CRISPR-Cas9 mediated generation of a conditional poly(A) binding protein nuclear 1 (Pabpn1) mouse model reveals an essential role for hematopoietic stem cells

Poly(A) binding protein nuclear 1 (PABPN1) is known for its role in poly(A) tail addition and regulation of poly(A) tail length. In addition, it has been shown to be involved in alternative polyadenylation (APA). APA is a process regulating differential selection of polyadenylation sites, thereby influencing protein isoform expression and 3ʹ-UTR make-up. In this study, we generated an inducible Pabpn1flox/flox mouse model using crRNA-tracrRNA:Cas9 complexes targeting upstream and downstream genomic regions, respectively, in combination with a long single-stranded DNA (ssDNA) template. We performed extensive in vitro testing of various guide RNAs (gRNAs) to optimize recombination efficiency for in vivo application. Pabpn1flox/flox mice were generated and crossed to MxCre mice for validation experiments, allowing the induction of Cre expression in the bone marrow (BM) by poly(I:C) (pIC) injections. Validation experiments revealed successful deletion of Pabpn1 and absence of PABPN1 protein. Functionally, knockout (KO) of Pabpn1 led to a rapid and robust depletion of hematopoietic stem and progenitor cells (HSPCs) as well as myeloid cells, suggesting an essential role of Pabpn1 in the hematopoietic lineage. Overall, the mouse model allows an inducible in-depth in vivo analysis of the role of PABPN1 and APA regulation in different tissues and disease settings.

www.nature.com/scientificreports/ In the past, different transgenic mouse models were developed expressing mutated Pabpn1, which can be used to study the pathology of OPMD in mice 13,14 . While these mouse models allow analysis of the role of mutated PABPN1, their use to study the role of PABPN1 in APA in vivo remains limited. Several studies could show that not only mutation of Pabpn1, but also changes in PABPN1 protein levels lead to robust deregulation in APA 10,12 . In order to study PABPN1 and APA in vivo in different tissues and cell populations, an inducible PABPN1 knockout (KO) mouse model is needed. Vest et al. reported such a mouse model in 2017 and phenotyped heterozygous PABPN1 KO mice 14 . In this mouse model the integrated loxP sites flank exon 1 and 2, but not the Pabpn1 promoter. However, several Pabpn1 isoforms do not include exon 1 and 2 and thus this generated allele may not be considered a null. We therefore aimed to develop an inducible Pabpn1 KO (null) mouse model that allows deletion of all expressed Pabpn1 isoforms and introduced loxP sites flanking exon 3 and 4. The generation of transgenic mice was successful and the targeting strategy was verified by a proof-of-principle study demonstrating the deletion of Pabpn1 in hematopoietic cells after expression of Cre under the control of the Mx1 promoter.

Results
gRNA design and testing. The primary aim of this study was the generation and validation of a floxed Pabpn1 transgenic mouse model that can be crossed to any Cre driver line to induce tissue-specific or inducible hetero-or homozygous KO of Pabpn1. To ensure loss of all protein-encoding isoforms, Pabpn1 isoforms and functional domains were analyzed using Ensembl (www. ensem bl. org). Several isoforms are reported, and all of them share expression of exon 3 and 4 of the Pabpn1 isoform ENSMUST00000022808. 13. In addition, the RNA binding domains are encoded in these exons. Therefore, we aimed to introduce loxP sites flanking exon 3 and 4 (Fig. 1A). For generation of transgenic mice, a CRISPR/Cas9-based approach was used 15,16 . 3 gRNAs per targeting locus were designed (upstream of exon 3: gRNAs A1-3; downstream of exon 4: gRNAs B1-3) using CRISPOR (http:// crisp or. tefor. net/) 17 (Fig. 1A). The targeting sequences were located at least 250 bp up-or down-stream of the respective exon to exclude interference of the introduced loxP sites with splicing. To ensure efficient recombination in vivo, we performed pre-testing of gRNA efficiency in vitro.
Towards this end, we made use of a disrupted lacZ reporter gene harboring the respective targeting sequence (gRNAs A1-3 or gRNAs B1-3; Fig. 1B). Targeting sequences correspond to the Pabpn1 introns targeted by the respective gRNAs and were inserted between the lacZ fragments in the pCRISPR report vector. gRNA oligonucleotides were annealed and cloned into the px330 Cas9 vector, enabling gRNA/Cas9 expression. To test gRNA efficiency, HeLa cells were transfected with the px330 Cas9 vector encoding for gRNA A1-3 or B1-3. Co-transfection with the respective pCRISPR reporter plasmid and a luciferase-encoding control plasmid (pUHC131.1) was performed. 24 h later, cells were lysed and LacZ as well as luciferase activity were measured. The normalized LacZ signal of the respective tested gRNA is a direct indicator of gRNA targeting efficiency. Testing revealed the highest efficiency for gRNAs A3 and B2 (Fig. 1B).
