Generation and Initial Characterization of FDD Knock In Mice

Background Mutations in the integral membrane protein 2B [1], also known as BRI2 [2], a type II trans-membrane domain protein cause two autosomal dominant neurodegenerative diseases, Familial British and Danish Dementia [3]. In these conditions, accumulation of a C-terminal peptide (ABri and ADan) cleaved off from the mutated precursor protein by the pro-protein convertase furin [4], leads to amyloid deposition in the walls of blood vessels and parenchyma of the brain. Recent advances in the understanding of the generation of amyloid in Alzheimer's disease has lead to the finding that BRI2 interacts with the Amyloid Precursor Protein (APP), decreasing the efficiency of APP processing to generate Aβ [5], [6], [7]. The interaction between the two precursors, APP and BRI2, and possibly between Aβ and ABri or ADan, could be important in influencing the rate of amyloid production or the tendency of these peptides to aggregate. Methodology/Principal Findings We have generated the first BRI2 Danish Knock-In (FDDKI) murine model of FDD, expressing the pathogenic decamer duplication in exon 6 of the BRI2 gene. FDDKI mice do not show any evident abnormal phenotype, with normal brain histology and no detectable amyloid deposition in blood vessel walls or parenchyma. Conclusions/Significance This new murine mouse model will be important to further understand the interaction between APP and BRI2, and to provide insights into the molecular basis of FDD.

