 proteins could be detected by Western blot assays. This was probably due to protein expression in the cells that was below the detectable levels of the anti-NRP1 antibodies.
Next, after assembling each novel soluble NRP1 isoform into a mammalian expression vector (pcDNA 3.0) and completely verifying the DNA sequences of various expression constructs, we determined the expression of various soluble NRP1 isoforms in mammalian cells. We separately transfected COS-7 cells with expression vectors encoding s12NRP1, sIIINRP1, or sIVNRP1 proteins and assessed the expression of the soluble NRP1 isoforms using Western blot analyses. As shown in Fig. 3, the NRP1 isoform proteins were expressed in the COS-7 cells as s12NRP1 (90 kDa), sIIINRP1 (60 kDa), or sIVNRP1 (67 kDa).
We next determined whether the novel NRP1 isoforms are capable of binding the NRP1 ligand VEGF165. Recent studies have shown that a recombinant membrane-bound NRP1 variant with a deleted b2 domain had decreased VEGF165 binding with no effects on Sema3A binding . However, another recombinant NRP1 variant composed of the b1b2 domain only and missing 30 amino acids at the C terminus of the b2 domain had no binding affinity for VEGF165 . Interestingly, sIIINRP1 contains a1a2 and b1b2 domains missing 48 amino acids at the C terminus of the b2 domain, whereas sIVNRP1 has intact a1a2 and b1b2 domains. Therefore, we hypothesized that sIIINRP1 may have decreased VEGF165 binding and that sIVNRP1 would have unaffected binding for VEGF165. As expected, similar to s12NRP1, sIVNRP1 is capable of binding VEGF165 (Fig. 4A). However, sIIINRP1 was also found to be able to bind VEGF165 (Fig. 4A). In addition, we assessed the abilities of s12NRP1, sIIINRP1, and sIVNRP1 to bind Sema3A by performing similar assays. Since sIIINRP1 and sIVNRP1 contain a1a2 and b1 domains necessary for binding to Sema3A, as expected from previous studies , both soluble NRP1 isoforms were capable of binding Sema3A (data not shown).
To determine whether the binding affinity for VEGF165 is different for sIIINRP1 relative to that displayed by s12NRP1, we performed a semiquantitative binding assay using normalized concentrations of each soluble NRP1 isoform. Various concentrations (12.0, 6.0, 2.4, or 1.2 nM) of sIIINRP1 or s12NRP1 were mixed with VEGF165-AP at a fixed VEGF165-AP protein concentration of 3.0 nM [25,27]. Soluble NRP1 bound to VEGF165-AP was coimmunoprecipitated with an anti-AP antibody conjugated to agarose beads and analyzed by Western blot analysis using an anti-NRP1 antibody (N-18; Santa Cruz Biotechnology) reactive against all soluble NRP1 isoforms. There was no significant difference in binding for VEGF165-AP between sIIINRP1 and s12NRP1 (Fig. 4B). Similar amounts of s12NRP1 and sIIINRP1 proteins were coprecipitated with the VEGF165-AP fusion proteins at each dilution. Neither s12NRP1 nor sIIINRP1 was detected at 1.2 nM dilution (data not shown). We also found that VEGF165 binding for sIVNRP1 was similar to that displayed by s12NRP1 (data not shown). Thus, these data suggest that although the sIIINRP1 and sIVNRP1 isoforms lack certain amino acids present in s12NRP1, these two novel soluble isoforms have similar capacities for binding to their cognate ligands.
