6.4. Transgenic mice expressing truncated type XV collagen (IV)

To gain insight into the biological function of type XV collagen, transgenic mice expressing mutant type XV collagen were generated by the microinjection technique (Hogan et al., 1994). This approach was designed to complement the study of consequences of the lack of type XV collagen in mice (Eklund et al., 2001). It is known for many collagens that mutations in their genes that cause the synthesis of abnormal collagen chains, which are still able to associate with other chains, cause more severe phenotypes than null alleles or mutations that prevent chain association, due to the dominant negative effect or “procollagen suicide” (see 2.3. and 2.4.2, reviewed by Aszodi et al., 1998; Myllyharju & Kivirikko, 2001). According to that, truncated type XV collagen molecules were anticipated to form heterotypic molecules with the endogenous type XV collagen chains, and thus result in the reduction of the endogenous type XV collagen level via procollagen suicide and/or interference in the normal function of this collagen via mutant α1(XV) chain -induced conformational change in its structure. In addition, the consequences of the internally truncated α1(XV) chain as the sole source of type XV collagen were studied by crossing the Col15a1^-/- mice with the mutant transgenic lines.

Microinjections of two different minigene constructs, both having an in-frame 123-amino acid deletion in the collagenous region driven by 3 or 8-kb endogenous 5´-flanking sequences, resulted in the establishment of six transgenic lines, 3XV-1, -8, and –29 for the shorter, and 8XV-01, -02, and -166 for the longer promoter fragment, respectively, with relatively low transgene copy numbers ranging from 2 to >5. In lines with the shorter promoter fragment, the offspring from heterozygote matings displayed essentially a Mendelian distribution of genotypes, thus the mutation did not interfere with normal development. Furthermore, the transgenic mice were viable, fertile, and phenotypically indistinguishable from their nontransgenic littermates. Neither did the histological analysis at the light microscopic level reveal any abnormalities. RT-PCR analysis revealed expression of the transgene in all lines, but the pattern and level were different compared with those of the endogene. However, the 3 kb-promoter fragment used to drive the expression of the minigene resulted in a marked decrease in the basic transcription activity compared to shorter or longer fragments in a luciferase-reporter gene assay, suggesting the presence of a silencer element between positions –2926 bp and –2064 (see above and III). In the same assay, the 8-kb promoter fragment conferred 3 times higher basic activity compared with the 3-kb fragment. The deviation from the endogenous gene expression pattern in the tissues of 3XV-mice of the transgenic lines 3XV-1, -8, and –29 is thus likely the result of the lack of regulatory elements in the shorter promoter fragment. Use of the longer promoter fragment resulted in a more endogenous-like transgene expression pattern in two of the three transgenic lines.

The Western blot analysis suggested that the expression of the transgene protein was copy-number dependent and correlated with the mRNA levels observed in the RT-PCR. Thus, with equal loading of dilute tissue homogenates in line 3XV-29, a discrete 180-kDa-protein band was detected with both antibodies in transgene positive mice, but not in wild-type or Col15a1^-/- control tissues. The same band appeared also in the concentrated tissue homogenates of wild-type and 8XV-166 mice, indicating that it represents wild-type and transgene products, and the latter is associated with or has the same mobility as the former, although it has a 10 kDa lower molecular weight. The observed protein size is reasonably well in line with those previously reported (Myers et al., 1996; Hägg et al., 1997b; Li et al., 2000).

Interestingly, when the mice expressing the truncated type XV collagen were crossed with mice lacking this collagen, there was a clear deviation from the expected Mendelian ratios, especially in the 3XV-29-derived COLXV29 line with the higher copy number of the transgene. The transgene was preferably detected in combination with the endogenous alleles, and there were less than expected offspring with the transgene and heterozygous for the endogenous allele, or lacking the endogenous allele. Thus, at least to some extent, the truncated transgene product is not tolerated when the level of the endogene gene product decreases, suggesting a ratio dependence of transgene vs. endogene. Unfortunately, the mice were lost before more offspring could be generated to confirm the numbers, and before the initiation of a histological analysis. Also, pregnancy terminations would have been needed to see whether spontaneous abortions of a portion of the transgene positive offspring took place.

