Gene Therapy (2001) 8, 882–890  2001 Nature Publishing Group All rights reserved 0969-7128/01 $15.00 www.nature.com/gt RESEARCH ARTICLE Adenovirus-mediated transfer of type IV ␣5 chain cDNA into swine kidney in vivo: deposition of the into the glomerular

P Heikkila¨1, A Tibell2, T Morita1, Y Chen1,GWu2, Y Sado4, Y Ninomiya5, E Pettersson3 and K Tryggvason1 1Division of Matrix Biology, Department of Medical Biochemistry and Biophysics, Departments of 2Transplantation Surgery, and 3Nephrology, Huddinge Hospital, Karolinska Institutet, Stockholm, Sweden; 4Division of Immunology, Shigei Medical Research Institute, Okayama; 5Department of Molecular Biology and Biochemistry, Okayama University Medical School, Okayama, Japan

Gene therapy of (hereditary nephritis) aims a FLAG epitope in the recombinant ␣5(IV) chain. The results at the transfer of a corrected type IV collagen ␣ chain gene indicate that correction of the molecular defect in Alport syn- into renal glomerular cells responsible for production of the drome is possible. Previously, we had developed an organ glomerular basement membrane (GBM). A GBM network perfusion method for effective in vivo gene transfer into composed of type IV collagen molecules is abnormal in glomerular cells. In vivo perfusion of pig kidneys with the Alport syndrome which leads progressively to kidney failure. recombinant adenovirus resulted in expression of the ␣5(IV) The most common X-linked form of the disease is caused chain in kidney glomeruli as shown by in situ hybridization by mutations in the gene for the ␣5(IV) chain, the ␣5 chain and its deposition into the GBM was shown by immunohisto- of type IV collagen. Full-length human ␣5(IV) cDNA was chemistry. The results strongly suggest future possibilities expressed in HT1080 cells with an adenovirus vector, and for gene therapy of Alport syndrome. Gene Therapy (2001) the recombinant ␣5(IV) chain was shown to assemble into 8, 882–890. heterotrimers consisting of ␣3(IV) and ␣4(IV) chains, utilizing

Keywords: Alport syndrome; type IV collagen; basement membrane; glomerulus; gene therapy; adenovirus

Introduction Gly-Xaa-Yaa repeat which allows for flexible kinks in the triple helix. In addition to the collagenous domain, the Alport syndrome is an inherited kidney disease charac- type IV collagen molecules have a noncollagenous globu- terized by progressive hematuria, development of renal lar NC1 domain at the carboxyl end, and the aminotermi- 1,2 failure and frequently also hearing loss. The only avail- nal has a noncollagenous 7S domain. Six genetically dis- able treatment is hemodialysis and/or kidney transplan- tinct type IV collagen ␣ chains have been described. The tation. The underlying cause of the disease is a defective ␣1(IV) and ␣2(IV) chains are ubiquitous and are present structure of the type IV collagen framework of the glom- in triple-helical molecules in a 2:1 ratio.9 The other ␣ erular basement membrane (GBM). The disorder results chains have variable and more restricted tissue distri- in deterioration of the GBM. The disease affects about bution. The current understanding of type IV collagen 1 1:5000 males. It has been estimated that 85% of cases are synthesis in the renal glomerulus is illustrated in Figure caused by mutations in the X chromosomal gene enco- 1a and b. In the GBM, ␣1(IV) and ␣2(IV) dominate during ␣ 3–5 ding the 5(IV) collagen chain. The less frequent auto- embryonic development (Figure 1a), but after birth these ␣ somal forms are caused by mutations in the 3(IV) or are replaced by trimers containing ␣3(IV), ␣4(IV) and ␣ 6,7 4(IV) collagen chain genes located on chromosome 2. ␣5(IV) chains. Due to their high cysteine content the Type IV collagen is a basement membrane-specific col- ␣3:␣4:␣5 trimers are thought to be necessary for forming lagen type which is the main structural component of a stronger, more cross-linked GBM collagen network (see 8 these extracellular structures. Similarly to other col- Figure 1b).10–12 In X-linked Alport syndrome caused by a lagens, type IV collagen is a triple-helical protein con- mutation in the ␣5(IV)␣chain gene, the ␣3(IV) and ␣ ␣ sisting of three chains. The collagen chains have (Gly- ␣4(IV)␣chains are usually absent from the GBM, even Xaa-Yaa)n repeats, glycine being the only amino acid though their genes residing on chromosome 2 are intact.13 small enough to fit into the center of the triple helix. The This is presumably due to intracellular degradation of the ␣ type IV collagen chains have many interruptions in the chains in the absence of ␣5(IV) that is essential for the formation of the ␣3:␣4:␣5 trimer. Instead, the Alport GBM contains the embryonic type of collagen IV mol- ␣ ␣ Correspondence: K Tryggvason, Department of Medical Biochemistry and ecules consisting of 1 and 2 chains (Figure 1c). How- Biophysics, Karolinska Institutet, S-171 77 Stockholm, Sweden ever, since these apparently do not provide sufficient Received 14 April 2000; accepted 14 September 2000 mechanical strength to the GBM and sufficient resistance Type IV collagen gene transfer P Heikkila¨ et al 883 a b