Generation and genotyping of Pabpn1 flox/flox mice. We then made use of the Easi-CRISPR method described by Miura et al. and Quadros et al. 15,16 . Briefly, a long ssDNA donor template is injected together with pre-assembled CRISPR RNA-trans-activating RNA-CRISPR-associated protein 9 (crRNA-tracrRNA:Cas9) ribonucleoprotein complexes into mouse zygotes (Fig. 1C). The crRNA molecule directs the complex to the target site by complementary base-pairing, while the tracrRNA part is bound by Cas9. After binding to the target site, Cas9 cleaves the DNA, generating a DNA double-strand break (DSB). By providing template DNA strands, in this case a ssDNA donor template, the Cas9-induced DSB can be used for homology-directed repair (HDR). Excision and recombination take place in the 1-cell or early 2-cell state. crRNAs corresponding to gRNA sequences A3 and B2 and a long ssDNA megamer were ordered from IDT. In the 1126 nucleotide (nt) long megamer, a 60 nt homology arm was followed by the upper loxP site, exon 3 and 4 including intronic sequences, the lower loxP site and a 60 nt downstream homology arm. crRNA-tracrRNA:Cas9 complexes and the ssDNA template were injected into one of the pronuclei of C57Bl6/J zygotes, which were subsequently transplanted into pseudo-pregnant foster mice.
From two rounds of injections, 17 mice were born (Fig. 1D). In order to identify potential founder mice, we developed a stringent genotyping strategy. In a first step, integration of the upstream (us) and the downstream (ds) loxP site was evaluated by PCR1 and PCR2, respectively. Successful integration of loxP sites leads to a band  To exclude the presence of point mutations, we performed amplicon sequencing of PCR2 and 4. In total, 3 animals (IDs: 5, 11, 15) passed this final QC (Fig. 1D). The mouse with ID5 was crossed to C57Bl6/J wildtype animals to generate heterozygous offspring. These mice were again extensively genotyped, and loxP integration was verified (data not shown). Heterozygous mice were crossed, generating homozygous Pabpn1 flox/flox mice.  2E). In addition, reduced frequencies of myeloid cells in the BM were observed, leading to a relative increase in B cell frequencies (Fig. 2F). Taken together, we successfully validate the function of our Pabpn1 flox/flox mouse model.  www.nature.com/scientificreports/ Characterization of MxCre Pabpn1 flox/flox mice. In order to better and more extensively characterize the effects of homozygous Pabpn1 KO in the hematopoietic system, we used a milder pIC treatment protocol, with 3 instead of 5 injections, and performed analysis already at day 10 (Fig. 3A). This treatment regime did not lead to complete loss of LSK and LS-K cells as observed with the previous treatment approach (Fig. 2) and thus allowed us to perform a more in-depth analysis of the hematopoietic phenotype. MxCre (Ctrl), MxCre Pabpn1 flox/flox mice (KO) and Pabpn1 flox/flox mice, not carrying the MxCre allele (fl/fl) and serving as an additional control, were injected with pIC and analyzed. Hemavet analysis revealed reduced numbers of mature blood cells, with the strongest effect observed in the myeloid lineage and platelet counts in KO animals (Fig. 3B, Supplementary Fig. 3A). This reduction in mature blood cells was accompanied by a strongly reduced total BM cellularity (Fig. 3C). Further analysis of the blood, spleen and BM confirmed significantly reduced numbers of myeloid cells (Fig. 3D-F). Interestingly, we also observed reduced numbers of pre-pro B cells and immature B cells, while numbers of mature B cells were not affected (Fig. 3F). Analysis of T cell maturation in the thymus revealed a decrease in the immature double negative (DN) population ( Supplementary Fig. 3B). In depth characterization of BM of Pabpn1 KO mice revealed reduced frequency of more committed progenitors contained in the Lin− Sca-1− c-Kit+ (LS-K) compartment (Fig. 3G). We observed reduced numbers of common myeloid progenitors (CMP) and granulocyte-macrophage progenitors (GMP), while megakaryocyte-erythrocyte progenitors (MEP) and common lymphoid progenitors (CLP) were unchanged. In addition, analysis of erythroid maturation revealed an increase in the more mature erythroblast populations and a decrease in maturing erythroblasts ( Supplementary Fig. 3C). The HSPC containing LSK compartment was massively reduced in KO animals, all HSC and MPP populations except MPP2 were depleted (Fig. 3H). Overall, we observed strong effects on the hematopoietic system following depletion of Pabpn1. Especially mature myeloid cells, myeloid progenitors and HSPCs were particularly dependent on Pabpn1 at the analyzed timepoint.