In order to study whether the truncated α1(XV)-chains could associate to form mutant type XV collagen molecules, tissues from COLXV1, a line containing only transgenic type XV collagen, were stained with a type XV collagen antibody. Although the transcripts for the transgene were expressed in a number of tissues, the protein product was detected only in the capillaries of the brain matrix. Previous studies have shown that type XV collagen is a prominent component of most, but not all, capillaries (Hägg et al., 1997b), and that there are developmental shifts in its expression. The brain is one of the organs affected by these shifts, as the capillaries in the forebrain and most superficial layers contain this collagen only during a very short period of development (from 12 to 14 dpc), while the expression in the mature brain is restricted to a subset of blood vessels, preferably those of larger caliber (III). Also, line 3XV-29 showed abnormal expression in the brain capillaries. The fact that two independent transgenic lines sharing the same 3-kb promoter fragment show “misexpression” of the transgene in brain capillaries, and that the line with the 8-kb promoter does not, indicates that this expression pattern is due to the regulatory properties of the shorter promoter. Interestingly, the type XV collagen 5´-flanking sequence contains several putative binding sites for the POU domain transcription factors expressed in mammalian forebrain (reviewed Rosenfeld, 1991). Based on the early and distinct spatio-temporal expression pattern of POU-domain genes in the forebrain, they are believed to have a role e.g. in the specification of neuronal phenotypes. The 3-kb promoter contains three out of four (within –5080) POU-domain motifs Brn-2 at –901 to –886 bp, Tst-1/Oct-6 at -1385 to –1371 bp, and Brn-2 at –1421 to –1406 bp. (In addition, one Brn-2 motif is located just outside the 3-kb promoter at –3248 to –3233 bp, whereas one Tst-1/Oct-6 motif is in the 5´-end of the large second intron.) Thus, truncation of the promoter at 3-kb could remove factors important in silencing type XV collagen expression in brain capillaries, which in conjunction with the presence of strong potent activators may be the reason for the observed “misexpression”.

Type XV collagen is known to provide structural integrity to microvessels, since the lack of it leads to collapse in the capillary walls in heart and skeletal muscle (Eklund et al., 2001). Function beyond a mere structural role is supported by the facts that the C-terminal fragment of type XV collagen restin can inhibit endothelial cell migration in vitro (Ramchandran et al., 1999) and that the surrounding matrix is implicated in endothelial cell migration and proliferation, both processes occurring in angiogenesis. In the transgenic mice, the immunostaining patterns for type IV collagen and the laminin γ 1-chain were unchanged, and thus the misexpression did not dramatically change the BM composition (not shown). In addition, the number of capillaries was apparently equal in both control and transgenic mice (not shown). Thus, the misexpression of type XV collagen in brain capillaries of the lines with the shorter promoter fragment did not appear to have major functional consequences.

The lines generated using an 8-kb promoter were expected to confer better tissue specificity. This was partly observed in two of the lines, although some elements needed for correct tissue-specificity were still missing (see above). As with the previous lines, the transgene positive mice appear phenotypically normal, viable, and fertile. However, a slight decline in the litter size (average 5 pups per litter compared with 7-9 for 3XV-lines), and a less than expected number of transgene positive mice from heterozygote matings (observed 35/62, expected 46,5/62), suggest that a portion of transgene positive mice were lost prenatally. Pregnancy terminations indicated a high number of abortions at 12 and 14 dpc (11/25), as well as distorted genotype distribution in the remaining offspring (observed 8/25, expected 18,75/25), indicating that most of the aborted fetuses were most likely transgene positive, probably homozygous for the transgene. The abortions begin between 10 to 11 dpc, as termination at 10 dpc revealed only 1 aborted fetus out of 9, with the genotype distribution in line with the Mendelian ratios. Further work includes studying the genotype ratio in other lines and obtaining histological data from aborted transgene positive fetuses to reveal the cause of death. It must be determined whether the death is primarily caused by mutated collagen XV, or its secondary consequence, for example, due to its false expression in tissues. For that, the knowledge of the expression pattern in developing and mature mice is necessary.