c d

Figure 1 Illustration of type IV collagen synthesis and incorporation into the GBM network in embryonic, adult and Alport syndrome kidney. (a) In the embryonic glomerulus, ␣1(IV) and ␣2(IV) chains assemble into triple-helical collagen molecules in a 2:1 ratio, and are secreted and deposited into the GBM network (broken lines). (b) Postnatally there is a developmental shift from ␣1(IV) and ␣2(IV) chains to ␣3(IV), ␣4(IV) and ␣5(IV) chains that form trimers in a 1:1:1 ratio.10 These molecules form a more tightly cross-linked meshwork (solid lines), due to a higher content of cysteine residues in the component chains. (c) In X-linked Alport syndrome, absence or abnormal ␣5(IV) chains lead to degradation of the ␣3(IV) and ␣4(IV) chains ␣ ␣ ␣ ␣ ␣ and consequent absence of 3: 4: 5 which are substituted by the embryonic 12: 2 trimers and, thus, a weaker GBM. (d) Gene therapy of X-linked Alport syndrome aims at restoring the situation in a normal adult (b) by introducing a vector synthesizing recombinant ␣5(IV) chains into the glomerular cells. to proteolysis,14 the consequence is deterioration of the gene transfer. The recombinant ␣5(IV) chain was shown structure, hematuria and development of Alport to be incorporated into triple-helical type IV collagen syndrome. molecules containing endogenous ␣3 and ␣4 chains, indi- Alport syndrome is an attractive candidate disease for cating that correction of the molecular defect in Alport gene therapy due to its high kidney specificity and syndrome might be possible. The second objective of this because the isolated blood circulation of the kidneys study was to explore if in vivo perfusion of pig kidneys makes them a good target for organ-specific gene trans- with these viruses would result in expression of the fer. The principle of gene therapy of X-linked Alport syn- recombinant ␣5(IV) chain and deposition of the polypep- drome is depicted in Figure 1d. This requires transfer of tide chain into the GBM. The experiments resulted in the appropriate type IV collagen ␣ chain gene to the efficient gene transfer into glomeruli and expression of endothelial and epithelial cells of the glomerulus, recombinant ␣5(IV) chain, as determined by in situ expression of the protein and intracellular assembly of hybridization and immunolocalization. Importantly, the the exogenous recombinant chain into triple-helical mol- recombinant ␣5 chain was deposited into the pig GBM, ecules together with the endogenous or ␣3, ␣4or␣5 indicating that it was assembled intracellularly into tri- chains, and, finally, secretion of the protein and ple-helical molecules together with endogenous pig type incorporation of the protein into the GBM type IV col- IV collagen ␣ chains and secreted from the cell. The lagen network (see Figure 1d).15 For the development of results strongly suggest the feasibility of gene therapy for gene therapy of Alport syndrome, we have previously Alport syndrome by targeted gene transfer to renal developed an organ perfusion system for adenovirus- glomeruli. mediated gene transfer into renal glomeruli in vivo.16 Using this procedure, we obtained a transfer efficiency of up to 85% of pig glomeruli using an adenovirus contain- Results ing the ␤-galactosidase reporter gene. Surprisingly, the kidney perfusion method resulted in efficient transfer Adenovirus vectors for expression of the type IV only to glomerular cells, while cells in other regions of collagen ␣5 chain the kidney did not take up significant amounts of the For expression of the ␣5 collagen chain two constructs virus. were made (Figure 2). The first adenovirus Ad-A5wt vec- The present study was undertaken to develop further tor contained the full-length human ␣5(IV) chain coding a gene therapy procedure for Alport syndrome. The first sequence,17 and the second one contained the same objective was to obtain expression of full-length human cDNA with an inserted FLAG-tag coding sequence. The ␣5 collagen cDNA with an adenovirus vector in cultured FLAG tag was used to enable easy purification of the cells, and explore whether the recombinant ␣5(IV) chain recombinant ␣5 chain, as well as to distinguish it from is incorporated into triple-helical molecules. Two con- the endogenous one in tissues following in vivo gene structs were made. One construct contained full-length transfer in pigs. The eight-residue FLAG tag sequence normal human ␣5(IV) chain cDNA, while the other one was placed in a 10-residue noncollagenous sequence in had an extra sequence, a FLAG-tag, that enables purifi- the fifth interruption in the collagenous domain of the ␣5 cation of the recombinant protein for functional studies, chain. This interruption coincides with interruptions in and allows for its localization in the tissue after in vivo all other ␣(IV) chains, and it is not known to have any

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 884 ␣5(IV) chain was detected as a band of about 200 kDa (Figure 3), using either an anti-FLAG antibody or mono- clonal anti-␣5(IV) antibody H53 made against a peptide sequence in the third interruption in the collagenous domain.20 The increase in size in vivo versus in vitro can be explained by glycosylation and other cellular post- translational modifications. The H53 antibody recognized polypeptide chains produced with both constructs, while the anti-FLAG antibody only recognized the ␣5(IV) chain with the FLAG marker. Both antibodies recognized two additional bands of about 100 kDa and 130 kDa in the medium. These bands probably represent degradation products.