To evaluate further the binding properties of the novel soluble NRP1 isoforms, we performed competition assays to determine whether each soluble NRP1 isoform can inhibit VEGF165-AP binding to cell surface VEGF receptors (Figs. 5A and 5B). Human MDA-MB-231 breast carcinoma cells were used since these cells lack VEGFR-2  but express NRP1 at high levels on their surface [5,17]. As shown in Fig. 5A, the binding of VEGF165-AP to the cells measured by AP activity was concentration-dependent in the assay. However, when various soluble NRP1 isoforms and Ig-Basic-AP (a negative control) are separately included in the binding assays, the novel soluble NRP1 isoforms, but not Ig-Basic-AP, prevented VEGF165-AP from binding to the cells to an extent similar to that displayed by s12NRP1 (Fig. 5B). The amount of AP activity bound to the MDA-MB-231 cells as a percentage of the control was 12, 13, and 24% for s12NRP1, sIVNRP1, and sIIINRP1, respectively, at the highest concentration of soluble NRP1 protein (24.0 nM). These data suggest that sIIINRP1, sIVNRP1, and s12NRP1 are capable of competing with VEGF receptors for VEGF165 binding.
Since the novel sIIINRP1 and sIVNRP1 isoforms can bind and sequester VEGF165 as effectively as s12NRP1 does, we postulated that they would be able to antagonize the biological functions of full-length NRP1 in promoting cell migration. It has been previously reported that autocrine VEGF165 signaling is responsible for the invasive behavior of MDA-MB-231 breast carcinoma cells. The proposed mechanism is that autocrine VEGF165 prevents the MDAMB-231 cells from entering apoptosis and promotes NRP1-mediated cancer cell migration . We determined whether the two novel soluble NRP1 isoforms inhibit NRP1-mediated human breast cancer cell migration. As shown in Fig. 6, the two novel soluble NRP1 isoforms, but not Ig-Basic-AP, were able to inhibit MDA-MB-231 cell migration toward NIH 3T3 CM to an extent similar to that of s12NRP1. MDAMB-231 cell migration was inhibited 51% by s12NRP1, 52.5% by sIVNRP1, and 55.5% by sIIINRP1, respectively, when calculated as a percentage of the control at the highest concentration of soluble NRP1 protein. These data demonstrate that the two novel soluble NRP1 isoforms can act to sequester VEGF165, thus inhibiting autocrine-mediated breast cancer cell migration.
NRP1 plays critical roles in neuronal guidance, cardiovascular development, angiogenesis, and tumor progression. Studies show that regulation of NRP1 function in these processes is at multiple levels. One mechanism is the modulation of NRP1 gene expression. The expression of full-length NRP1 has been found in normal and cancerous tissues and in endothelial, hematopoietic, and cancerous cell lines . NRP1 is regulated by tumor necrosis factor-α  or VEGF  in EC and by epidermal growth factor in human gastric or pancreatic cancer cells [31,32]. It is likely that the regulation of NRP1 by these growth factors is mediated by transcription factors such as Sp1 . Another mechanism for the regulation of NRP1 function is the expression of naturally occurring soluble NRP1 isoforms, which often act as antagonists. Klagsbrun and colleagues have previously reported two naturally occurring soluble NRP1 isoforms, s12NRP1 and s11NRP1. Both s12NRP1 and s11NRP1 are generated during pre-mRNA processing by reading through into the introns of the NRP1 gene [19,20]. The s12NRP1 and s11NRP1 isoforms are expressed in multiple normal human tissues , whereas s12NRP1 was also found in cancerous tissues [5,15]. Our results show that two novel NRP1 isoforms, sIIINRP1 and sIVNRP1, are produced by an alternative splicing mechanism that often occurs in cell surface receptors such as sFlt-1 , NRP2 , and FGFR . Although sIIINRP1 and sIVNRP1 were initially found in several human EST clones through searching the NCBI human EST database, these novel isoforms are indeed naturally occurring. sIVNRP1 was present in most normal human and cancerous tissues that were analyzed, whereas sIIINRP1 was detected only in cancerous and normal brain tissues and not in other normal human tissues that were analyzed. However, our semiquantitative RT-PCR analyses using β-actin as an internal control showed that the expression levels of sIIINRP1 and sIVNRP1 were lower than that of s12NRP1 in the tissues and cells analyzed (Fig. 2A). We detected the protein synthesis