Generation of a Matrix Gla (Mgp) floxed mouse, followed by conditional knockout, uncovers a new Mgp function in the eye

The ability to ablate a gene in a given tissue by generating a conditional knockout (cKO) is crucial for determining its function in the targeted tissue. Such tissue-specific ablation is even more critical when the gene’s conventional knockout (KO) is lethal, which precludes studying the consequences of its deletion in other tissues. Therefore, here we describe a successful strategy that generated a Matrix Gla floxed mouse (Mgp.floxed) by the CRISPR/Cas9 system, that subsequently allowed the generation of cKOs by local viral delivery of the Cre-recombinase enzyme. MGP is a well-established inhibitor of calcification gene, highly expressed in arteries’ smooth muscle cells and chondrocytes. MGP is also one of the most abundant genes in the trabecular meshwork, the eye tissue responsible for maintenance of intraocular pressure (IOP) and development of Glaucoma. Our strategy entailed one-step injection of two gRNAs, Cas9 protein and a long-single-stranded-circular DNA donor vector (lsscDNA, 6.7 kb) containing two loxP sites in cis and 900–700 bp 5′/3′ homology arms. Ocular intracameral injection of Mgp.floxed mice with a Cre-adenovirus, led to an Mgp.TMcKO mouse which developed elevated IOP. Our study discovered a new role for the Mgp gene as a keeper of physiological IOP in the eye.


Results
CRISPR/Cas9-mediated gene editing for the generation of an Mgp.floxed mouse. The MGP gene is highly expressed in the trabecular meshwork of at least human 29,[34][35][36][37] , porcine 44 and mice 38,39 . Because the Mgp KO 18 is lethal and because of the need to address the potential relevance of MGP in trabecular meshwork physiology, we set up to generate an Mgp.floxed mouse that would then allow the creation of a trabecular meshwork cKO (Mgp.TMcKO). Our strategy made use of the CRISPR/CAS9-mediated gene editing technology strategies to allow insertion of two loxP sequences in the same Mgp allele (in cis). A diagram with the overall design of the needed elements for this gene editing is shown in Fig. 1. We identified insertion sites in the mouse genome, generated optimized gRNAs and Cas9 expression plasmids and constructed a donor DNA vector with two restriction sites-flanked loxP sequences (Fig. 1).
The Mgp gene contains four exons with initiation of translation at 66 nt in of exon 1 and termination at 145 nt in of exon 4 (MGI: 96976; GeneID: 17313). Our Mgp.floxed mouse design to allow the subsequent generation of an Mgp cKO, entailed deletion of a region containing exons 3 and 4, which encode 70% of the protein's C-terminus (aa 12 to 104 Mgp) and will produce a small truncated inactive protein. This truncated protein, encoded by remaining exons 1 and 2, would be predicted to be 31 aa, from which 19 aa comprise the signal peptide. Thus, the secreted product would be approximately 1.5 kDa. The exon 4 deletion comprises also the mRNA's 3′ UTR and would most likely affect mRNA levels. In addition, this DNA region includes critical functional binding sites 51 whose elimination by themselves will render an inactive Mgp protein. To find loxP insertion sites, intron 2 and the 3′ proximal genomic region of Mgp were searched using https ://crips r.mit.edu/ and https ://zlab.bio/ guide -desig n-resou rces databases to identify potential Cas9 guides RNA with low off-target mutagenesis. We identified sites at 374 nt upstream of exon 3 in the sense strand for the 5′ loxP site, and 904 nt downstream of exon 4 in the antisense strand. Predictive algorithms yielded two CRISPR RNA (crRNA) sequences with numerical off-target sites scores of 77 and 79 for the 5′ loxP insertion region, and two sequences with scores of 65  www.nature.com/scientificreports/ www.nature.com/scientificreports/ the 3′ loxP insertion region. The four crRNA candidates were fused to the Cas9 binding scaffold trans-activating CRISPR RNA (tracrRNA) to generate single guide-RNAs (gRNAs). For the generation of each gRNA, two complementary 20 nt oligonucleotides per site were synthetized containing an addition of 4 overhanging nt (5′ TATA to the sense strands and 5′ AAAC to the antisense strands to allow cloning) ( Table 1 for the two selected crRNAs) (Fig. 2). Complementary oligonucleotides were annealed and cloned into an in-house custom-made T7 expression plasmid (pT7.gRNA g79/92) between two BsaI proximal cutters located downstream of the T7 promoter and upstream of the 77 bp tracrRNA (Cas9 binding site). In vitro transcription with T7 RNA polymerase produced single 97 bp gRNAs which were purified on RNeasy columns. Cas9 protein cDNA 13 was cloned into prokaryotic expression plasmid pET-28a(+), expressed in bacteria and purified by Ni-NTA agarose. DSB cutting efficiency of the four transcribed gRNAs was assayed in vitro with the Cas9 protein and with C57BL/6J (B6) mouse genome DNA fragments amplified with primers P1f/P1r and P2f/ P2r ( Table 1). The primers surround the 5′ and 3′ potential loxP insertions sites and produced 785 bp and 985 bp fragments respectively. These PCR amplicons (250 ng) were incubated with 600 ng each of the corresponding gRNA and 100 nM Cas9 at 37 °C. Gel electrophoresis analyses of resulting fragments showed that gRNAs g79 and g92 (5′ and 3′ loxP sites respectively) were the more efficient and each rendered DNA fragments whose size corresponded to the utilization of the expected DSB close to 100% efficiency (Fig. 2).