Chain composition of secreted type IV collagen molecules containing the recombinant ␣5(IV) chain In order to study with which ␣ chains the recombinant ␣5(IV) chain can assemble in a triple-helical collagen mol- ecule, the chain was expressed in human HT1080 cells that normally express the ␣1(IV)–␣5(IV) collagen chains, but not ␣6(IV). The ␣1(IV), ␣2(IV) and ␣5(IV) chains nor- mally synthesized and secreted by these cells were easily detectable in Western blots from the medium, while the ␣3(IV) and ␣4(IV) chains were synthesized in such small Figure 2 Two adenovirus constructs expressing human ␣5(IV) collagen amounts that they were only detectable in the cell lysate. chains and schemes of the respective ␣5(IV) polypeptide chains. (a) Aden- ␣ Following infection of the cells with the adenovirus con- ovirus Ad-A5wt containing full-length human 5(IV) cDNA with the struct containing ␣5(IV) cDNA with the FLAG-tag, the rabbit globin poly(A) sequences under the control of the adenovirus major ␣ late promoter. Below, the resulting 1695 amino acid long human ␣5(IV) 5(IV) chain was detectable in cell lysates and culture chain. Noncollagenous domains are indicated by gray boxes and the large media by Western blotting using anti-FLAG antibodies. collagenous (Gly-Xaa-Yaa) repeat domain by a white box. The vertical The recombinant ␣5-FLAG chain present in the medium bars in the white box represent 22 short interruptions in the collagenous was mainly observed as a single chain without assemb- sequence. The sequence of the fifth interruption is shown. (b) Adenovirus ␣ ␣ ling with the other endogenous (IV) chains, but some construct containing full-length human 5(IV) cDNA with the 24 nucleo- of it was assembled with other ␣(IV) chains. To study tide sequence insertion coding for the FLAG sequence. The resulting 1696 ␣ amino acid long polypeptide chain contains the FLAG tag in the fifth the chain composition of trimers containing the 5-FLAG interruption of the collagenous domain. The sequence of the modified fifth chain, medium protein was immunoprecipitated with the interruption is indicated. The 8-residue sequence of the FLAG tag is boxed. anti-FLAG antibody. The immunoprecipitate was found to contain ␣3(IV) and ␣4(IV) chains, in addition to the ␣5-FLAG chain, as shown by Western blotting with chain special role, other than to provide a kink in the molecule. specific monoclonal antibodies21 (Figure 4). Furthermore, The interruption does, for example, not contain cell bind- a minor band was visible by staining with an antibody ing sites or cross-linking amino acids and, furthermore, against the ␣2(IV) chain. The ␣1(IV) chain was not con- none of the over 300 different missense mutations now tained at all in immunoprecipitates obtained with the identified in the ␣5(IV) chain in Alport syndrome is anti-FLAG antibody. The ␣6(IV) chain was never located in this interruption.5,18 Consequently, it was con- observed, which can simply be explained by the fact that sidered likely that the FLAG sequence would not inter- the ␣6(IV) chain is normally not expressed by this cell fere with assembly of this chain into a triple-helical mol- line. The immunoprecipitated medium from cells ecule. In addition to this modification, the translation initiation signal in both constructs was modified to con- tain the optimal context for initiation of translation according to Kozak.19 The two types of adenoviruses were made by homologous recombination, and the recombinant viral plaques were purified by plaque assays. The cDNA sequence encoding human ␣5(IV) cDNA was shown to be correct as determined by DNA sequencing (not shown).

Characterization of recombinant type IV collagen ␣5 chains expressed with adenovirus vectors in human cells Expression of the ␣5(IV) chain was first detected by in Figure 3 Characterization of recombinant human ␣5(IV) chains produced vitro translation of a 180 kDa polypeptide demonstrating with adenovirus in 293 human embryonic kidney cells. The cells were infected with the Ad-A5FLAG (FLAG) or Ad-A5wt (wt) adenoviruses, that the construct encodes a protein of the right size (data − ␣ or mock-infected ( ). Samples from cell lysates or media were analyzed not shown). Adenoviral expression of the 5(IV) in on a Western blot using antibodies against the ␣5(IV) chain (H53), or infected 293A cells was detected by immunoblotting of the FLAG epitope (M2). The sizes of the molecular weight standard from the protein from cell lysates and media. The recombinant the top are 212, 170, 116 and 76 kDa.

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 885 fer efficiency to pig glomeruli in vivo was as high as 80%.16 As previously reported,16 only glomerular cells showed expression of the transgene, while other kidney structures were practically completely negative for expression of ␤-galactosidase (not shown). In the experi- mental series of this study, virus concentration of 5 × 1011 p.f.u. gave the best results, with close to 50% gene trans- fer efficiency and no pathological changes observed in kidney morphology or function during the 4-day post- operation observation period (Table 1). In contrast, two gene transfer experiments with virus concentrations of 6 × 1012 and 1.6 × 1013 resulted in interstitial nephritis (Table 1). Figure 4 Chain composition of type IV collagen molecules containing the ␣ ␣ To analyze expression of the recombinant 5(IV) chain recombinant 5(IV) chain produced with adenovirus in HT1080 cells. The ␣ media from the cells infected with the Ad-A5FLAG (FLAG) or Ad-A5wt in the kidney, mRNA for the 5 chain was visualized by (wt) adenoviruses were immunoprecipitated using anti-FLAG antibodies. in situ hybridization (Figure 5) using a 172 base pair The precipitations were analyzed on Western blots using antibodies probe from cDNA encoding the NC1 domain of the against all six ␣ chains which are indicated in the upper line. The sizes human ␣5(IV) collagen chain. The expression pattern in of the molecular weight standards (kDa) are indicated. The immunopreci- the perfused kidneys was similar to that obtained by X- pitate was shown to contain ␣3(IV) and ␣4(IV) chains, in addition to the recombinant ␣5(IV) chain. gal staining, and no expression was seen in the control kidneys (Figure 5B). This means that the amount of endogenous ␣5(IV) chain mRNA is very low, or that por- expressing the wild-type construct, used as a control in cine mRNA, which has not been sequenced, is not the immunoprecipitations, was negative for all ␣ chains completely homologous with the human one. in immunostaining (Figure 4). These results demon- Translation of the recombinant ␣5 chain mRNA in the strated that the recombinant ␣5-FLAG chain is capable of perfused kidneys was demonstrated by immunofluores- incorporating into ␣3:␣4:␣5 chain trimers, the authentic ence staining using an FITC-conjugated anti-FLAG anti- type IV collagen isoform of adult GBM, while it does not assemble with ␣1(IV) and ␣2(IV) chains.