of full-length NRP1 in human U87MG glioma and MDA-MB-231 breast cancer cells using an anti-NRP1 antibody against C-terminal amino acids (clone C-19; Santa Cruz Biotechnologies, Santa Cruz, CA, USA). However, this antibody and several other anti-NRP1 antibodies including commercially available and two well-characterized anti-NRP1 antibodies [4,16] could not detect the expression of s12NRP1, sIIINRP1, and sIVNRP1, probably due to the lack of C-terminal amino acid sequences in these soluble NRP1 isoforms or low sensitivities of the antibodies in this type of analyses (data not shown). This suggests that the protein expression of these soluble NRP1's was below the detectable levels by direct Western blot analyses. Low levels of expression of sIIINRP1 and sIVNRP1 isoforms also suggest that these two soluble NRP1 molecules may not achieve functional inhibitory levels under physiological and pathological conditions. This would also undermine the relevance of these NRP1 splice variants in various pathophysiological processes. Moreover, we found that the expression level of sIIINRP1 mRNA was higher in GBM (WHO grade IV) than in lower grade gliomas when the expression level of sIIINRP1 was normalized to the expression of β-actin in RT-PCR analyses (Fig. 2B and data not shown). At present, we cannot explain why the level of sIIINRP1 mRNA in malignant GBM was higher than that in their less aggressive counterparts. A possible explanation is that in GBM increased levels of VEGF expression are much higher than the elevated levels of its antagonists such as sFlt-1 and sNRP1. Active tumor angiogenesis has otherwise been ensured by predominant bioavailable VEGF proteins within the tumors during the glioma progression .
The binding capacity of NRP1 molecules to Sema3A and VEGF has been studied in great detail. It has been demonstrated that the extracellular domains of the NRP1 molecule, a1a2 and b1, bind to Sema3a, while the b1b2 domains are the docking site for VEGF165 [37–39]. Recently, two reports further dissected the binding properties of NRP1 to VEGF165 using recombinant NRP1 variants that contained various extracellular domains [24,25]. While one study showed that the a1a2 and b1 domains are required for Sema3A binding , the other investigation demonstrated that the VEGF165 binding site is located in the b1b2 domains . However, there were slightly different observations between these two studies in terms of VEGF165 binding. When the b2 domain was deleted in a recombinant soluble NRP1 that has only a1a2 and b1 domains, no VEGF165 binding was found . In contrast, when the b2 region was removed in a membrane-bound NRP1 variant that contained a1a2, b1, and c domains, the capacity of binding to VEGF165 was only partially reduced . This suggests that although the a1a2 domain by itself does not bind to VEGF165, the a1a2 domain is able to modulate the binding of VEGF165 to b1b2. This conclusion was supported by the observation that the a1a2/b1b2 recombinant proteins were 50 times more potent than the b1b2 domain alone in inhibiting VEGF165 binding to cell surface receptors. A truncated b1b2 domain lacking 30 amino acids from its C-terminus and also missing the a1a2 domains did not bind to VEGF165 . The binding properties of sIIINRP1 and sIVNRP1 proteins to Sema3A and VEGF165 assessed under the indicated conditions (Figs. 4 and and5)5) corroborate these conclusions. Despite slightly differing structures compared to previously discovered isoforms, the novel sIIINRP1 and sIVNRP1 isoforms have biochemical and biological properties similar to those displayed by s12NRP1 and s11NRP1 molecules. Both sIIINRP1 and sIVNRP1 bind to Sema3A and VEGF165 with affinities similar to those displayed by the s12NRP1 protein containing the entire a1a2 and b1b2 domains . Although the sIIINRP1 protein has more amino acids missing (48 aa) from its b2 domain C-terminus than the truncated b1b2 NRP1 variant lacking 30 amino acids and also the a1a2 domain , the sIIINRP1 isoform binds VEGF165 at a capacity similar to that of sIVNRP1 as well as s12NRP1 proteins. This may suggest that the a1a2 domains present in sIIINRP1 indeed contribute to VEGF165 binding. Moreover, due to expression levels of sIIINRP1 and sIVNRP1 proteins in our experimental system, these results may not accurately reflect the binding affinities of the various NRP1 splice variants.