Discussion
Using the Easi-CRISPR technique 15,16 , we successfully generated an inducible Pabpn1 KO mouse model, with loxP sites flanking exon 3 and 4. We achieved a targeting efficiency of 18% (3/17 animals). In our hands, the Easi-CRIPSR strategy was a quick and reliable approach to generate floxed transgenic animals. A common issue with CRISPR-Cas9 targeting approaches are off-target effects 19 . gRNAs were chosen based on their reported off-target score. gRNA A3 was predicted to have 90 potential off-targets with at least 3 mismatches and gRNA B2 presented 100 anticipated off-targets with at least 2 mismatches. A high number of mismatches decreases the likelihood of off-target mutations. Because of the absence of homologous regions, an off-target event would lead to a mutation, but not to loxP site integration. All of the predicted off-targets are localized on other chromosomes than Pabpn1 leading to loss of mutation by further crossing of the generated mice. We therefore did not systematically screen for off-target events. Overall, we minimized the likelihood of off-target events by gRNA choice and backcrossing of mice.
In contrast to the transgenic inducible Pabpn1 We functionally validated our mouse model by crossing Pabpn1 flox mice to the MxCre line followed by Cre induction using pIC, leading to efficient deletion of the floxed allele in hematopoietic cells. Recombination was observed on the DNA level, and reduced levels of PABPN1 protein in total BM cells was confirmed by Western blot and intracellular flow cytometry analysis. Unfortunately, analysis of sorted cell populations was hindered by the strong phenotype observed in hematopoietic cells leading to a loss of the respective cell populations. The residual level of PABPN1 protein still detected in total KO BM cells is presumably due to the presence of nonhematopoietic cell types not expressing Cre.
As a proof-of-concept experiment to characterize Pabpn1 KO mice, we performed extensive profiling of BM cells. This revealed depletion of HSPCs and reduced frequencies of myeloid progenitors and mature myeloid cells in the blood, thymus, BM and spleen. In addition, B and T cell maturation was impaired and reduced numbers of non-mature B and T cells were observed. Overall, short-lived mature blood cells and progenitors seem to be more strongly affected by loss of Pabpn1. Interestingly, 10 days after the first pIC injection also the HSPC compartment was massively depleted. This compartment usually contains highly quiescent long-lived cell subsets 20 . The HSPC compartment could directly be affected by loss of Pabpn1 or loss of mature cells could lead to activation of this compartment as observed for other genes 21 . This activation in the absence of Pabpn1 could cause cell death of HSPCs. Overall, our findings are in line with our previous HSPC Pabpn1 KD studies, in which we observed impaired HSC function 12 . In the future, additional studies need to be conducted to better understand the kinetics and the mechanism of the observed phenotype.
Overall, we describe the successful generation of an inducible Pabpn1 KO mouse model. In the future, this mouse model can be used to facilitate functional in vivo research on the APA regulator PABPN1 using different tissue-specific Cre drivers and disease models.    Table 1, Sigma-Aldrich; design gRNA A: http:// crisp or. tefor. net/ crisp or. py? batch Id= aQ9aK iq6R1 spZW5 ypsBG; design gRNA B: http:// crisp or. tefor. net/ crisp or. py? batch Id= idFOo 1Vhuw IGjPT eLMe3) in the px330 Cas9 vector (42230, Addgene). The respective oligos were annealed for 2 h (starting at 95 °C continuously cooled to RT). The px330 Cas9 vector was digested with BbsI enzyme for 2 h at 37 °C. Annealed oligos and the digested px330 Cas9 vector were purified by agarose gel electrophoresis (2% for oligos; 1% for vector). Gel extraction using the QIAEX II Gel Extraction Kit (Qiagen) according to manufacturer's guidelines (elution in 20 µl) was performed. 100 ng digested vector was ligated with 1.2 ng annealed oligos for 1.5 h at 22 °C using T4 ligase (New England BioLabs Inc.). Bacterial (E. coli) transformation was performed with the ligated px330 Cas9 vectors. Transformed bacteria were plated on agar plates (containing 100 µg/ml ampicillin) and incubated overnight at 37 °C. Colonies were picked and colony PCR (Supplementary Table 2) was performed (Supplementary Table 3). Successful integration of target sequence oligos in the px330 Cas9 vector leads to a product of approximately 500 bp. Presence of this product was verified by agarose gel electrophoresis (1%). Positive clones were used for overnight culture and subsequent DNA plasmid Mini-prep using the NucleoSpin Plasmid Kit (Macherey Nagel) according to manufacturer's guidelines was performed. Enzymatic digestion was used to test for oligo integration using the enzymes BbsI and EcoRI (both New England BioLabs Inc.). Re-ligation of px330 Cas9 vector without oligo integration leads to 2 products after the digestion step (3.2 kb and 5.3 kb). Successful oligo integration leads to abrogation of the BbsI restriction enzyme sites and results in the generation of only one 8.5 kb product. Digestion was performed for 2 h at 37 °C. Product size was analyzed by agarose gel electrophoresis (1%).