Analysis of expression of a recombinant ␣5(IV) chain mRNA and protein in the porcine kidneys following in vivo organ perfusion To study if the recombinant ␣5(IV) can be expressed in kidney glomeruli in vivo, we used a perfusion method developed for adenovirus-mediated gene transfer in pigs.16 Briefly, the porcine kidney was isolated from the systemic blood circulation by clamping the renal artery and vein, after which the organ was perfused with a sol- ution containing the adenovirus and red blood cells (hematocrit 17%) using an external pump and oxygen- ation system. Samples containing between 1.5 × 1010 and 1.6 × 1013 p.f.u. of Ad-A5FLAG adenovirus and same amounts of Ad5CMVlacZ adenovirus were injected into the renal artery before perfusion. The Ad5CMVlacZ vec- tor was used as an internal control to evaluate general gene transfer efficiency in each experiment. Gene transfer efficiency, as measured by X-gal staining of tissue sec- tions, varied between 10 and 50% of positive glomeruli, depending on the virus concentration (Table 1). The high- Figure 5 (A) In situ hybridization of the ␣5(IV) chain mRNA in pig est transfer efficiency was obtained with concentrations kidney tissue following adenovirus-mediated gene transfer in vivo reveals 13 above 1.6 × 10 . The results were similar to those pre- expression in a large proportion of the glomeruli. (B) The control kidney viously reported for the Ad5CMVlacZ virus, where trans- shows no signals.

Table 1 Semiquantitative assessment of adenovirus-mediated expression of human FLAG-tagged collagen IV ␣5 chain following perfusion with different virus concentrations

No. of animals Dose of Ad-␣5FLAG Immunostaining for Kidney per experiment virus (p.f.u.) per experiment ␣5-FLAG expression histology

3 1.5 × 1010 negative normal 3 5.0 × 1011 + normal 1 6.0 × 1012 + interstitial nephritis 1 1.6 × 1013 ++ interstitial nephritis

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 886 a b c two significant steps forward towards gene therapy of Alport syndrome. First, it was demonstrated that one can use adenovirus to produce the recombinant ␣5(IV) chain in cultured human cells so that this chain is incorporated into type IV collagen trimers with the ␣3:␣4:␣5 chain composition that is essential for normal structure and function of the adult GBM. Second, the adenovirus- mediated expression of the recombinant human ␣5(IV) chain in pig kidneys in vivo resulted in the production of an ␣5(IV) chain that also was deposited extracellularly into the GBM proper. Type IV collagen trimers composed of ␣3, ␣4 and ␣5 Figure 6 Immunolocalization of recombinant human ␣5(IV)-FLAG chain in a swine renal glomerulus following adenovirus-mediated gene transfer chains in a 1:1:1 ratio are the predominant isoform of the in vivo. (a) Immunohistochemical staining with anti-FLAG antibody GBM as shown in biochemical assays or indirectly by FITC conjugate shows the human ␣5(IV)-FLAG chain in linear struc- localization of these three chains in the GBM using chain- tures, in addition to two patchy sites. (b) Immunostaining for the endogen- specific monoclonal antibodies.12,13 In this study, the ous ␣4(IV) chain of the GBM using AlexaFluor 546-labeled secondary recombinant human ␣5(IV) chain expressed in a human ␣ antibody localizes the protein to similar linear structures as the 5(IV)- fibrosarcoma cell line was associated with ␣3(IV) and FLAG chain. (c) Merge of images in a and b demonstrates codistribution ␣ ␣ ␣ of the ␣5(IV)-FLAG and ␣4(IV) chains in the glomerulus, except for two 4(IV) chains, but not with the 1(IV) or 2(IV) chains sites where excess expression of the ␣5(IV)-FLAG chain can be seen. of type IV collagen. This was demonstrated by using a FLAG antibody that selectively immunoprecipitated the ␣3(IV), ␣4(IV) and ␣5(IV)-FLAG together. Although it was not possible to demonstrate if the three chains were present in the trimers in a 1:1:1 ratio, it can be considered body (Figure 6a), while no staining was detected in the likely, as the ␣5(IV) chain is known to assemble with control kidneys (not shown). Similarly to the in situ ␣3(IV) and ␣4(IV) in basement membranes such as the hybridization results, the recombinant ␣5(IV) polypep- GBM where these three chains are all present.12,13 The tide chain was only seen in glomeruli. Importantly, the faint staining for the ␣2(IV) chain is not considered sig- anti-FLAG staining was consistent with a linear GBM-like nificant, because the ␣2(IV) chain is expressed in these staining, indicating that the recombinant ␣5(IV) chain cells in considerably higher amounts compared with that was truly incorporated into the GBM proper (Figure 6a). of the ␣3(IV) and ␣4(IV) chains. However, it is possible Co-staining of the same sections for the the type IV col- that minor amounts of the ␣2(IV) chain can be associated lagen ␣4 chain using an AlexaFluor-labeled secondary with the network composed of ␣3(IV), ␣4(IV), and ␣5(IV) antibody (Figure 6b) indicated that the linear structures chains by noncovalent interactions as shown by Gunwar positive for the ␣5-FLAG chain were the GBM proper. et al.12 The present findings on chain composition were Merge of the two colors (Figure 6c) gave a yellowish color obtained under cell culture conditions, but the chain of the GBM structures, but at two locations the ␣5(IV) assembly of the ␣(IV)-FLAG chain in the perfused pig chain staining did not overlap with that of the ␣4(IV) kidneys of this study is likely to occur in the same way. chain. This possibly represented mesangial regions. In the present study the adenoviral gene transfer into pig kidneys in vivo resulted in expression of human ␣ Discussion 5(IV) chain in the glomeruli. This was clearly demon- strated both by mRNA and protein analyses using in situ The present work was carried out to explore the possi- hybridization and indirect immunofluoresence, respect- bility of developing treatment for Alport syndrome by ively. As previously described the perfusion method only gene therapy. The basic disorder of this disease is abnor- directed the gene transfer to glomerular cells, which are mal structure of the GBM network made of triple-helical the actual target cells for gene therapy of Alport syn- type IV collagen molecules that have an ␣3:␣4:␣5 chain drome. The recombinant ␣5(IV)-FLAG chains were composition. Mutations in any of the respective genes can clearly incorporated into the GBM proper, as demon- lead to the disease. strated by the double staining for the recombinant Gene therapy of Alport syndrome aims at the transfer ␣5(IV)-FLAG and endogenous ␣4(IV) chains. It can be of a corrected type IV collagen ␣ chain gene into renal considered likely that the ␣5(IV)-FLAG chains assembled glomerular cells that are responsible for production of the intracellularly into trimers with endogenous porcine GBM. The prerequisites for gene therapy of Alport syn- ␣3(IV) and ␣4(IV) chains, because this is the combination drome include: (1) availability of an appropriate gene normally found in cells that simultaneously synthesize delivery system into cells of renal glomeruli; (2) these chains. Furthermore, it is known that single ␣5(IV) expression of the delivered type IV collagen gene in those chains do not form trimers, and that they are degraded cells; (3) proper post-translational modifications and fold- inside the glomerular cells if they do not find their ␣3(IV) ing of the respective ␣ chain which facilitates intracellular and ␣4(IV) chain partners, as happens in autosomal association into an ␣3:␣4:␣5 trimer; and (4) incorporation Alport syndrome where either the ␣3(IV) or ␣4(IV) chains of those trimers extracellulary into the GBM proper, are absent. The sequence encoding porcine ␣5(IV) chain which could restore the deteriorated GBM structure. is not known and, therefore, it is not known how homolo- We have previously demonstrated the possibility of gous it its with the human sequence. However, in this targeting expression of foreign genes into cells of the study the deposition of the recombinant human ␣5(IV) renal glomeruli in vivo, using adenovirus containing a chain into the GBM in swine kidneys implies that the reporter gene.16 The results of the present study represent swine ␣5(IV) can be replaced by the corresponding ␣