Previous studies have also shown that either naturally occurring s12NRP1 isoform or a recombinant a1a2 b1b2 NRP1 variant could inhibit VEGF165 binding to cell surface VEGF receptors and suppress the stimulatory activities of VEGF165 on cell migration and proliferation [20,25]. Since our data suggest that sIIINRP1 and sIVNRP1 are capable of binding to VEGF165, we predicted that the presence of either of these novel NRP1 isoforms would inhibit NRP1-mediated cell migration. The inhibition of human MDAMB-231 breast carcinoma cell migration by sIIINRP1 and sIVNRP1 supports this hypothesis. Furthermore, since the MDA-MB-231 cells express only NRP1 and VEGFR-1 at their cell surface [17,18], the presence of various soluble NRP1 isoforms suppressed NRP1-mediated breast cancer migration.
One of the important functions of soluble cell surface receptors is to sequester the cognate ligands of the full-length receptor, thus inhibiting the physiological processes mediated by the full-length receptor. For example, when overexpressed in a rat prostate carcinoma AT2.1 cell model, s12NRP1 has been shown to increase apoptosis and decrease blood vessel number and integrity, thus strongly inhibiting rat prostate tumor angiogenesis and progression . We attempted to demonstrate that sIIINRP1 and sIVNRP1 have similar inhibitory effects displayed by s12NRP1 on tumor angiogenesis and growth. We separately introduced s12NRP1, sIIINRP1, and sIVNRP1 isoform or Ig-Basic-AP cDNAs into different types of mammalian expression vectors and expressed them in human U87MG glioma cells individually to determine their inhibitory functions on angiogenesis and tumor growth in our U87MG glioma xenograft models in nude mice [18,26,40,41]. However, although strong expression of each of these soluble NRP1 proteins was detected in the transfected U87MG cells 3 days after the transfection, we failed to isolate any drug-resistant U87MG cell clones that stably expressed one of these soluble NRP1 isoforms. Because NRP1 has been shown to mediate autocrine survival effects displayed by VEGF , it is possible that overexpression of one of the three soluble NRP1 isoforms in U87MG gliomas cells disrupted such antiapoptotic function mediated by the full-length NRP1 receptor, thus causing the death of U87MG glioma cells that expressed exogenous soluble NRP1 isoform proteins.
In summary, we have identified, expressed, and initially characterized two novel alternatively spliced soluble human NRP1 isoforms that we have designated sIIINRP1 and sIVNRP1. These two NRP1 spliced isoforms were initially found in the NCBI human EST database and they are naturally occurring in normal human and cancerous tissues. In addition, our data show that despite slight structural differences in the b2 domain and linker region between sIIINRP1, sIVNRP1, and the previously reported s12NRP1 isoform, the three NRP1 variants showed similar abilities to bind to their cognate ligands, Sema3A and VEGF165, and inhibited NRP1-mediated MDA-MB-231 breast cancer cell migration. Our results demonstrate that sIIINRP1 and sIVNRP1 are functional antagonists against VEGF receptors, suggesting that these soluble NRP1 molecules could be useful pharmacological entities to regulate angiogenesis and neuronal development during physiological and pathological processes.