Several attempts to insert two loxP sites in cis in the Mgp gene by injecting two separate donor DNAs failed to generate a sole Mgp.floxed mouse allele. Four microinjection attempts were made with single-stranded oligomers, and 2 attempts with supercoiled donor vectors without any success (Table 1S). To overcome this problem, we designed a single donor vector containing both, the 5′ and 3′ loxP sites in cis. The vector, pMgp.Floxed donor, was generated by In-Fusion cloning of three DNA elements amplified from the B6 mouse genome and subsequent seamless insertion into a basic in-house generated pUC backbone plasmid (Fig. 3). The first element, a 916 bp 5′ homology arm (plus 15 fusion vector bp), contained sequences from 226 bp upstream to 655 bp downstream of exon 2 (primers P3f/P3r). The second element, a 2246 bp donor fragment containing both the 5′-and 3′-loxP sequences plus KpnI restriction sites, extended from 655 bp downstream of exon 2 to 1000 bp downstream of exon 4 (primers P4f/P4r). The third element, a 670 bp 3′ homology arm fragment (plus 15 fusion vector bp) extended from 1000 to 1655 bp downstream of exon 4 primers (P5f/P5r) ( Table 2, Fig. 3). To allow for In-Fusion cloning, the forward primers corresponding to the sense strand of each element, contained a 15 bp sequence that complemented with 15 bp of the 5′ end of the antisense strand of the previous element (Table 2). In order to secure accurate insertion of the loxP DNA sequences plus restriction sites in the donor fragment element, these forward and reverse primers (P4f/P4r) were made to contain 160 bp oligonucleotides each (Table 2). After In-Fusion cloning, the resulting 6738 pMgp.Floxed donor plasmid containing a total of 3802 bp of mouse Mgp gene with inserted loxP sites in the noncoding region, was confirmed by sequencing (Fig. 3).
Microinjection of CRISPR/Cas9 with long-single-stranded-circular DNA (lsscDNA) to generate the Mgp-floxed mice. Because of our failed previous attempts and, of the reported success of using singlestranded donor DNAs 10,52,53 , our plasmid donor vector was converted to a single-stranded circular DNA before injection. For this, a new gRNA with a vector-matching sequence at 480 nt downstream of exon 4 was generated (AAT GGT CCC ATA TGT GAC TA(tgg)). The double-stranded (ds) DNA donor vector was then incubated (1 h, 37 °C) with the gRNA and a mutated form of Cas9 (mutation D10A) which causes a single-stranded (ss) DNA break, rather than the DSB caused by the Cas9 WT protein 13 . The resulting nicked DNA was purified by phenol/ www.nature.com/scientificreports/ chloroform extraction and treated with Nuclease III, which removed nucleotides from the 3′ end of nicked DNA and thus rendered a long-single-stranded-circular DNA molecule (lssc). After the treatment, the lssc DNA was purified by a Qiaquick spin column and dialyzed (Fig. 3b). A 2 µl mix containing purified Cas9 (400 nM), selected 5′ and 3′ gRNAs (g79 and g92, 100 ng each), and the lssc form of the pMgp.Floxed vector (20 ng) was injected into 69 embryos (out of 139 embryos obtained from 12 superovulated females). Ten embryos lysed during injections, and the remaining 59 were cultured overnight to yield 21, 2-cells embryos. The 21 embryos were implanted into one recipient CD-1 female mouse which produced 6 pups: #53, #54, #55, #56, #57, #58 (hereafter termed #1 to #6) (Fig. 4a). Genotype characterization of the pups was conducted by PCR according to the strategy depicted in Fig. 4b. Four primer pairs (P6f/P6r, P7f/ P7r, P8f/P8r, P9f/P9r) were designed to identify the absence of 5′ vector sequences, the presence of the 5′-and 3′ loxP sites and the absence of 3′ vector sequences, respectively ( Table 3). Presence of vector sequences would denote random integration of the donor DNA in the genome. Results of the DNAs from each of the six founders amplified with each of the primer sets are shown in gels (1), (2), (3) and (4) from Fig. 4c. Controls included DNA from the B6 mouse (negative), a non-template control, and a DNA mix of B6 plus donor vector (positive). Of the six pups, only one pup (#5) was simultaneously positive for 5′ loxP and 3′ loxP (505 bp and 324 bp bands in Fig. 4c gels (2) and (3) and negative for vector sequences (677 bp and 834 bp in Fig. 4c gels (1) and (4). The Oligonucleotides annealed for the generation of crRNAs used in the transcription vectors (left), pET-28a(+) plasmid used to produce the human codon optimized Cas9 protein (center) and mouse DNA diagram showing location of the primers (arrows) used to amplify the DNA fragments used in the assay (right). (b) Incubation of in vitro transcription from gRNA plasmids, Cas9 and PCR-generated DNA (left); electrophoretic separation of the resulting DNA fragments in a 2% agarose/TAE gel stained with ethidium bromide (right). M: 1 kb Plus DNA Ladder (GeneRuler, ThermoFisher). For the 5′ site, the P1f/P1r 785 bp amplified fragment incubated with the gRNAs resulted in a doublet band corresponding to the 425 bp and 358 bp fragments; the gRNAg79 was more efficient than gRNAg77. For the 3′ site, the P2f/P2r 985 bp amplified fragment incubated with the gRNAs resulted in 837 bp and 248 bp fragments; the gRNA92 was more efficient than gRNAg65. Size of the fragments correspond to the predicted DSB (relevant gel bands encased in green boxes for easier visualization). Lower MW bands correspond to primer dimers. Incubations controls without the gRNAs and Cas9 protein were negative. www.nature.com/scientificreports/ female with floxed allele, founder #5, was crossed 3X with a B6 male, and genotype of their F1 genome was reconfirmed with the same PCR primers and same strategy described above (presence of loxP sites and absence of vector sequences) (not shown). Genotype of six F1 pups (#5.5, #5.6, #5.7, #5.8, #5.9 and #5.10) was additionally evaluated by southern blot hybridization (Fig. 5). Figure 5a shows the strategic design: three PCR-generated probes were made to uniquely identify the 5′, the internal, and the 3′ DNA fragments resulting from a KpnI-BsrgI digestion of the Mgp.WT and Mgp.floxed alleles (Fig. 5a). Sequences of Probe 1 (425 bp) obtained by amplification with P10f/P10r primers, Probe 2 (610 bp) amplified with primers P11f/P11/r, and Probe 3 (660 bp) amplified with P12f/P12r are shown on Table 4. Hybridizations of the digested WT allele rendered an 8.9 kb DNA band with the three probes. Because of the insertion of the KpnI sites with the loxP sequences ( Table 2), hybridization of the digested Mgp.floxed allele with Probe 1 rendered the expected 2.9 kb band, while hybridization with Probe 2 and Probe 3 rendered 3.7 kb and 2.1 kb DNA fragments respectively. Mice #5.5, #5.8, #5.9 and #5.10 (2 males and 2 females) were heterozygotes for the Mgp.floxed allele (Mgp floxed/+ ), while mice #5.6 and #5.7 were WT (Mgp +/+ ) (Fig. 5b). This result confirms the sequential position of the 5′-and 3′-loxP sites on a single allele.