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 887 chain of man. This is not unusual for heterotrimeric base- plasmid containing the SV40 polyA signal was created. ment membrane , as for example, recombinant The SV40 polyA signal was amplified by PCR from a mouse and human chains expressed simul- pSG5 vector (Stratagene) using a linker sequence in a for- taneously in vitro and in vivo can form hybrid trimers.22,23 ward primer to create EcoRI, MunI and BglII sites to the A large body of data has shown that only trimeric col- 5Ј end, and an XhoI site to the 3Ј end of the PCR frag- lagen molecules are incorporated into fibers or networks ment. The fragment was subsequently inserted into the of the . Therefore, it is likely that the pBluescript vector using the EcoRI and XhoI sites. deposition of the recombinant ␣5(IV)-FLAG chain in the A subclone A5–3ЈXX containing the 3Ј end of the GBM represents depostion of trimeric molecules. The ␣5(IV) cDNA linked to SV40 poly A sequence was cre- most likely composition of these trimers is ␣3:␣4:␣5 ated by inserting an EcoRI(2934)–ApaI(3969)fragment because this isoform normally forms the type IV collagen from the PL-31 cDNA clone and an ApaI(3969)– network of adult GBM. However, direct evidence for EcoRI(5270) fragment from the MD-6 cDNA clone14 into such trimers would require isolation of such molecules a pBluescript vector cut with EcoRI. The resulting and their biochemical analysis which could not be carried EcoRI(2934–5270) fragment was ligated into the SV40 out within this study. polyA signal containing plasmid that was digested with The current results indicate that a genetic collagen dis- MunI and EcoRI. The subclones in which the EcoRI site ease such as Alport syndrome might be treatable by tar- (5270) was mutated, due to ligation to the MunI site, were geted gene transfer to the renal glomeruli, and most of selected and subsequently cut with XbaI and EcoRI. The the criteria for carrying out actual gene therapy experi- XbaI(2404)–EcoRI(2934) fragment from the HT14 cDNA ments have now been met. Several dog24,25 and mouse26–28 clone17 was then inserted to create the plasmid A5–3ЈXX, models for both X chromosome linked and autosomal that contained the 3Ј half of the cDNA. The numbering of recessive Alport syndromes are now available for gene the nucleotides starts from base 1 at the 5Ј of the cDNA.17 therapy studies. The X-linked forms described in A plasmid A5–5ЈNX containing the 5Ј half of the cDNA Samoyed dogs24 and in a family of mongrel dogs from was created by ligating the following three DNA frag- Navasota, Texas25 render themselves useful for the type ments into the pBluescript vector that was linearized of gene transfer experiments carried out in this study, as with NotI and XbaI: (1) NotI–AvaI(420) from the JZ-4 the canine disease is essentially identical to that in cDNA clone17 subcloned to the pBluescript vector; (2) humans. Affected dogs of both strains develop hematuria AvaI(420)–AccI(768) PCR fragment amplified from the and proteinuria within a few months after birth, and pro- HT14 cDNA template using a forward primer extending gression to renal insufficiency by 8–15 months. Adeno- to the AvaI(420) site (HT14 clone starts at site 444); (3) virus-mediated gene transfer with ␣5(IV) cDNA initiated AccI(768)–XbaI(2404) from the HT14 cDNA clone. at the onset of hematuria symptoms might give a positive To generate plasmid construct pA5-UFL that contains therapeutic response that could extend the life span of the full-length coding sequence for the ␣5(IV) chain, the affected dogs. Such an experiment would be important XbaI(2404)–XhoI fragment was recovered from the A5– as a ‘proof of principle approach’. Since adenovirus only 3ЈXX plasmid and ligated into the A5–5ЈNX plasmid that provides expression of the transgene for about 6–8 weeks, was cut with XbaI and XhoI. it will not be a question of life-lasting treatment with such To facilitate recombinant protein purification and dis- a vector. However, since the half-life of basement mem- tinction from the endogenous ␣5(IV) collagen the full- brane proteins, such as type IV collagen, is up to 2 years, length ␣5(IV) cDNA was modified to contain a nucleotide the life span of the affected dogs might be expanded sig- sequence encoding the FLAG epitope, an octapeptide nificantly if the treatment turns out to be effective. with the amino acid sequence DYKDDDDK. The FLAG In conclusion, the present results provided important sequence was added to the sequence encoding the fifth advances concerning two key prerequisites for gene ther- interruption in the collagenous domain by oligonucleo- apy of Alport syndrome. First, a recombinant ␣5(IV) tide-directed mutagenesis. A forward primer homolo- chain expressed by an adenovirus is capable of assemb- gous to fragment 1033–1052 in the cDNA sequence and ling with endogenous ␣3(IV) and ␣4(IV) chains of cul- a mutagenic primer 5Ј-TTCTCCTATAGTTATCTTGTCA- tured cells into triple helical collagen molecules. Second, TCGTCGTCCTTGTAGTCTCTAGGAATTACAAGTCCA- the recombinant ␣5(IV) chain expressed by adenovirus in 3Ј containing the FLAG encoding sequence were used to swine renal glomerular cells in vivo is deposited into the amplify a 254 bp megaprimer which after purification GBM. These results together with the previous develop- was used as a primer in a second PCR reaction with a ment of a glomerulus-specific gene transfer method reverse primer homologous to nucleotides 1702–1721. strongly support the possibilities for gene therapy of The HT14 cDNA clone was used as a template. The 700 Alport syndrome. bp PCR product was digested with MscI(1120) and HincII(1683) and inserted using the same enzymes into A5–5ЈNX which was earlier modified by deleting the Hin- Materials and methods cII site from the polylinker. The sequence of the PCR fragment and the cloning sites of the resulting plasmid DNA construction and analysis were verified by sequencing, and the NotI–XbaI(2404) A 5270 base pair plasmid containing full-length human insert was ligated into the A5–3ЈXX plasmid to generate ␣5(IV) collagen cDNA with an SV40 polyA signal was the pA5-UFLAG plasmid. The inserts of the pA5-Uwt constructed from overlapping cDNA clones and PCR and pA5-UFLAG plasmids were sequenced to check the products into a pBluescriptSK− vector (Stratagene Clon- presence of possible mutations, and then used in an in ing Systems, La Jolla, CA, USA). The SV40 poly A vitro translation assay (TNT coupled reticulocyte lysate sequence was included in the early cloning steps, but not system; Promega, Madison, WI, USA) to ensure the trans- in the final expression constructs. First, a pBluescript lation of a full-length polypeptide chains (185 kDa).