Introduction BRI 2 is a type II trans-membrane protein of unknown function. The BRI gene belongs to a multigene family comprising at least three homologues in both mice and humans, BRI 1 , BRI 2 and BRI 3 (also referred to as ITM2A, ITM2B and ITM2C, or E25A, E25B and E25C, respectively) [1,2,3,8,9]. It possesses a BRICHOS domain, a conserved motif common to members of the BRI, ChM-I, SP-C and CA11 protein families, thought to have a role in the targeting of the respective proteins to the secretory pathway or to intracellular processing [10]. Proteins sharing the BRICHOS motif are dissimilar, and associate with a diverse range of phenotypes, varying from dementia to cancer and respiratory distress. BRI 2 was first described in relation to Familial British Dementia (FBD) [2], an autosomal dominant neurodegenerative disease characterized by the early onset of personality changes, memory and cognitive deficits, spastic rigidity and ataxia. In FBD, a Cterminal 34 amino acid (aa) peptide of BRI 2 accumulates as amyloid, leading to severe amyloid angiopathy of the brain and spinal cord with perivascular amyloid plaque formation, parenchymal plaques affecting the limbic areas, cerebellum, cerebral cortex, neurofibrillary tangles of hippocampal neurons and periventricular white matter changes [11]. Familial Danish Dementia (FDD), previously known as heredopathia ophthalmootoencephalica, is an autosomal dominant disease characterized by the accumulation of an amyloidogenic C-terminal 34 aa peptide of BRI 2 [3]. FDD is characterized by early onset cataracts, deafness, progressive ataxia and dementia [3,12]. Neuropathological examination of patients with FDD shows diffuse brain atrophy with a particularly severe involvement of the cerebellum, cerebral cortex and white matter, as well as the presence of very thin and almost demyelinated cranial nerves, and widespread amyloid angiopathy in the small blood vessels and capillaries of the cerebrum, choroid plexus, cerebellum, spinal cord and retina. In FDD, parenchymal compact plaques are consistently absent, whereas neurofibrillary tangles (NFTs) are the major histological finding in the hippocampus [3,12]. BRI 2 is physiologically cleaved, at the C-terminus, by furin, a calcium-dependent serine endoprotease, producing a 23 aa soluble C-terminal fragment. FBD and FDD are due to mutations in the BRI 2 gene located on chromosome 13q14 [2,8]. In FBD, a single base substitution at the stop codon of BRI 2 generates a longer open reading frame, resulting in a larger, 277 aa precursor (BRI 2-ABri , compared to the 266 aa long normal protein) [2]. In FDD, a decamer duplication in the 39 region of the BRI 2 gene, right before the stop codon, leads as well to the production of a longer, 277 aa protein (BRI 2-ADan ) [3]. The genetic defect is different, but the outcome is the same: the generation of a longer 34 aa C-terminal fragment, ABri in FBD and ADan in FDD, which accumulates as amyloid [2].
More recently, a new link between BRI 2 and Ab has been found. It has been shown that BRI 2 binds APP in a region comprising the extracellular juxtamembrane domains of both proteins, in a cis fashion [5,6,7]. This interaction leads to interference on the physiological processing of APP: BRI 2 restricts the docking of c-secretase to APP and the access of aand bsecretases to their cleavage site on APP itself. The overall result of this interaction is the reduction of the amyloidogenic processing of APP, without a direct inhibition of the general activity of the secretases [5]. BRI 2 maturation is required for this function. In fact, only mBRI 2 (and not the immature precursor) binds mature APP and inhibits its processing on the plasma membrane and in endocytic compartments [16]. Matsuda et al have further shown that BRI 3 , a member of the BRI family, binds and inhibit the action of aand b-secretases on APP [17]. Others have argued that the BRI 2 -23 wild type C-terminal peptide, released from BRI 2 by furin/furin-like cleavage, can inhibit Ab aggregation in vitro [18].
The interaction between these two amyloidogenic proteins can be of interest, especially for the design of new therapeutic strategies for AD and FBD/FDD. To this end, the processing of the BRI 2 precursor, both in the wild type allele and its mutant form must be understood. Transgenic mice bearing the British or Danish mutations have been generated, and are under investigation as models of dementias and cerebral amyloid angiopathies [19,20]. A transgenic FDD model which over-expresses the Danish mutant form of BRI 2 shows amyloid deposition in the walls of blood vessels of the cerebrum and cerebellum, parenchymal amyloid deposition and reactive gliosis, ADan amyloidosis and some signs of cerebellar ataxia [20]. Transgenic over-expression of human mutant genes has been extensively used to generate mouse models for human neurodegenerative disorders due to the need of overexpressing enough amount of the mutant protein in order to observe a phenotype. It is however possible that some disorders may, at least partially, be due to loss of function rather than a gain of toxic function due to mutations. If this were the case, a transgenic approach would be counterproductive. Therefore, we have reasoned that models reproducing the genetics of human pathologies may be worth studying since they may unveil important disease mechanisms that a transgenic approach may miss or, possibly, hide. Thus, we have generated a genetically faithful model of FDD by introducing the pathogenic human mutation in the mouse BRI 2 gene by a knock-in technology. Here, we describe the generation of this Knock-In murine model of FDD (FDD KI ) and its initial characterization.

Ethics Statement
Mice were maintained on a C57BL/6 background for several generations (at least 15). Mice were handled according to the Ethical Guidelines for Treatment of Laboratory Animals of Albert Einstein College of Medicine. The procedures were described and approved in animal protocol number 20040707.