Two intercrosses of the Mgp floxed/+ pups led to the generation of the first homozygous breeding pair Mgp floxed/floxed mice. Litters from following generations of these mice were genotyped for about four generations  Table 2. Cloning details are described in "Methods" section. (b) Conversion of the 6,738 bp ds pMgp. Floxed donor plasmid to the injection grade long-single-stranded-circular (lssc) molecule. To produce the lssc molecule, the ds DNA plasmid was first nicked by incubation with a gRNA (designed to target sequences 480 bp downstream of Ex4) and a mutated Cas9.D10A protein that produces ss breaks. The nicked plasmid was then treated with Nuclease III to degrade one strand from the 3′ end and purified. Yellow: 5′ homology arm; grey: edited donor fragment with loxP sequences; green: 3′ homology arm. black: pUC background vector.  (Fig. 2Sa). All experiments were performed in independently generated MIA cells lines, at passages 1 and 2 (see "Methods" section). Two days post-infection, DNA, RNA, and protein from these cells were extracted and assayed for the presence of a floxed DNA fragment, levels of Mgp transcripts, and the absence of the Mgp protein. For the DNA evaluation, PCR primers P14f/P14r (sequence in methods) were designed to yield a 661 bp amplimer band of the floxed/recombined allele, and a 2804 bp of the unrecombined one (Fig. 2Sb). A second primer pair, P15f/ P15r (392 bp, "Methods" section), was designed to be outside the Mgp loxP region of the gene and thus validate amplification levels of the treated and control templates. For evaluation of the Mgp-specific transcript levels, extracted RNAs from the treated and control dishes were reverse-transcribed and Taqman-PCR amplified with mouse Mgp Taqman probe annealing to exon 1-2 boundaries. The Myoc gene and 18S Taqman probes were used   (a) Schematic approach to generate Mgp.floxed alleles using a CRISPR/Cas system entailing two gRNAs (5′ and 3′) and a one long-singlestranded-circular DNA (lssc) composed of the donor/edited region flanked by two in cis loxP sites plus 5′ and 3′ homology arms (Fig. 3). Sixty-nine embryos were injected and of those, 21/2-cell surviving embryos were implanted in one recipient CD-1 female, which delivered six pups (#1 to #6). (b) Diagram of the Mgp.floxed allele and PCR strategy to analyze the genome of the F0 pups. Arrows indicate primers used to PCR-amplify pups DNA to identify each genotype. P6f/P6r (to identify absence of 5′ vector sequences, 677 bp product); P7f/P7r (to identify presence of 5′ loxP site, 505 bp product); P8f/P8r (to identify presence of 3′ loxP, 324 bp product) and P9f/P9r (to identify absence of 3′ vector sequences, 834 bp product). Primer sequences are shown in Table 3 To first confirm positive gene delivery, a set of whole globes (n = 2 to 4 eyes per group), were embedded in OCT at 7 days post-injection to be able to observe GFP expression. Cryosections were evaluated for GFP fluorescence on an Olympus X71 microscope equipped with a DP80 monochrome camera. All injected eyes showed that the intracameral viral injections delivered the GFP protein mainly to the trabecular meshwork/Schlemm's canal region (Fig. 6a). To then determine whether the Mgp gene was floxed/recombined in the angle tissue after Cre delivery, a second set of eyes was analyzed at 75 days post-viral injections using the eye globes from the IOP physiology experiments (see below). DNA was extracted from the iridocorneal angle strips of individual eyes and subjected to PCR amplification using the same P14 and P15 primer-pairs described above for the cells (Fig. 6b). Group 1 showed the Mgp.floxed allele recombined (661 bp fragment) at slightly lower full floxing efficiency of the 2804 bp unrecombined DNA. While in vitro we observed a complete floxing of the DNA, the slightly lower in vivo efficiency may be the result of the challenging technical delivery of the intraocular injection as well as of variations of the microsurgery of the trabecular meshwork strip, rather than to the ability of floxing the gene. No recombined DNA fragment was observed in control groups 2, 3 and 4. Likewise, a western blot with protein extracts showed the presence of Mgp protein only in control groups 2, 3 and 4 and not in the experimental Group 1 (Fig. 6c, left). Although it is known that the natural flow of aqueous humor delivers intracamerally injected viruses mainly to the trabecular meshwork, the extent of the differential delivery to other anterior segment tissues was not known. For this, we tested the presence of the Mgp floxed/recombined allele in all anterior segment tissues bathed by the aqueous humor. DNA and protein from trabecular meshwork, iris and cornea were extracted at 7 days post-injected whole globes, amplified and assayed by gel electrophoresis and western blot. Results shown in Fig. 6c right/top corroborated that the preferred targeted tissue after intracamerally injection of the Ad.GFP. Cre is the trabecular meshwork, which shows a close to complete recombination of the gene. It also shows that a small fraction of the virus enters the iris and the cornea and recombines their Mgp.floxed allele. Importantly, the iris and cornea tissues do not express the Mgp gene and thus, do not produce Mgp protein 38   www.nature.com/scientificreports/ injection baseline IOP measurements, IOPs were monitored weekly for 10 weeks using the TonoLab. The average IOP baseline values of all Mgp.floxed mice eyes used here (n = 80 eyes) was 8.2 ± 0.11 mmHg and it was not significantly different than that of the 8.0 ± 0.10 mmHg baseline of the B6 (n = 22) (p = 0.31). The mean absolute Figure 5. Confirmation of Mgp.floxed F1 genotype and loxP insertion. Female founder #5 was crossed with a B6 male and the genome of their six F1pups (#5.5 to #5-10) evaluated by southern slot hybridization. Tail DNA was digested with KpnI-BsrgI restriction enzymes, run in 0.7% agarose gels, blotted to a membrane and hybridized to three DIG-labeled oligonucleotide probes. Two of the probes located outside the target vector to assure correct genomic loxP insertion. The sequence of the probes and of the primers to generate them are shown in Table 4; experimental details are described