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 888 To increase the expression level of the recombinant Protein analysis, purification, antibodies and protein, the translation initiation signal of ␣5(IV) cDNA immunoblotting was replaced by a modified Kozak translation initiation Three days after infection, the cells were collected and sequence.19 A forward primer 5Ј-AAGGAAAAAAGCG- suspended in an SDS sample buffer. The medium was GCCGCAAGCTTGCCGCCACCATGGAACTGCGTGG- collected and centrifuged at 8000 r.p.m. for 10 min, and AGTCAGCCT-3Ј homologous to ␣5(IV) cDNA at pos- used in immunoprecipitation or concentrated by precipi- ition 203–225, plus containing a modified translation tation with 70% ethanol at −20°C for 1 h. After centrifug- initiation signal and restriction sites for NotI and HindIII, ation at 8000 r.p.m. for 1 h at 4°C, the pellets containing and a reverse primer homologous to 417–440, was used the precipitated proteins were suspended in a SDS to amplify a 270 bp PCR fragment using the HT14 cDNA sample buffer and analyzed by SDS-PAGE. clone as a template. The PCR product was digested with To purify intact, recombinant ␣5(IV) chains containing NotI–AvaI(420) and subcloned, together with the a FLAG tag, medium from adenovirus-infected cells was AvaI(420)–XbaI(2404) insert from the pA5-Uwt and pA5- immunoprecipitated using an anti-FLAG M2 antibody UFLAG plasmids into the A5–3ЈXX plasmid that was lin- covalently conjugated to Sephadex matrix (Eastman earized with NotI–XbaI to generate plasmids pA5-Uwt- Kodak Scientific Imaging Systems, New Haven, CT, 205 and pA5-UFLAG-205, respectively. USA), as described in the manufacturer’s instructions. To produce the transfer plasmids for generation of the The protein was eluted into reductive SDS sample buffer adenoviruses to be used for gene transfer of the ␣5(IV) and analyzed by SDS-PAGE using 6% gels. Western blot cDNAs, the NotI–BglII inserts of pA5-UFL-205 and pA5- analysis was performed using standard methods. The UFLAG-205 were ligated into a dephosphorylated primary antibodies used were rat monoclonal antibodies pAdBM5pAG vector (Quantum Biotechnologies, Quebec, H11, H22, H31, H44, H53 and H63 against the type IV Canada) digested with BamHI. The unligated ends were collagen ␣1to␣6 chains, respectively,20,21 or antibodies treated by Klenow, phenol extracted and subsequently raised against the FLAG epitope (Eastman Kodak Scien- blunt-end ligated to form circular transfer plasmids tific Imaging Systems). The blots were incubated with the called pAdBM5pAG+Uwt and pAdBM5pAG+UFLAG. primary antibodies, followed by incubation with the sec- ondary antibody and chemiluminescent detection (Du Construction and purification of recombinant Pont NEN, Boston, MA, USA). The secondary anti- adenoviruses bodies were peroxidase-conjugated (DAKO, Glostrup, Recombinant adenoviruses (Ad-A5FLAG and Ad-A5wt) Denmark). containing the 5.2 kb cDNA encoding the human ␣5(IV) type IV collagen chain with and without a FLAG tag In vivo administration of adenovirus into the pig kidney were produced using an Adeno-Quest Adenovirus by organ perfusion expression system (Quantum Biotechnologies). Briefly, In vivo perfusion was carried out essentially as previously E1/E3-deleted replication-defective serotype 5 human reported.16 Briefly, young farm pigs were anesthesized, adenoviral (AdCMVlacZ⌬E1/⌬E3) DNA and the recom- and the left kidney was exposed via paramedial incision binant transfer vectors pAdBM5pAG+Uwt and and using extraperitoneal dissection. The renal artery, pAdBM5pAG+UFLAG linearized with ClaI were co- vein and ureter were clamped and cannulated. The can- transfected into 293A cells. The recombinant virus pla- nules were connected to silicon tubings connected to the ques purified by consecutive plaque assays and viral perfusion device. Perfusion was performed in a closed- clones expressing the ␣5(IV) chain were identified first circuit mode where the oxygenated and heated (37°C) by PCR and then by Western blotting using anti-FLAG- perfusate was recirculated continuously through the kid- M2 and anti-␣5(IV) chain H53 antibodies. The resulting ney. The perfusate solution (about 250 ml) contained pre- adenoviruses contained the ␣5(IV) chain cDNA under the viously separated and blood group compatible porcine control of the adenovirus major late promoter and red blood cells (17% hematocrit value), 10000 U heparin enhancer and the rabbit globin poly A sequences. The (Lo¨vens, Malmo¨, Sweden), 100 mg cefuroxim (Zinacef; adenoviral stocks were purified twice by ultracentrifug- Glaxo Welcome, Mo¨lndahl, Sweden) in Krebs-Ringer sol- ation through a CsCl2 gradient, followed by desalting ution. Before connection of the renal arterial inlet to the with Econopac columns (Bio-Rad Laboratories, Hercules, perfusion device, 18 ml of 0.9% NaCl solution, containing CA, USA). The titers of the virus stocks were assessed by 180 mg lidocain (Xylocain; Astra-Zeneca, So¨dertalje, plaque assays. Human adenovirus AdCMVlacZ, used as Sweden) and 5000 U heparin, was infused into the renal a control, was amplified and purified as above. The viral artery followed by infusion of the adenovirus dissolved preparations were tested for replication competence by in 9–14 ml final volume of phosphate-buffered saline pH extended cultivation on HeLa cells. 7.4. The kidney was exposed to 15 min warm ischemia before perfusion with the oxygenated perfusate. The Infection of human cells in vitro urine was collected continuously and added back to the HT1080 or 293A cells were cultured in DMEM (Life Tech- perfusion fluid. Perfusions were carried out for 120 min nologies, Gaithersburg, MD, USA) containing 10% FCS with a perfusion flow of 53 ml/min and an arterial press- and penicillin/streptomycin until 80% confluency. The ure varying between 150 and 200 mmHg. Thereafter, the cells were infected with adenoviruses Ad-A5FLAG or cannules were removed and the incisions in the vessels Ad-A5WT at MOI 1000 in serum-free medium for 3 h. and ureter were sutured. Heparin 1000 U was given i.v. The virus containing medium was rinsed off and the cells immediately before reperfusion and 15 min after reper- were fed by serum-free medium containing 100 ␮g/ml fusion. Fragmin (2500 U) (low molecular weight heparin, ascorbate (Sigma, St Louis, MO, USA). Ascorbate (50 Pharmacia Upjohn, Stockholm, Sweden) was adminis- ␮g/ml) was added daily to ensure appropriate post- tered s.c. at the end of the operation and the following translational modifications of collagenous proteins. morning. Solu-Cortef (hydrocortisone, 100 mg; Pharma-