Methods
Cloning of the pCRISPR-report vector. In addition, oligos encoding for target sequences (Supplementary Table 1, Sigma-Aldrich) were ordered for cloning into the pCRISPR-Report vector (pTAL-Rep) 17 for gRNA testing. Oligos were annealed and purified by gel extraction as previously described (see "Cloning of the px330 Cas9 vector" section). The pCRISPR-Report vector was digested using BstBI and NruI and purified by gel extraction as previously described (see "Cloning of the px330 Cas9 vector" section). 100 ng digested vector was ligated with 1.2 ng annealed oligos for 1.5 h at 22 °C using T4 ligase (New England BioLabs). Bacterial (E. coli) transformation was performed using the ligated pCRISPR-Report vectors encoding for lacZ fragments separated by targeting sequences that are equivalent to the mouse Pabpn1 targeting sequences for upstream gRNAs A and downstream gRNAs B. Transformed bacteria were plated on agar plates (containing 100 µg/ml ampicillin) and incubated overnight at 37 °C. Colonies were picked and colony PCR was performed (Supplementary Tables 3,  4). Successful integration of oligos in the pCRISPR-Report vector leads to a product of approximately 900 bp length and was verified by agarose gel electrophoresis (1%). Positive clones were used for overnight cultures and subsequent DNA plasmid Mini-prep using the NucleoSpin Plasmid Kit (Macherey Nagel) according to manufacturer's protocols was performed. Enzymatic digestion was used to test for oligo integration using the enzymes SnaBI (pCRISPR-Report for testing of A gRNAs) or EcoRI (pCRISPR-Report for testing of B gRNAs) (all New England BioLabs). Re-ligation of pCRISPR-Report vector without oligo integration leads to 1 product after the digestion step (7.4 kb). Successful oligo integration leads to the presence of the respective restriction enzyme sites and the generation of two products (A: 0.8 kb + 6.6 kb; B: 1.2 kb + 6.2 kb). Digestion was performed for 2 h at 37 °C. Product size was analyzed by agarose gel electrophoresis (1%). Positive clones were selected and overnight culture and subsequent DNA plasmid Midi-prep using the Macherey Nagel NucleoBond Xtra Midi Plasmid Kit according to manufacturer's protocols was performed.  www.nature.com/scientificreports/ Reporter assay. 25,000 HeLa cells/well were seeded in 12-well plates in DMEM GlutaMAX (+ 10% FCS, + 1% Pen/Strep). Cells were incubated for 2-3 h. Transfection mix was set up (Supplementary Table 5) and mixed with 100 µl RotiFect mix (5 µl RotiFect (Carl Roth) in 100 µl Opti-MEM I) and incubated for 30 min at RT. 200 µl transfection mix were added per well, and cells were incubated overnight at 37 °C. Cells were washed with PBS and lysed using 1× passive lysis buffer for 3 min at RT. Lysates were centrifuged at 13,000 rpm for 5 min at 4 °C. 30 µl supernatant per replicate and condition were transferred to a fresh Eppendorf tube, and 700 µl Z-buffer (1.60 g Na2HPO4.7H2O, 0.55 g NaH2PO4.H2O, 0.075 g KCl, 0.012 g MgSO4, pH 7.0, in 100 ml dH2O, 0.35% 2-Mercaptoethanol) were added and mixed. Subsequently, 200 µl ONPG solution (0.55 g ONPG in 100 ml dH2O) were added per tube and incubated at RT until a yellow color-shift was observed. Reaction was stopped using 500 µl 1 M NaOH per tube, and the signal was measured (420 nm). For luciferase measurements, 10 µl of lysate per replicate and condition were pipetted into a white 96-well plate. This plate was measured in a luciferase detection instrument (Perkin Elmer Wallac 1420 Victor2 Microplate Reader). To calculate normalized signals, the following equation was used: LacZ/Luciferase*mean(Luciferase).