in "Methods" section. (a) Diagrams of the Mgp.WT and Mgp.floxed alleles showing the location of the restriction enzymes, the three probes and the distance between the sites in kb. Expected fragment size in the both alleles upon the double enzyme digestion (note that KpnI sites were introduced during the generation of the donor vector, shown in Table 2); yellow and green original vector 5′ and 3′ homology arms. (b) Southern blot membrane containing the DNA from the six pups hybridized sequentially to the three probes after stripping in between hybridizations; relevant gel bands encased in green boxes for easy visualization. flox: floxed. M1: DIG-labelled DNA molecular markers II (La Roche); M2: 1 kb Plus DNA Ladder, which cross-reacts with the probes (GeneRuler, ThermoFisher). All genomic DNA fragments from both alleles showed the expected size. Mice # 5.6 and #5.7 were WT; mice #5.5, #5.8, #5.9 and #5.10 were heterozygous for the correct recombined allele.  (Fig. 7a). www.nature.com/scientificreports/ Additional IOP parameters from the four groups are shown in Fig. 7b,c. The mean integral IOP (area under the curve, AUC) (defined as the mean of the cumulative IOP that each eye received for the duration of the experiment) was 1129.9 ± 32 mmHg-d for the experimental Mgp.TMcKO group (Fig. 7b,c). For the three control groups, at the same time-period, the mean integral IOPs were 669.1 ± 3.9, 644.4 ± 6.7 and 610.7 ± 8.4 mmHg-d respectively (Fig. 7b,c). The mean integral IOP difference between experimental and control groups was highly significant for all comparisons (Fig. 7b,c). Considering the whole study period, the Peak IOP in the Mgp.TMcKO mice group at 75 days was 32 mmHg (full range of all eyes 14.0 to 32 mmHg). For the three control groups, at the same time period, the Peaks' IOP were 11.0 mmHg (range 8.3 to 11 mmHg), 9.6 mmHg (range 8.3 to 9.6 mmHg) and 9.3 mmHg (range 7.0 to 9.3 mmHg) respectively. At the last 75 days point, the mean ΔIOP from baseline was

Discussion
To the best of our knowledge, and despite the relevance of the MPG gene in an increased number of human diseases, no Mgp.floxed animal model has to date been generated. In this study, we show that a stepwise strategy led to the successful construction of the Mgp.floxed mouse. The generation of this mouse is particularly important because the Mgp.KO is lethal and the only way to study the function of the gene in specific post-natal tissue is by generating tissue-specific cKOs and ablating the gene locally. We attribute the success of the first F0 with floxed allele Mgp floxed/+ mouse to a combination of events that included selecting the correct loxP insertion sites in the Mgp gene, low off-target gRNAs, and the construction of a singular donor plasmid. This particular donor plasmid contained a 3802 bp Mgp donor DNA fragment which included both 5′ and 3′ loxP sites in cis, inserted restriction sites for subsequent validation of the mouse, and long homology arms to facilitate accurate integration. The inserted restriction sites allowed confirmation of the accurate recombination in the F1 generation. Based on our previous failed attempts (Table 1S), in this study we injected the embryos with two gRNAs, the purified Cas9 protein, and the entire plasmid donor DNA (6.7 kb) in a single-stranded circular form. This long-single-strandedcircular DNA was generated by using a vector-specific gRNA, a mutated Cas9 and the nuclease III 3′ end activity. Our rationale for using the lsscDNA was that the published methods for generating long linear SSDNA involved error-prone synthesis methods and raised the concern that unwanted mutations might be co-introduced with the loxP sites. However, we really don't have the data to emphasize this method as being novel and better than other donor methods. The use of a long-single-stranded linear DNA (lssDNA) had previously been reported to increase efficiency of obtaining floxed alleles 10,52,53 with such efficiency depending on its length 52,54 . In here, a much larger donor plasmid DNA fragment (3.8 kb) contained in a ss circular DNA donor vector, resulted in the correct floxed allele. At this time though, we are unable to say whether a contaminant double-stranded molecule in the lssc preparation, could have contributed to the generation of the Mgp.floxed allele.
Because of the high abundance of the Mgp gene in the trabecular meshwork/outflow pathway tissue 29-31 , of its expression induction by glaucomatous-causing agents 34-36 as well as its involvement in IOP homeostatic response 35 , our laboratory has had a long-time interest in identifying the functional role of this gene in this tissue 36,45,55 . It is well established that dysfunction of the trabecular meshwork in its ability to regulate IOP leads to the development of glaucoma, a blinding disease with enormous global health repercussions. Until now, eye studies on this gene have mainly been conducted using primary human cells and organ cultures but not in living animals. Our first transgenic mouse, Mgp-Cre.KI 38 confirmed in vivo the localized high expression of the gene to the eye outflow pathway region. To focus now on function, in this study we generated a second mouse, Mgp.floxed, which would allow us to subsequently generate an Mgp.TMcKO. For this, and given our ability to deliver genes to the trabecular meshwork by intracameral injections in different animal species 56-59 , we chose an adenovirus vector to target the Cre-recombinase enzyme to the trabecular meshwork. Although adenovirus delivery expresses the transgene for a short, limited time we show here that 7 days post-injection appeared to be sufficient for the delivered enzyme to recombine most of the trabecular mouse genomic DNA in vivo (Fig. 6). Concurrently, we also saw that a very low proportion of the virus entered surrounding tissues facing the anterior chamber other than the trabecular meshwork. Iris and cornea DNA were minimally recombined. This finding (Fig. 6c-right), together with our previously published result that Mgp is not expressed in the iris nor cornea 38 , validated the intracameral injection route of administration of the Ad.GFP.Cre as a reliable system for the generation of the Mgp.TMcKO.