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 889 cia Upjohn, Stockholm, Sweden) was given i.v. at reper- Principles of Molecular Medicine. Humana Press: Totowa, NJ, fusion. To avoid circulatory instability, calciumchlor- 1998, pp 665–668. idehexahydrate 100 mg/ml (Apoteksbolaget, Umeå, 3 Barker D et al. Identification of mutations in the COL4A5 col- lagen gene in Alport syndrome. Science 1990; 248: 1224–1227. Sweden) and a dose of 0.3 ml/kg i.v. was administrated ␣ 5–10 min before reperfusion. Zinacef was given at a dose 4 Hostikka SL et al. Identification of a distinct type IV collagen chain with restricted kidney distribution and assignment of its of 100 mg i.v. at the start of the operation and after 8 and gene to the locus of X chromosome-linked Alport syndrome. 16 h as a prophylactic antibiotic treatment. The animals Proc Natl Acad Sci USA 1990; 87: 1606–1610. were maintained on i.v. fluids overnight after which they 5 Tryggvason K. Mutations in type IV collagen genes and Alport gradually resumed oral feeding. Four days later, the ani- phenotypes. In: Tryggvason K (ed.). Molecular Pathology and Gen- mals were killed and the kidneys were harvested for later etics of Alport Syndrome. Karger: Basel, 1996, pp 154–171. analysis by histochemistry and immunohistochemistry. 6 Mochizuki T et al. Identification of mutations in the ␣(IV) and The study was approved by the local Animal Ethics Com- ␣4(IV) collagen genes in autosomal recessive Alport syndrome. mittee and was performed in accordance with the Swed- Nat Genet 1994; 8: 77–81. ish law concerning animal welfare and treatment of 7 Lemmink KK et al. Mutations in the type IV collage ␣3 research animals. (COL4A3) gene in autosomal recessive Alport syndrome. Hum Mol Genet 1994; 3: 1269–1273. In situ hybridization 8 Hudson BG, Reeders ST, Tryggvason K. Type IV collagen: Struc- ture, gene organization, and role in human diseases. Molecular For in situ hybridization, tissues from pig kidneys were basis of goodpasture and Alport syndromes and diffuse leiomy- fixed in paraformaldehyde, dehydrated, embedded in omatosis. J Biol Chem 1993; 268: 26033–26036. paraffin and sectioned. The sections were post-fixed, 9Ku¨ hn K. Basement membrane (type IV) collagen. Matrix Biol incubated in PBS containing 0.1% active diethyl pyrocar- 1994; 14: 439–445. bonate (Sigma), equilibrated in 5 × SSC and prehy- 10 Miner JH, Sanes JR. Collagen IV ␣3, ␣4, and ␣5 chains in rodent bridized for 2 h at 55°C. The sections were hybridized basal laminae: sequence, distribution, association with , overnight at 55°C with a 172 bp fragment encoding the and developmental switches. J Cell Biol 1994; 127: 879–891. NC1 domain of the ␣5(IV) chain. Digoxygenin (DIG)-11- 11 Leinonen A et al. Complete primary structure of the human type UTP-labeled antisense and sense riboprobes were gener- IV collagen ␣4(IV) chain; comparison with structure and ␣ ated by in vitro transcription with T7 and T3 RNA poly- expression of the other (IV) chains. J Biol Chem 1994; 269: merase (Boehringer Mannheim Biochemicals, Mannheim, 26172–26177. Germany). After washing in 50% formamide and stan- 12 Gunwar S et al. Glomerular basement membrane; identification of a novel disulfide cross-linked network of ␣3, ␣4 and ␣5 chain dard sodium citrate, the sections were incubated with an of type IV collagen and its implication for the pathogenesis of alkaline phosphatase-coupled anti-DIG antibody and Alport syndrome. J Biol Chem 1998; 273: 8767–8775. developed using nitro blue tetrazolium and 5-bromo-4- 13 Nakanishi K et al. Immunohistochemical study of ␣1–␣5 chains chloro-3-indolyl phosphate (NBT/BCIP) solutions of type IV collagen in hereditary nephritis. Kidney Int 1994; 46: (Boehringer Mannheim).Tissues from the right untreated 1413–1412. kidney were used for control analysis. 