Human glioma cell lines U251MG, U87MG, U373MG, and T98G; human embryonic kidney (HEK) 293T; human breast carcinoma MDA-MB-231 cells; and COS-7 cells were from American Type Culture Collection (ATCC; Manassas, VA, USA). Human glioma cell lines LNZ-308, LN18, LN235, LN428, and LN443 were previously described . COS-7 cells were cultured in DMEM with glutamine (2.0 mM; Mediatech, Herndon, VA, USA), 10% FBS (Sigma, St. Louis, MO, USA), and 1% penicillin/streptomycin at 37°C in 5% CO2. Glioma cell lines, HEK 293T cells, and MDA-MB-231 cells were cultured under the same conditions except that FBS from HyClone (Salt Lake City, UT, USA) was used. The following reagents were used for this study: goat polyclonal anti-NRP1 antibody (N-18, 1:300), mouse monoclonal anti-NRP1 antibody (A-12, 1.0 μg/ml; Santa Cruz Biotechnology), rabbit anti-NRP1 antibody (Ab-1, 1:500; Oncogene Sciences, San Diego, CA, USA), rabbit anti-NRP1 antibody (NP1ECD1A ), rabbit anti-goat immunoglobulins conjugated to horseradish peroxidase (1:500, P0449; DAKO, Carpinteria, CA, USA), monoclonal anti-human placental AP antibody conjugated to agarose beads (A2080; Sigma); AP Assay Reagent A (Q501), cell lysis buffer (Q504; GenHunter, Nashville, TN, USA), FastStart Taq DNA polymerase (Roche, Indianapolis, IN, USA), Lipofectamine PLUS, Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA), total RNA from normal human tissues (Clontech, Palo Alto, CA, USA), VEGF165-AP/pCEP4 DNA construct from Dr. J. Raper, University of Pennsylvania (Philadelphia, PA, USA), Sema3A-AP/pCEP4 DNA construct from Dr. A. Kolodkin, Johns Hopkins University (Baltimore, MD, USA). All other reagents were from Invitrogen, Sigma, or Fisher Scientific (Hanover Park, IL, USA).
The NCBI human EST clone database was searched by BLAST using the entire coding sequence of full-length NRP1. We looked for EST clone sequences encoding human NRP1 cDNA and also containing polyadenylation signals. Six clones with the following accession numbers were identified: BF939225, BG120342, AA976114, AI953301, AA987327, AA973607. The cDNA for the discovered IMAGE clones with Accession Nos. AA976114, AI953301, and BG120342 were purchased in bacteria transformed agar stabs from Incyte Genomics, Inc. (St. Louis, MO, USA) or ATCC. DNA from each EST clone was confirmed by full-length automated sequencing.
First, full-length NRP1 was cloned by RT-PCR from the total RNA isolated from human U87MG glioma cells. Two RT-PCR products containing NRP1 bases 1–935 and 880–2772 were produced and then TA separately cloned into a TOPO 2.0 vector (Invitrogen). To assemble the full-length NRP1 sequence, NRP1 bases 1–904, cut with BamHI and PstI, and NRP1 bases 905–2772, cut with PstI and SacI, were ligated into the BamHI/SacI cut of the pBluescript SK(−) vector (Stratagene, San Diego, CA, USA). PstI was used to cut at the naturally occurring site at 904 of NRP1. To assemble a s12NRP1 construct, the sequence between base 880 and the end of exon 12 (1924) was amplified by PCR, with the downstream primer introducing the intron-derived amino acids (GIK), a stop codon, and a SacI and an EcoRI site. Bases 1–904 digested with BamHI and PstI and the PCR product of the C-terminal portion of s12NRP1 digested with PstI and SacI were then ligated into the BamHI/SacI cut of pBluescript SK(−) to assemble the complete s12NRP1 sequence. Both sIIINRP1 and sIVNRP1 were then assembled. For each isoform, the 5′ portion of full-length NRP1 was added since the EST clones did not contain the 5′ portion of NRP1. For sIIINRP1, AA976114 was used for cloning. The construct was first assembled in pBluescript SK(−) by PCR using primers upstream, 5′-CTTCCCAGTATAGCACC-3′ (880–896), and downstream, 5′-TATAGAATTCTGTCACATTTCGTATTTTA-3′, specific for intron 12 sequence and including an EcoRI restriction site. The PCR product digested with PstI and EcoRI and the