The Mgp.TMcKO developed elevated IOP. We observed a steady IOP increase in the mouse cohort injected with the Ad.GFP.Cre while no increase was observed on any of the uninjected cohorts of the Mgp.floxed mice, of the Mgp.floxed mice injected with Ad.GFP or of wild type mice B6 injected with Ad.GFP.Cre. The increase was gradual. It started at 5 days post injection, became significant at 12 days and continue to increase to 75 days, when the experiment was terminated. Because the biochemical characterization of the trabecular meshwork tissue of the Cre-treated mice showed the absence of the Mgp protein, these results strongly indicate that the presence of Mgp appears to be essential for the maintenance of physiological pressure in the mouse eye.
The potential mechanisms as to how the specific ablation of this gene in the outflow pathway leads to an elevated IOP in the mice are currently being investigated. The fact that the observed increase of pressure occurs gradually is an indication that the Mgp gene is not causing a sudden obstruction of the outflow pathway but rather causes a metabolic cellular change whose accumulation would lead to changes in the ECM that increase IOP. Given the classical known function of Mgp as an inhibitor of calcification, one would first suspect that the outflow tissue without this gene would undergo an extracellular calcification process which in turn could be translated in an increase of stiffness. It has been established that the trabecular meshwork from glaucomatous patients is stiffer than that of the controls 60 , and that glaucomatous effectors such as TGFβ2 induce stiffness 55,61 . The calcification process in the vascular system is classified in macrocalcification and microcalcification depending on the pattern and size of the calcified lesions 62 . In atherosclerosis, macrocalcification is manifested by the formation of large calcified lesions (plaques) while microcalcification is associated with spotty, granular calcification and the progress of the disease 63  www.nature.com/scientificreports/ the gradual increase of IOP observed here in the Mgp.TMcKO would be more in tune with the presence of a microcalcification rather than a large calcification lesion. Mgp is secreted in matrix calcifying vesicles. Another interesting possibility would be that of the involvement of Mgp in maintaining the extracellular calcium homeostasis, a recently described action of the gene shown to occur during sperm maturation 33 . In the trabecular meshwork, Mgp is carboxylated, therefore active. In the sperm, it was shown that uncarboxylated, inactive MGP which loses its ability to bind calcium, contributes to high Ca concentrations in the epididymal lumen 33 . It would be interesting to determine whether the absence of Mgp in the outflow pathway cells could also disturb the Ca levels on the ECM which could in turn affect extracellular calcium entry and modulate trabecular meshwork function 64 . Another potential mechanism responsible for the elevated IOP in the Mgp.TMcKO could involve BMP2. It is well established that MGP binds to, and sequesters BMP2 36,65,66 . To this end, we had earlier shown that overexpression of BMP2 in rat trabecular meshwork by intracameral injection of Ad5BMP2 induced elevated IOP 67 . An increase in BMP2 triggered by the absence of Mgp could alter the physiology of the trabecular meshwork through several pathways. As reasoned in Buie et al. 67 , besides provoking an increase in calcification 68,69 , BMP2 could influence IOP by its consequent increases in Col1A1 67 , a trabecular meshwork ECM component, originally reported to be enhanced under glaucomatous conditions 70 . More recently Col1A1 has been found to be genetically linked to glaucoma cohorts 71 . In the same Buie's paper 67 , it was also discussed the well-known crosstalk between BMP2 and the Wnt/β-catenin pathways and the activation of each of the pathways by the other 72,73 . There is an apparent paradox though, that while BMP2 activates the Wnt/β-catenin pathway 73 and causes elevated IOP 67 , an antagonist of the Wnt/β-catenin pathway, SFPR1, causes also elevated IOP 74 . However, some studies have shown that BMP2 can also inactivate Wnt/βcatenin 75 and that its activation or inhibition of the pathway is dependent on the status of the expression of p53 and SMAD4 tumor genes, like in colorectal cancer cells 76 . It would seem then plausible that in the trabecular meshwork, BMP2 could be inhibiting the Wnt/β-catenin pathway.
The close relationship of MGP with elastin brings another interesting avenue on to the potential role of Mgp in maintaining tissue integrity and physiological IOP. On the eye's outflow tissue, elastin is an essential component of the trabecular meshwork ECM 77 . Together with collagens and proteoglycans, elastin forms part of the beams of the corneoscleral region upon which lie the trabecular meshwork cells. A network of elastin-like fibers is also present in the juxtacanicular region and the thickness of these fibers has been associated with glaucomatous specimens 77 . Findings in the vascular and chondrocytes/osteoblast fields have demonstrated that elastin functions as the scaffold of mineralization. Further, MGP colocalizes with elastin in the arterial elastic lamina which is the first site of ectopic mineralization in the Mgp.KO 78 . It is also known that there is a correlation between calcification and degradation of elastic fibers 79 and that elastic fragmentation precedes vascular calcification 80,81 . A recent finding further showed that an Mgp.KO mouse with elastin haploinsufficiency (Mgp −/− ; Eln +/− ) exhibited reduced calcification 82 . It is intriguing that elastin degradation is also the hallmark of Marfan's syndrome, a connective tissue disease caused by a mutation in fibrillin1, a protein essential for the formation of the elastic fibers 83 . About 2% of Marfan's patients exhibit glaucoma 84,85 . Moreover, elastin degradation and fragmentation led to the release of elastin peptides, which colocalized with areas of microcalcification in the aorta of Marfan's patients 86 . Altogether, these data from other systems bring out the possibility that the absence of Mgp in the trabecular meshwork could induce elevated IOP by just inducing an alteration on its ECM's elastic network 82 . Elastin fragmentation followed by initiation of microcalcification could further contribute to ECM changes that could lead to increased resistance to aqueous humor outflow facility. These changes in the elastic network would affect the structure and integrity of the trabecular meshwork tissue which is so critical for the maintenance of physiological pressure.