FDD KI Mice Construction
To clone the ADan mutation allele, a fragment obtained form BAC (containing mouse BRI 2 genome) was amplified by PCR with the following primers: Fw: 59-cccAAGCTTtttttttttttttttaaagacaac-39; Rev: 59-gggAA-GCTTgaagtggtcagcagggag-39, obtaining a 7536 bp fragment, flanked by 2 HindIII sites, and comprising part of Intron 2 to a 39UTR region, up to 3850 bp from the BRI 2 stop codon. Such a fragment was cloned into a pBS vector, into HindIII sites (pBS-Ex3-6 HindIII), and used as a template for subsequent cloning.
A 340 bp fragment containing the ADan (duplication 786-795 of cDNA) mutation, plus a humanizing substitution aRg (accRgcc = ThreonineRAlanine) at the 12 th codon of exon 6, was obtained by serial PCRs, using pBS-Ex3-6 HindIII as a template. The final external primers contained restriction sites for HincII (59) and EcoRI (39). The mutated fragment (HincII-ADan-EcoRI) was inserted into the SmaI and EcoRI sites of a pBS vector yielding pBS-ADan0.3.
A NotI-XhoI fragment from pBS-ADan2.9, comprising the RA, was inserted into NotI and SacII sites of a Soriano PGK-Neo-dTA vector (blunt-ended by the SacII and XhoI digestion followed by treatment of T4-exonuclease and Klenow polymerase respectively) at 59of the Neo cassette. The XhoI site was reconstituted upon ligation of the fragment into the vector.
The construct's Left Arm was extracted from the template with EcoRI and HincII, generating a 1.1 kbp into Intron 5 up to the corresponding HincII site, 59 of Exon 6 (start of Right Arm). The fragment was thus inserted into the PGK-Neo-dTA vector at 39 of the Neo cassette and 59 of the dTA cassette.
The resulting construct was thus: The resulting construct was linearized with SalI and purified prior to injection in ES cells strain 129 by electroporation. ES culture was performed on feeder layer, and further electroporation and handling was also performed according to the methodology employed at Dept of Cell Biology, Albert Einstein College of Medicine, and according to Wakayama et al. [21]. In particular, after electroporation, ES cells were re-plated in 55 cm 2 dishes and let grow until visible clones would appear. Clones were then picked and transferred to 96 well plates in triplicates. Triplicates were either screened by PCR or frozen for subsequent use and further analysis.
Injection of the two Danish ES cell clones into C57BL/6J blastocysts was performed at the Albert Einstein College of Medicine gene-targeting facility, according to the current facility protocol (http://www.aecom.yu.edu/home/SharedFacilities/ ViewFacility.asp?ID = 30).

PCR Analysis
The PCR screening was performed using the Expand Long Template PCR System (Roche-applied-Science) with Betaine, according to the manufacturer instructions. PCR analysis of recombinant ES cells and mice was conducted with the following primers and digestion strategies to identify the correct recombinant clones and strains: ES cells: Due to the specific sequences present in the amplified PCR product, no digestion of the amplicon could be performed on the mice PCR analysis. PCR products were thus sequenced to ascertain that the targeted sequence was correctly inserted in the genomic DNA. The sequence has been deposited in GenBank (accession number GQ424832).

Southern Blot Analysis
Twenty mg of genomic DNA was digested with BamHI overnight, run on a 1% TAE agarose gel and transferred on a Hybond-N+ membrane (Amersham).
The probe was prepared by PCR from a BAC clone (RP24) with the following primers: -Left arm: Fw: 59-GACAGAGGTTCTGCCCTCAG-39 Rev: 59-ACCGAGTCGTAGGACAGTG-39 Probe size: 547 bp -Right arm: Fw: 59-CTGTGCTGCCTGACACTACTTC-39 Rev: 59-TCTGTCCATACTCCCTGTCCTT-39 Probe size: 515 bp. One mg of PCR probe was labelled with 5 mL of 32 P-dCTP (3000 Ci/mmol, ICN) and purified through a Push Column (Stratagene) according to the manufacturer's protocol. Membranes, containing the cleaved genomic DNA, were hybridized at 65uC and subsequently washed 4 times in SSC buffer (Sigma). Film was exposed to the hybridised membranes at -80uC and then developed.

General Pathology Analysis
Wild-type and FDD KI animals were studied at 18 months of age. After anesthesia, animals were perfusion-fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.2) (Sigma), after which brains and organs were removed, embedded in paraffin and sectioned. Eight mm sections were cut and mounted on poly-l-lysine-coated slides. After deparaffinization in Xylene ad rehydration, sections were stained with hematoxylin and eosin (H&E), Congo red standard methods. Mineralization was visualized in H&E sections as a deep basophilic amorphous and/or granular material.