first 904 bases of the full-length NRP1 digested with BamHI and PstI were then ligated together into a BamHI/EcoRI-digested pBluescript SK(−). sIVNRP1 was also first assembled in pBluescript SK(−). BG120342 was first cut with SalI and NotI, followed by BspHI, at the naturally occurring site at 1531 of the NRP1 full-length sequence. This digestion yielded a fragment containing the 3′ region of sIVNRP1 (the alternatively spliced region) including the poly(A) stretch. The sIVNRP1 construct was assembled by ligating the pBluescript SK(−) vector digested by BamHI and NotI, the first 1531 bp of NRP1, and the EST clone sequence containing the 3′ region of sIVNRP1 cDNA. After assembly, full-length NRP1, s12NRP1, sIVNRP1, and sIIINRP1 were subcloned into a pcDNA3.0 vector (Invitrogen). Later, to remove the poly(A) tail included in the EST clone and increase mammalian expression, sIVNRP was resubcloned into the pcDNA 3.0 vector. This was done using PCR amplification as was described in the assembly of sIIINRP1. DNA sequences of all four expression constructs, full-length NRP1, s12NRP1, sIVNRP1, and sIIINRP1 in pcDNA3.0, were completely verified by automated sequencing. No changes in amino acid sequences of the various NRP1 proteins were found.
Total RNA was isolated from various snap-frozen human tumor tissues using the TRIZOL reagent (Invitrogen/Life Technologies). Human glioma tissues were from Dr. R. Nishikawa, Saitama Medical School (Saitama, Japan). Human lung cancer tissues were from Dr. J. Siegfried, University of Pittsburgh Cancer Institute. Identities and grades of glioma and lung cancer tissues were separately verified using hematoxylin and eosin-stained tumor tissue sections by pathologists at Saitama Medical School and the University of Pittsburgh. Total RNA from various normal human tissues was purchased from Clontech. Reverse transcriptase reactions were performed with RT and without RT. PCR was performed on the various RT reactions to determine the expression of each isoform. Primers were designed either to be specific for each isoform or to amplify s12NRP-1, sIIINRP-1, and sIVNRP-1 in the same reaction. The downstream primer (5′-TATAGAATTCTGTCACATTTCGTATTTTA-3′, matching from 9 to 28 bp from the 5′ end of intron 12) was used for all PCR. The upstream primer was used to cause the specificity of each reaction. To amplify s12NRP-1, sIIINRP-1, and sIVNRP-1 simultaneously, the upstream primer (5′-GTGTTCATGAGGAAGTTC-3′, nucleotide positions 1528–1545) was used (open reading frame position 1528–1545, within exon 9). To amplify sIIINRP-1 specifically, the upstream primer (5′-GCAAACGCAAGGCGAAGGT-3′, 5′ end at NRP1 open reading frame position 1597) was designed across exons 9 and 12 with 17 bp in exon 9 and 2 bp in exon 12. The identity of the 114-bp sIIINRP-1 PCR product was completely verified by cloning into a TOPO 2.1 vector (Invitrogen) and automated sequencing. To amplify sIVNRP-1 specifically, the upstream primer (5′-GGGCTGTGAAGTGGAAGGTG-3′, 5′ end at NRP1 open reading frame position 1742) was designed across exons 10 and 12, with 17 bp in exon 10 and 3 bp in exon 12. Primers for human β-actin (5′-CGGGAAATCGTGCGTGACAT-3′, upstream, and 5′-GGAGTTGAAGGTAGTTTCGTG-3′, downstream) were included in some reactions at a 15-fold lower primer concentration than NRP1 primers. PCR conditions for reactions amplifying all three isoforms and reactions specific for sIIINRP-1 were 95°C for 4 min, (95°C for 30 s, 56°C for 60 s or 58°C for 50 s, 72°C for 30 s) for 40 cycles and for sIVNRP-1, 95°C for 4 min (95°C for 1 min, 58°C for 1 min, 72°C for 1 min) for 35 cycles and then 72°C for 10 min. To quantify the relative levels of gene expression among different NRP1 isoforms, the captured images of gel electrophoresis were imported into the Image Pro Plus program (version 4.1, Media Cybernetics, Silver Spring, MD, USA) and analyzed.