The lack of phenotypic consequences in lines with the 3 kb-promoter is likely due to aberrant gene regulation, which did not confer endogeneous like expression. In addition, there are a number of other reasons why the truncated type XV collagen would not cause any obvious phenotypic changes in transgenic mice, despite being expressed at both at mRNA and protein levels. The deleted portion may not be functionally important. The fact that the two deleted COL-domains are conserved between man and mouse argues against that hypothesis (Hägg et al., 1997a). A continuous, non-interrupted triple-helical structure appears to be most critical in the fibril-forming collagens (reviewed by Myllyharju & Kivirikko, 2001), while the nonfibrillar collagens that are characterized by a number of interruptions in their triple-helical structures may tolerate e.g. glycine mutations, as seen in the COL4A4 gene (Boye et al., 1998), or even 9-bp deletions in the collagenous domain, as in the COL9A3 gene, with no evident consequences (Paassilta et al., 1999). Nevertheless, the 123-residue truncation should result in the looping-out of the corresponding portion of the normal-length α1(XV)-chains, and thus interfere with the formation of higher order structures, which contain type XV collagen, and/or result in the degradation of the entire molecule.

Secondly, the expression level of the transgene may be too low. There are examples from transgenic mice harboring mutations in the fibrillar collagen genes of a copy number dependency in the severity of the phenotypes. A glycine mutation in the COL2A1 gene caused severe chondrodysplasia (Garofalo et al., 1991) in three lines having 10-50 copies of the transgene, whereas one line with 2 copies appeared normal. A 15-amino acid deletion in the Col2a1 gene causing chondrodysplasia displayed also copy-number dependency, as the line with only one copy did not show any abnormalities, whereas the offspring in two lines with 12 (homozygous) and 15 copies (heterozygous) had a severe phenotype (Metsäranta et al., 1992). In this case, it was concluded that the expression of transgene mRNA at levels equivalent or higher than the endogenous mRNA is necessary to cause the phenotype.

Thirdly, the mutation in transgenic lines described here may also cause a subtle phenotype too discrete to be detected at the light microscopic level without any challenging. This is supported by the fact that lack of this collagen in Col15a1^-/- mice causes a relatively mild phenotype affecting skeletal muscle and the cardiovascular system (Eklund et al., 2001). The muscle phenotype was first detected at the light microscopic level, as focal histopathological signs for regeneration followed by degeneration. A large number of samples was needed to confirm the phenotype, since there was variation within a sample in the frequency of the damaged areas, and between the muscle groups in their susceptibility to damage. However, the muscle damage was greatly induced when the mice were subjected to running in the motor-driven treadmill. The detection of cardiovascular phenotypes required the use of sophisticated equipment and experimental set-ups. Electron microscopy was also needed to pinpoint collapsed capillaries and endothelial cell degeneration in the heart and skeletal muscle of null mice, findings, which were likewise undetectable under the conventional microscope. To study the cardiovascular phenotype further, the mice were subjected to exercise-induced cardiovascular stress and studied for signs of cardiac injury. Furthermore, isolated perfused hearts were used to monitor cardiac function. Taken together, the sedentary lifestyles the mice live in their cages is far from the natural situation, thus further stimuli were needed to detect and confirm the mild phenotype in Col15a1^-/- mice. The effect of generated mutations may be also transient, as described for laminin α4 null mice, which exhibit anemia and hemorrhages when newborn. Within two weeks, however, both symptoms completely disappear, thus restoring the normal phenotype (Kortesmaa, 2000).

A general tendency for collagens is that the dominant-negative mutations, which produce structurally altered α-chains, lead to more severe phenotypes than null mutations (reviewed by Aszodi et al., 1998). This has been the case for type X collagen, where mice expressing mutant collagen develop a more severe skeletal phenotype (Jacenko et al., 1993) than the mice lacking type X collagen (Kwan et al., 1997; Gress and Jacenko, 2000), which were actually first reported as nonphenotypic (Rosati et al., 1994). Here, prenatal lethality was observed in one of the mouse lines expressing the transgene in tissues, known to express the gene normally, whereas the lack of type XV collagen in Col15a1 null mice is known to cause a relatively mild phenotype affecting skeletal muscle and the cardiovascular system (Eklund et al., 2001). Thus, expression of mutant α1(XV) chains appears to result in more severe phenotypic consequences than a null allele, and conditions previously not known to be caused by mutations in this collagen may be identified in the future.