Western Blot
Western blot was conducted on protein extracts from brains or neuronal culture of FDD KI mice and wt littermates, as described previously [5] for cells lysates and brain tissue. Briefly, primary neuronal cells were lysed in Hepes-Triton buffer (20 mM Hepes/ NaOH pH 7.4, 1 mM EDTA, 150 mM NaCl, 0.5% Triton X-100, plus protease inhibitors (PIs)) on ice for 30 min. The lysates were cleared by spinning at 20,000 g for 10 min. Equal amount of proteins of cleared lysates were loaded on SDS-PAGE and transfered onto nitrocellulose membranes; APP and BRI 2 (wild type and Danish mutants) were detected by 22C11 (Chemicon, MAB348) or BRI 2 antibody, generously provided by Dr. Haruhiko Akiyama, respectively [24].

Generation and Characterization of FDD KI Mice
The targeting strategy for the generation of the Danish mutant FDD KI mice entailed the replacement of the BRI 2 exon 6 with a mutated exon 6 carrying the FDD mutation. We generated a targeting vector for the introduction of FDD BRI 2 mutation. The vector used the floxed PGK-neo selection cassette and contains a 59 homologous region and the negative selection cassette, PGK-dt. The 39 homologous region introduced the FDD mutation and a BamHI site into the BRI 2 mouse gene. The 10 nucleotide insertion found in patients with FDD was introduced before the stop codon. In order to humanize the mouse C-terminal region of BRI 2 , an alanine (A) was substituted for threonine (T) at codon 250 (Fig. 1a) of the murine BRI 2 gene.
The linearized targeting vector was transfected into 129 ES cells by electroporation. In the presence of the positive selection drug, G418, only those clones in which the PGK-neo selection cassette was integrated and the PGK-dt cassette was removed by homologous recombination would survive. ES cell clones carrying the targeting vector by random, non-homologous integration, were eliminated due to expression of diphtheria toxin.
After selection in G418-containing medium, ES cell clones carrying the proper homologous recombination and the tADan (targeted ADan) allele were identified by PCR for 59 region (i.e. Left Arm: if homologous recombination had occurred these primers would amplify a product of 1.67 kb) and for the 39 region (i.e. Right Arm: if homologous recombination had occurred these primers would amplify a product of 3.4 kb). Out of the ,600 screened ES clones, we found three clones in which the Danish mutation was inserted in one of the BRI 2 alleles.  Fig. 1b. Also, PCR amplification and digestion, as specified in the Material and Methods section, was used to check the proper insertion of the construct in the genomic DNA and the removal of the Neo cassette (not shown).

342) ES clones are shown in
The occurrence of homologous recombination was confirmed by sequencing the PCR products and by performing Southern blot analysis (Fig. 1c). DNA derived from individual BRI 2 tADan/+ ES clones was digested with BamHI, gel separated, blotted into a nylon membrane and hybridized with either the 59 or the 39probe. The 59 probe hybridizes with a ,11.9 kb fragment derived from the wild-type locus. Homologous recombination at the 59 homologous region yields a ,8.9 kb fragment upon BamHI digestion due to the introduction of the BamHI site and the PGK-neo selection cassette.

ES clones BRI 2
ADan/+ 344 and BRI 2 ADan/+ 339 carry a wild type allele (11.9 kb) and a recombined allele (8.9 kb). Of note, the 11.9 kb and 8.9 kb bands had a similar intensity, indicating that 50% of the BRI 2 alleles are wild type and 50% are recombined, and proving that the ES cells we have selected are a clonal populations. Similar results were obtained when homologous recombination at the 39 site was assessed. In this case the 39 probe detected a wild-type ,11.9 kb fragment and a recombinant 4.7 kb fragment, due to the introduction of the BamHI site. These two Danish ES cell clones (129, agouti coat colour), carrying the correct site-specific homologous recombination, were injected into C57BL/6J blastocysts (black coat colour) at the Albert Einstein College of Medicine gene-targeting facility. The resulting chimeras with a high proportion of agouti coat colour (i.e. with a high relative contribution from the injected ES cells) were backcrossed to C57BL/6J mice to obtain heterozygous BRI 2 ADan/wt , which were identified by PCR and Southern analysis as described above (not shown) using tail DNA. Heterozygous mice were crossed to Meu40-Cre mice to obtain Meu40/BRI 2 tADan/wt animals. Cre is a bacteriophage P1-encoded recombinase that catalyzes site-specific recombination between two 34 bp loxP recognition sites, resulting in the excision of the intervening DNA sequences. The resulting mouse has been named FDD KI .