The proteins of cell lysates or CM were separated on 7.5% SDS polyacrylamide gels and then transferred to a nitrocellulose membrane. The rinsed and blocked membranes were then incubated with a goat polyclonal anti-NRP1 antibody (1:300, N-18; Santa Cruz) at room temperature for 1 h. The blot was washed and probed with a rabbit anti-goat antibody, conjugated with horseradish peroxidase (DAKO), at room temperature for 1 h. The blot was washed again and developed with enhanced chemiluminescence reagents (Amersham, Piscataway, NJ, USA).
sIIINRP1, sIVNRP1, and s12NRP1 constructs in pcDNA 3.0 vectors were separately transfected into COS-7 cells using the Lipofectamine PLUS reagent. The cells were lysed 48 h posttransfection. Expression of the expected sized proteins was confirmed by immunoblotting. To collect CM containing secreted proteins used in binding experiments, expression constructs of sIIINRP1, sIVNRP1, and s12NRP1 as well as Sema3A-AP, VEGF165-AP, or Ig-Basic-AP fusion proteins were transfected into HEK 293T cells separately by calcium phosphate precipitation as previously described . The medium was changed after transfection with serum-containing medium. For serum-containing CM, the CM was collected after the cells were allowed to grow for an additional 48 h. For serum-free CM, the cells were allowed to recover for 24 h in the presence of serum and then grown for an additional 48 h in medium without serum.
The binding capacities of sIIINRP1, sIVNRP1, and s12NRP1 to Sema3A-AP or VEGF165-AP were determined by immunoprecipitation for the ligand (Sema3A-AP or VEGF165-AP) followed by immunoblotting for the soluble NRP1 receptors. To assess abilities of binding to each ligand, equal volumes of the various soluble NRP1's containing CM (with serum) were mixed with VEGF165-AP, Sema3A-AP, or Ig-Basic-AP containing CM (with serum). As negative controls, soluble NRP1 containing CM was mixed with CM from mock (no DNA expression construct)-transfected cells. Protease inhibitors (100 μg/ml PMSF, 1.5 μg/ml leupeptin, 1.0 μg/ml pepstatin A, 1.0 μg/ml aprotinin) were added to each mixture. The mixtures of the CM were rotated at 4°C for 4 h for VEGF165-AP binding or 6 h for Sema3A-AP binding. Then, 50 μl of a 50/50 slurry of an anti-human placental AP antibody conjugated to agarose beads was added. The CM was rotated for an additional 6 h at 4°C for VEGF165-AP binding or 12 h at 4°C for Sema3A-AP binding. The beads were spun down and then washed with 1 ml of cold PBS. The samples were then eluted in sample buffer, separated on a 7.5% SDS–polyacrylamide gel, and immunoblotted for NRP1 as described above.
To perform semiquantitative binding assays of sIIINRP1 or sIVNRP1 and VEGF165, CM containing VEGF165-AP was added at a final concentration of 3.0 nM. Recombinant s12NRP1, sIIINRP1, sIVNRP1, or Ig-Basic-AP proteins were added separately at concentrations of 12.0, 6.0, 2.4, and 1.2 nM. CM from mock-transfected cells was used in negative controls instead of VEGF165-AP CM and also added to bring the volume up to 6 ml in other reactions. CM was rotated, immunoprecipitated, and immunoblotted as described above. The amount of the recombinant sIIINRP1, sIVNRP1, s12NRP1, or Ig-Basic-AP protein in each CM was normalized by SDS–polyacrylamide gel electrophoresis, followed by Coomassie blue staining. A known amount of bovine serum albumin was used as a standard for protein quantification. The concentration of VEGF165-AP in the CM was calculated by assaying the AP activity with Gen-Hunter AP Assay Reagent A, yielding a concentration in U/ml, and then using the activity of pure human placental AP of 1500 U/mg to calculate the VEGF165-AP molar concentration. One unit of AP hydrolyzes 1.0 μmol/min p-nitrophenyl phosphate at 37°C.