In summary, in this study we have used a combination of strategies to generate an Mgp.floxed mouse (Mgpfloxed/floxed ). The use of a plasmid with a 3.8 kb donor DNA containing long homology arms, the 5′ and 3′ loxP sequences in cis, and its conversion to a single-stranded circular form before microinjection, together led to the success obtaining a floxed founder animal. The functional loxP recombination and ablation of the gene, fully characterized by viral gene delivery of the Cre-recombinase enzyme, opens the door to the use of this mouse to study Mgp roles in different tissues and diseases. Using this approach for the study of glaucoma, we report here the generation of the first Mgp.cKO in the trabecular meshwork, which is the tissue responsible for the regulation or intraocular pressure in the anterior segment of the eye. We find that the absence of Mgp protein in this tissue resulted in elevated IOP in the living animal. These findings show that the presence of Mgp is essential  www.nature.com/scientificreports/ to maintain the physiological IOP and protects the eye for developing elevated pressure. They have uncovered a new function for the Mgp gene. The mechanisms by which Mgp is responsible for maintaining eye pressure are yet to be investigated. Whatever those might be, Mgp appears as a good therapeutic target for the treatment of glaucoma.

Methods
Mice. All animal work was performed as approved by the Institutional Animal Care and Use Committee at the University of North Carolina at Chapel Hill (UNC) and conducted in accordance with the ARVO Statement on the Use of Animals in Ophthalmic and Vision research. All animals were housed in temperature-controlled rooms under standard 12-h cycle lighting with food and water provided ad libitum. The mouse strain used to obtain the embryos was C57BL/6J (B6) (Jackson Laboratory, Bar Harbor, ME, USA 000664). The mouse strain used to implant the blastocysts was CD-1 (Charles River Laboratory, Wilmington, MA, USA 022). The mouse strain used for experimental controls was also B6. For the generation of the plasmid donor vector (pMgp.Floxed) In-Fusion cloning was conducted with three DNA elements (5′ homology arm, donor fragment and 3′ homology arm) using an In-Fusion HD cloning kit (Takara 639649). The three DNA elements were amplified from genomic B6 DNA obtained from tail tips by the Hot Shot method 88 . The elements expand an Mgp region from 226 bp upstream of exon 2 to 1655 bp downstream of exon 4 (total 3802 bp including loxP sites and added restriction enzyme sites). They were amplified with primers containing 15 bp end sequences overlapping the EcoRV/BamH1 linearized pL453 vector and each other, plus the desired insertions of loxP and restriction enzymes sites sequences (Table 2). Amplifications were conducted using Q5 HD polymerase, 5 × reaction buffer and 5 × GC buffer (NEB M0491L), 0.4 µM corresponding primers and 20 ng of genomic DNA, at 98 °C 3 min, 95 °C 30 s, annealing temperature 30 s. (5′ element 58 °C, middle donor fragment 64 °C, 3′ element 70 °C), 72 °C 1 min, 35 cycles. DNA fragments were purified by gel electrophoresis. The 5′ element (100 ng), the middle donor fragment (100 ng), the 3′ element (100 ng), and the linearized and purified pL453 backbone (120 ng) were combined in one 10 µl cloning reaction with the In-Fusion HD Cloning Plus enzyme mix (Takara 638920) at 50 °C for 15 min. A control reaction lacking the DNA from the three elements was conducted in parallel. Transformation was conducted on Stellar cells (Takara) and colonies grown on ampicillin plates. Isolated colonies were amplified, and their plasmid DNA confirmed by sequencing. One sequenced confirmed plasmid, pMgp.Floxed was grown and purified using a maxiprep kit (Qiagen 12963).
For the preparation of the injection-grade lssc DNA, 50 µg of pMgp.Floxed DNA were incubated at 37 °C for 60 min with a newly designed gRNA containing matching vector sequences, and a D10A mutated singlestranded cutting Cas protein (hCas9_D10A was a gift from George Church (https ://n2t.net/addge ne:41816 ; RRID:Addgene_41816) to produce nicked-pMgp.Floxed. The gRNA and the mutated Cas protein were cloned and produced from the pT7RNA and pET-28a(+) plasmids as indicated above. Forty 40 µg of nicked pMgp. Floxed DNA were incubated with 20 µl (100 units/µl) Nuclease III (NEB M0206) for 30 min at 37 °C, purified by a quick spin column and dialyzed.
Embryo injection and mouse production. Mouse embryos were collected from the oviduct of naturally mated B6 females that were superovulated by injection with PMS (ProSpec, East Brunswick, NJ, USA) and HCG (Cosmo Bio USA, Carlsbad, CA, USA). A 1 µl mix containing purified Cas9 (400 nM), 5′ and 3′ gRNAs (gRNA g79 and gRNA 92, 50 ng each), and the lssc form of the pMgp.Floxed vector (30 ng) was prepared and from there, a few pl were injected into the embryo pronucleus with a micromanipulator. The injected embryos were cultured in KSOM media (CytoSpring, Mountain View, CAL USA) overnight and embryos that developed to the two-cell stage were transferred into the oviducts of pseudopregnant females.