14 Kalluri R et al. Isoform switching of type IV collagen is develop- mentally arrested in X-linked Alport syndrome leading to Immunohistochemical and histochemical analysis of increased susceptibility of renal basement membranes to endo- kidney tissues proteolysis. J Clin Invest 1997; 99: 2470–2478. Kidney tissues were studied for ␤-galactosidase expression 15 Tryggvason K et al. Can Alport syndrome be treated by gene by X-gal staining.16 In immunohistological studies, anti- therapy. Kidney Int 1997; 51: 1493–1499. FLAG-M2 FITC conjugate (Sigma) and anti-␣4(IV) chain 16 Heikkila¨ P et al. Adenovirus-mediated gene transfer into kidney antibody H4320,21 were used. Cryosections were fixed glomeruli using an ex vivo and in vivo kidney perfusion system – first steps towards gene therapy of Alport syndrome. Gene Ther- using acetone/methanol (1:1), and blocked using 20% nor- apy 1996; 3: 21–27. mal goat serum. The M2 antibody FITC conjugate was ␮ 17 Zhou J, Hertz JM, Leinonen A, Tryggvason K. Complete amino diluted to 10 g/ml and the H43 antibody to 1:50, and the acid sequence of the human ␣5(IV) collagen chain and identifi- staining was carried out using AlexaFluor 546 labeled goat cation of a single base mutation in exon 23 converting glysine secondary antibody (Molecular Probes, Eugene, OR, USA). 521 in the collagenous domain to cystein in an Alport syndrome Tissues from the right untreated kidney were used as a patient. J Biol Chem 1992; 267: 12475–12481. control. 18 Martin P et al. High mutation detection rate in the COL4A5 col- lagen gene in suspected Alport syndrome using PCR and direct DNA sequencing. J Am Soc Nephrol 1998; 9: 2291–2301. Acknowledgements 19 Kozak M. Determinants of translational fidelity and efficiency We are grateful to Tiina Berg, Margareta Andersson and in vertebrate mRNAs. Biochimie 1994; 76: 815–821. Dr Ehab Rafael for assistance, and thank Ingvild Halbig 20 Kagawa M et al. Epitope-defined monoclonal antibodies against type IV collagen for diagnosis of Alport’s syndrome. Nephrol for care of the animals. This work was supported in part Dial Transplant 1997; 12: 1238–1241. by grants from the Swedish Medical Research Council, 21 Sado Y et al. Establishment by the rat lymph node method of the Novo Nordisk Foundation, Hedlund’s Foundation, epitope-defined monoclonal antibodies recognizing the six dif- and by a EU grant No. BIO4-CT96–0537. ferent ␣ chains of human type IV collagen. Histochem Cell Biol 1995; 104: 267–275. ␣ References 22 Yurchenco P et al. The chain of laminin-1 is independently secreted and drives secretion of its ␤- and ␥-chain partners. Proc 1 Atkin CL, Gregory MC, Border WA. Alport syndrome. In: Natl Acad Sci USA 1997; 94: 10189–10194. Schrier WW, Gottschalk CW (eds). Diseases of the Kidney. Little 23 Kortesmaa J et al. Recombinant laminin-8 (␣4␤1␥1) – production, Brown: Boston 1988, pp 617–641. purification and interactions with integrins. J Biol Chem 2000; 2 Tryggvason K, Heikkila¨ P. Alport syndrome. In: Jamison L (ed.). 275: 14853–14859.

Gene Therapy Type IV collagen gene transfer P Heikkila¨ et al 890 24 Zheng K et al. Canine X chromosome-linked hereditary neph- 26 Miner JH, Sanes JR. Molecular and functional defects in kidneys ritis: a genetic model for human X-linked hereditary nephritis of mice lacking collagen ␣3(IV): implications for Alport syn- resulting from a single mutation in the gene encoding the ␣5(IV) drome. J Cell Biol 1996; 135: 1403–1413. chain of type IV collagen. Proc Natl Acad Sci USA 1994; 91: 27 Cosgrove D et al. Collagen COL4A3 knockout: a mouse model 3989–3993. for autosomal Alport syndrome. Genes Dev 1996; 10: 2981–2992. 25 Lees G et al. Glomerular ultrastructural findings similar to her- 28 Lu W et al. A new model of Alport’s syndrome. J Am Soc Nephrol editary nephritis in 4 English cocker spaniels. J Vet Intern Med 1997; 8: 1818A. 1997; 11: 80–85.

Gene Therapy