Binding of VEGF165-AP fusion proteins to cell-bound VEGF receptors was performed using human MDA-MB-231 breast carcinoma cells, which express NRP1 at high levels , according to the manufacturer's protocol (GenHunter Corp.). Briefly, VEGF165-AP, Sema3A, or Ig-Basic-AP fusion proteins were preincubated in the presence of 1.0 μg/ml heparin for 30 min on ice prior to cell binding. The cell-binding assay was carried out in 12-well plates for 2 h at 4°C with gentle agitation followed by five washes with ice-cold PBS. Cells were lysed with cell lysis buffer (GenHunter). The endogenous AP activities were inactivated by incubating the cell lysates at 65°C for 10 min. Cell-associated VEGF165-AP or Ig-Basic-AP was quantified by measuring the generation of yellow color at 420 nm using p-nitrophenyl phosphate as a substrate (AP Assay Reagent A; GenHunter). In the competition assay, VEGF165-AP was preincubated with various soluble forms of NRP1 or Ig-Basic-AP in the presence of 1.0 μg/ml heparin for 30 min on ice, followed by incubation with cells for 2 h at 4°C and quantitation of cell-bound AP activity as described above.
The cell migration assay was carried out in Boyden chambers as previously reported . Briefly, MDA-MB-231 cells were harvested with trypsin and resuspended in DMEM containing 0.1% BSA, 1.0 μg/ml heparin, and 20.0 μg/ml ZVAD-FMK. The inhibitory effects of a mouse monoclonal anti-NRP1 antibody (A-12, 1.0 μg/ml; Santa Cruz) or various soluble forms of NRP1 were determined by preincubating cells with the indicated reagents for 30 min on ice. Cells incubated with antibody or soluble forms of NRP1 or Ig-Basic-AP were placed in the upper chamber and the lower chamber was filled with serum-free CM from NIH 3T3 cells as a chemoattractant. The assembled Boyden chambers were incubated at 37°C in 5% CO2 for 4 h. At the end of the incubation period, cells were fixed and stained. Nonmigrating cells on the upper surface were removed. The number of migrating cells was quantified by counting 10 random high-powered fields (200× total magnification) per filter.
We thank M. Jarzynka for critically reading the manuscript, Q. Zhang and M. Mumtaz for their help during this project. We also thank J. Raper for the VEGF165-AP construct and advice on the transfection of HEK 293T cells, A. Kolodkin for the Sema3A-AP and Ig-Basic-AP constructs and a rat anti-NRP1 antibody , R. Nishikawa for frozen human glioma tissues, and J. Siegfried for frozen human lung cancerous tissues. This work was supported by The Brain Cancer Program of the James S. McDonnell Foundation, The Brain Tumor Society, The Sidney Kimmel Foundation for Cancer Research, and developmental funds from the University of Pittsburgh Cancer Institute (S.-Y.C). F.C.C. was the recipient of a Howard Hughes Medical Institute Undergraduate Summer Research Fellowship. S.-Y.C. is a Sidney Kimmel Scholar.
1Abbreviations used: NRP, neuropilins; VEGF, vascular endothelial growth factor; Sema, semaphorin; VEGFR, VEGF receptor; FGF, fibroblast growth factor; HEK, human embryonic kidney; AP, alkaline phosphatase; ORF, open reading frame; WHO, World Health Organization; GBM, glioblastoma multiforme; AA, anaplastic astrocytoma; CM, conditioned medium; RT, reverse transcriptase; PCR, polymerase chain reaction.