Genotyping of neonatal mice (F0 and F1). Genomic DNA was extracted from the tail tip using the Hot Shot method indicated above 88 . For PCR, we used four primer pairs designed to target the inserted region externally and internally ( www.nature.com/scientificreports/ agarose/TAE gel. Negative controls included DNA from a parental non floxed mouse and a non template reaction. Positive control included the vector DNA. For the southern blot genotyping of the F1, tail tip genomic DNA was isolated using standard proteinase K extraction method (200 µl of 30 mM Tris/10 mM EDTA/1% SDS/12 µl of 10 µg/ml proteinase K followed by ethanol precipitation). Each pup DNA (10 µg) was digested with KpnI/BsrG1 overnight and run on a 0.7% agarose gel also overnight. After denaturation, the gel was blotted to a nylon membrane (Roche Life Science, Branford, CT, USA) by capillary action for 16 h. The blot was hybridized subsequently to three, DIG-labeled oligonucleotide probes in Hybing solution (DIG Easy Hyb, Roche) at 45 °C in rolling bottles for 16 h. The three oligonucleotide probes (Table 4), each specific for a region of the DNA, were obtained by PCR amplification in a total volume of 25 µl. Conditions were 95 °C 2 min, (94 °C 30 s, 72 °C 30 s (− 1 °C per cycle), 72 °C 1 min) for 14 cycles, then (95 °C 30 s, 58 °C 30 s, 72 °C 1 min) for 24 cycles, and ending at 72 °C for 2 min using the PCR DIG probe synthesis kit (Roche 11636090910). After hybridization, blots were sequentially washed with 2 × SSC at room temperature and 0.5 × SSC at 65 °C, exposed to X-ray film (Genesee Scientific, El Cajon, CAL USA). For stripping in between hybridizations, blots were incubated 2 × in stripping buffer (0.2 M NaOH, 0.1% SDS) at 37 °C for 15 min each. DNA Molecular Marker II Dig labelled (Roche) and 1 kb Plus DNA Ladder (GeneRuler, ThermoFisher, Waltham, MA, USA) were used to determine the sizes of the hybridized fragments.

Primary culture of mouse iridocorneal angle cells (MIA cells). Mice 2-4 months old were eutha-
nized by an overdose intraperitoneal (IP) injectable anesthesia (400 mg/kg ketamine/20 mg/kg xylazine/4 mg/ kg acepromazine) (Covetrus, Dublin, OH, USA), followed by cervical dislocation immediately prior to tissue collection. Whole globes were enucleated, cleaned and washed with PBS. Under a dissecting microscope, globes were then bisected a few mm posterior to the limbus using an Optical microsurgery blade (Wilson Ophthalmic, Mustang, OK) and iridectomy scissors. After removing the lens, separated anterior segments were cut into four quadrants and iris and ciliary body carefully removed with forceps. Strips of the angle region containing the trabecular meshwork were obtained by making anterior and posterior incisions of the iridocorneal region and placed them on 2% porcine gelatin-coated (Sigma) 35 mm dishes. The tissue was coverslipped with a drop of MEM Richter's Modification medium (IMEM, HyClone/ThermoFisher) supplemented with 20% fetal bovine serum (FBS, Gibco/ThermoFisher), 50 µg/ml gentamicin (Gibco/ThermoFisher) and cells allowed to grow for 3-4 weeks changing the media every other day. Upon confluency, cells from the right and left eyes of the same mouse were trypsinized, pooled, passed to 60 mm dishes to confluency, and either harvested and stored in liquid nitrogen, or used directly for the experiment. In total, the MIA cells used in this study originated from six Mgp.floxed mice (6 cell lines, MIA-F1, 2, 4, 5, 6 & 7 from 6 different breeding pairs) and from two B6 mice (MIA-B61 and MIA-B62, from 2 different breeding pairs). All cells were used at passages 1-2.

Adenoviral vectors and infection of MIA cells. Recombinant adenoviruses Ad-GFP-2A-iCre and
control Adeno.GFP were obtained commercially (Vectors Biolabs, Malvern, PA, USA, and Qbiogen, Carlsbad, Canada, respectively), grown and purified in our laboratory 58 . The Ad-GFP-2A-iCre (termed Ad.GFP.Cre in the manuscript) expresses both a codon improved Cre-recombinase (iCre) an eGFP marker. Cre and GFP are driven by the same CMV promoter and separated by 2A peptides. The Adeno.GFP (termed Ad.GFP in the manuscript) is an adenovirus 5 carrying a variant of the jellyfish Aequorea vitoria GFP driven by the CMV promoter 58 . Physical particles were tittered as viral genomes (vg)/ml by extracting an aliquot of purified viral DNA (DNeasy Blood and Tissue kit, Qiagen 69054), measuring its optical density (Nanodrop One, ThermoFisher) and converting each ng of DNA to the number of DNA molecules based on its MW. Viral infectivity (plaque-forming units per ml, pfu/ml) was measured with a QuickTiter Adenovirus Titer Immunoassay kit (Cell Biolabs, San Diego, CA, USA VPK-109) following manufacture recommendations. Viral lots used in this study had concentrations of 5-6 × 10 11 vg/ml (0.2-2 × 10 11 pfu/ml) (Ad.GFP.Cre) and 5 × 10 11 vg/ml (3.9 × 10 10 pfu/ml) (Ad.GFP) respectively.
MIA primary cells at passage 1 to 2 seeded on six-well dishes were grown to 70% to 80% confluency, washed twice with PBS, and exposed to the recombinant adenoviruses in 1 ml serum-free medium. After exposure to the virus for 2 h, complete media was added, and incubation continued for 2 days. Fluorescence images were captured on living cells with an inverted IX71 Olympus fluorescence microscope equipped with a DP80 monochrome camera and cellSense software (Olympus, Center Valley, PA, USA).