General Characteristics and Pathology of the FDD KI Mice
Mice presented with no growth abnormalities, thrived at appropriate age, as their wild type littermates. Up to age 18 months, no susceptibility to infections was noted. The animals are active and alert.
Before the pathology examination, at 18 months of age, they presented in good body condition with adequate body fat, with no discharges or secretions from nostrils, conjunctiva, aural, urogenital or anal openings. Some mice also presented heart valve melanosis, which is a common finding in aged mice of various strains. Other sporadic findings, which are common in aged mice, were small foci of hepatocellular necrosis, vacuolation of and degeneration/ regeneration of kidney tubular epithelial cells, mineralization in the kidney tubules in the pelvis and cortico-medullary junction (not shown). Overall, these mice have lesions that are commonly found in older mice and are considered age-related, spontaneous lesions. All of the lesions found were considered to be within the normal limits for age-related lesions. One of the animals presented with a uterine histiocytic sarcoma and hydronephrosis, both of which are found in older, untreated mice at low incidence.

Neuropathology
A few animals presented a small amount of dark basophilic material deposited on either side of the lateral thalamus consistent with mineralization (incidental, Fig. 2d), which is a common finding in aged mice of various strains, and the deposits develop along the basement membranes of the vasculature and may contain calcium and phosphorus. The finding of dark brown pigment along the meninges of the olfactory bulbs is consistent with melanosis (incidental, Fig. 2a), which is also a common finding in mice in dark pigment mouse strains such as C57BL/6 or having that strain the their background. All the brain tissue and the spinal cord were stained with Congo red for potential amyloid deposition. All of the stained tissue was negative for amyloid (not shown). Samples of H&E staining of cerebral cortex, hippocampus, cerebellum, thalamus and spinal cord are shown in Fig. 2a-e.
Brain sections of 18-month-old mice were examined for FDDrelated pathology, particularly for amyloid deposition in brain parenchyma and vessel walls. For comparison, we included sections of an age-matched female FDD-Tg mouse, which expresses the Danish mutant form of human BRI 2 under the mouse prion protein promoter [20]. H&E staining of Knock In mice (Fig. 3c, 4a) showed no significant loss of neurons or noticeable malformations, and this observation was further confirmed when mutant sections were compared with those of wild type littermates (Fig. 3a, 4d). Immunostaining using anti-GFAP (Fig. 3b, d) and anti-ubiquitin (not shown) antibodies did not reveal significant differences with wild type mice in levels of activated astrocytes or presence of ubiquitinated material, respectively. Staining using Congo red or ThS did not show the presence of amyloid deposits (Fig. 4b). Amyloid deposition in FDD-Tg mice is particularly strong in the cerebellum (Fig. 4e) but not in Knock In mice (Fig. 4b) and only FDD-Tg mice showed immunoreactivity using Ab 1700, specific for the ADan amyloid peptide [20] (Fig. 4f versus 4c). Immunoreactivity was not seen in non-transgenic littermates, FDD-Tg and Knock In mice using antibody 4G8 (not shown). These results indicated that the brain of Knock In mice maintained normal morphology of aged mice and were free of amyloid deposits at 18 months of age.