Intracameral microinjection of recombinant viral vectors. The microinjection system for nanoliteraccurate delivery to the intracameral space of the mouse consists of a 10 µl specially designed glass syringe (NanoFil syringe, World Precision Instruments, WPI, Sarasota, FL, USA) mounted on an UltraMicroPump (UMP3, WPI) which is connected to a small controller box (Micro2T SMARTouch, WPI) to program delivery conditions. The NanoFil syringe is attached to a quartz flexible tubing (SilFlex) (35 cm long, 100 µm ID, 460 µm OD), reinforced at both ends with teflon jackets. One end of the Silflex tubing was connected to the NanoFil syringe, while the other end was connected to a 1 cm PE-10 intramedic tubing piece (0.38 mm ID, Clay Adams, Parsippany, NJ, USA) which in turn was tightly connected to the 460 µm OD shank of the 33G NanoFil needle (total 40 mm long). This modification allowed to hold the NanoFil needle with a needle holder (Barraquer, Storz Ophthalmic Instruments, Rochester, NY, USA) and to more accurately drive it to the injection site. www.nature.com/scientificreports/ SMARTouch controller box was programmed to deliver 2 µl in 30 s (67 nl/s). The overall system is depicted in Fig. 3S. Each mouse was anesthetized by an IP injectable anesthesia (50 mg/kg ketamine/5 mg/kg xylazine/1 mg/ kg acepromazine) (Covetrus). Whiskers were trimmed and pupils dilated with a drop of 1% tropicamide ophthalmic solution (Covetrus). While resting slightly on its side with its tail to the right, the mouse was placed under a surgical stereo microscope (Leica M80) equipped with a DFC450 digital camera (Leica Microsystems, Buffalo Grove IL, USA). The NanoFil syringe was filled with the different viral treatments and the mouse eye was secured nasal to temporal with fine Straight Bishop-Harmon Tissue forceps (Storz E1500). These forceps, once closed at its front, leave a space between the shafts that helps protect the optic nerve from crushing when holding the eye during the injection. The NanoFil needle was inserted through the cornea a few mm from the retracted iris by holding the needle with the needle holder. To facilitate entrance, the cornea was superficially pricked with a 27G needle prior to the NanoFil needle insertion. When the NanoFil needle was inside the anterior chamber, the UMP3 micropump was turned on by the Micro2T box controller, and fluid entry monitored by direct visualization through the operating microscope. After delivery, the NanoFil needle was left in place for 30 s. and withdrawn gradually to minimize leaking. Topical antibiotic ointment (neomycin 3.5 mg/g, polymyxin B 10,000 U/g, and bacitracin 400 U/g) (Covetrus) was applied to the eyes, and animals returned to their cages, resting on heating pads for recovery. TaqMan Relative Quantification values between treated and untreated samples were calculated by the formula 2 − C T where C T is the cycle at threshold, ΔC T is C T of the assayed gene minus C T of the endogenous control (18S), and ΔΔC T is the ΔC T of the normalized assayed gene in the treated sample minus the ΔC T of the same gene in the untreated one (calibrator). Because of the high abundance of the 18S rRNA used as the endogenous control and in order to get a linear amplification, RT reactions from treated and untreated samples were diluted 10 4 times prior to their hybridization to the 18S TaqMan probe. Statistical analysis was performed by the Student's t-test.

DNA extraction and PCR analysis of the validation experiments.
Genomic DNA from MIA cells was purified using a DNeasy Blood and Tissue kit (Qiagen 69054) from 35 mm dishes. Harvesting of the cells was conducted by scraping the PBS-washed 35 mm dish with a cell lifter and centrifugation at 300×g for 5 min. The pellet was resuspended in 200 µl PBS, 200 µl of AL buffer, 20 µl proteinase K and extraction continued following manufacturer's recommendations. Final columns were eluted with 50 µl of HyPure water (HyClone/GE Healthcare Life Science, Logan, UT, USA). For the tissue, after PBS washing, iridocorneal strips from single eyes were disrupted by vortexing with 180 µl of the kit's ATL plus 20 µl proteinase K. The homogenate was transferred to an eppendorf tube and extraction continued as described for the cells. In addition to iridocorneal strips, the iris and cornea tissues were also dissected and pooled each from 3 eyes. Cornea samples were homogenized in a glass microtissue grinder (Kimble-Kontes, Vineland, NJ, USA) with 180 µl of ATL plus 20 µl proteinase K prior to transferring to the eppendorf tube.

Protein extraction and western blot analysis.
After the removal of the medium, treated and untreated primary MIA cells in 3 cc wells were washed twice with cold PBS, and harvested in 140 µl cold lysis buffer (100 µl RIPA buffer plus 40 µl of 1 × protease inhibitor) (Sigma-Aldrich and Roche/Sigma respectively). Lysed cells were centrifuged cold at 14,000×g for 10 min and the supernatant disrupted with a sonicator (Microson Ultrasonic XL 2000; Misonix, Farmingdale, NY, USA) equipped with a 2.4 mm microprobe (Misonix) at setting 3 for five pulses. The sonicate (soluble fraction) was collected and stored at − 80 °C until use. For the tissue, dissected iridocorneal angle strips, containing the trabecular meshwork, were rinsed in PBS, pooled from 3 eyes and placed Measurement of intraocular pressures (IOP). IOPs were measured unmasked. Mice were lightly anesthetized (42 mg/kg ketamine, 4 mg/kg xylazine and 0.8 mg/kg acepromazine) by IP injection and with an eye drop of 0.5% tetracaine (Covetrus). Measurements were obtained with a calibrated TonoLab selected for mouse settings (Colonial Medical Supply, Franconia, NH) and equipped with a foot pedal. At this anesthesia concentration, mice achieve recumbency in 3 to 5 min and all IOP measurements were taken at 3 min after recumbency. To take IOP measurements, whiskers were trimmed, and mice positioned with the visual axis horizontal to the TonoLab probe which was held at a distance 2-5 mm from the center of the cornea. IOPs were obtained as the average of 6 consecutive measurements by pressing the foot pedal. Only mean values with a standard deviation (expressed as percentage of the mean) less than 5% were accepted. The average of at least 3 such readings (18 measurements) was considered to be the absolute IOP for the given point. The Integral IOP (cumulative pressure received by each mouse during the entire duration of the experiment) was calculated using the Area Under the Curve (AUC) tool of the GraphPad Prism 5 software (GraphPad Software, Inc., La Jolla CA). Data were analyzed using the SigmaPlot software (Systat Software Inc., San Jose, CA, USA) and are presented as means ± SEM. All IOP measurements were taken between 11:30 am to 1:00 pm. Under these conditions, baseline values of the B6 strain were 8.0 ± 0.10 mmHg (n = 22 eyes).