Biochemical Analysis of BRI 2-ADan Expression
To determine whether the mutant proteins are expressed in vivo, we made protein extracts form neuronal cultures of wt, BRI 2 ADan/wt and BRI 2 ADan/ADan mice. Western blot analysis of these samples shows three phenomenona. 1) A common, major band of about 32 kDa is readily visible in all three samples. This band corresponds to the post-furin cleavage mature BRI 2 (mBRI 2 ) polypeptide because mBRI 2 is identical regardless of whether it is derived from the wild type or the mutant precursor protein. 2) A much less abundant protein of ,33 kDa which is visible in the wild type and BRI 2 ADan/wt sample but not the BRI 2 ADan/ADan lysate. This band corresponds to the wild type, immature BRI 2 precursor (imBRI 2 ). The absence of this band in the BRI 2 ADan/ADan neurons is attributable to the fact that these cells have both BRI 2 alleles mutated. 3) As expected, the BRI 2 ADan/wt sample expresses all three BRI 2 polypeptides, i.e. wild type imBRI 2 , BRI 2-ADan precursor and mBRI 2 . It is worth noting that, while immature murine BRI 2 (imBRI 2 ) is barely detectable, BRI 2-ADan is more markedly expressed (see BRI 2 ADan/ADan neurons, Figure 5) in this specific experiment. Whether this is related to the artificial absence of the wild type protein (this does not happen in FDD since the patience have only one mutated allele) or reflects some interesting biology of BRI 2-ADan , remains to be determined. Notably, APP expression and maturation are comparable in FDD KI and wt mice (Fig. 5, lower panel).

Discussion
In this manuscript we present an initial characterization of a Knock In model for a human neurodegenerative disorder, FDD. This Knock In model is strategically different from the conventional genetic approach to cerebral amyloidosis. Traditionally, mouse models for human dementias are based on a transgenic approach in which human mutant proteins that cause familial forms of dementia are over-expressed under the control of brainspecific promoters [25,26,27]. This approach has both limitations and advantages. The potential limitations are linked to the fact that these models are genetically incongruous with the human diseases. These dementias have an autosomal dominant way of transmission and affected subjects have a wild type and a mutant allele. On the contrary, the commonly used animal models overexpress, from an artificial promoter, several copies of a mini-gene coding the mutant protein. The mutant gene is therefore expressed in a spatio-temporal manner different from the natural alleles and in cells with two copies of wild type alleles. However, transgenic approaches have successfully reproduced cerebral amyloidosis in mice, and the reproduction of those lesions is one of the main parameters used to endorse an animal model as representative of the human disease. The opposite is true for a Knock In approach. In this genetic paradigm, the pathogenic human mutation is inserted in the mouse allele. Thus, the Knock In mouse is genetically faithful to the human pathology. However, the Knock In model may not successfully reproduce the amyloid lesion in the mouse. The model that we present here is a clear demonstration of the above-mentioned assumptions. A comparison of the FDD-KI model that we have created with the FDD transgenic model generated by Vidal et al [20] illustrates the superior power of a transgenic approach as far as the reproduction of amyloidogenic lesions are concerned.
In spite of this limitation, we believe that there is some merit to a Knock In approach. Human familial dementias may have a pathogenic component due to loss of function caused by the mutation [28,29,30,31,32]. This component may participate to the pathogenic process together with a gain of toxic function due to amyloidosis. A loss of function phenotype would be hidden by a transgenic approach for two reasons: 1) the two endogenous wild type alleles can support synthesis of sufficient amounts of functional wild type protein: 2) the over-expression of a partial loss of function mutant protein may augment rather then decrease that function, as may happen in the disease. It would be erroneous to assume that the absence of obvious neuropathological lesions (such as plaques, NFTs and/or neuronal cell loss) indicate absence of clinical pathology. Functional deficits may underlie the very first clinical manifestations of neurodegenerative diseases, and memory deficits may be caused not by gross anatomical changes but by subtle dysfunctions of the neuronal network. In this framework, determining the absence of neuropathological lesions in FDD KI mice is intrinsically important. The behavioral characterization of this mouse model will test this hypothesis. In addition, the FDD KI mouse will be useful to clarify the change of trafficking or processing of Danish mutant of BRI2 protein without potential artifacts due to over-expression in transgenic models. Finally, the FDD KI mouse will be instrumental in studying the interaction between mutant BRI 2 and APP in vivo and how the Danish mutation affects the inhibitory role of BRI 2 on APP processing in vivo.