Rapid target validation in a Cas9-inducible hiPSC derived kidney model

Recent advances in induced pluripotent stem cells (iPSCs), genome editing technologies and 3D organoid model systems highlight opportunities to develop new in vitro human disease models to serve drug discovery programs. An ideal disease model would accurately recapitulate the relevant disease phenotype and provide a scalable platform for drug and genetic screening studies. Kidney organoids offer a high cellular complexity that may provide greater insights than conventional single-cell type cell culture models. However, genetic manipulation of the kidney organoids requires prior generation of genetically modified clonal lines, which is a time and labor consuming procedure. Here, we present a methodology for direct differentiation of the CRISPR-targeted cell pools, using a doxycycline-inducible Cas9 expressing hiPSC line for high efficiency editing to eliminate the laborious clonal line generation steps. We demonstrate the versatile use of genetically engineered kidney organoids by targeting the autosomal dominant polycystic kidney disease (ADPKD) genes: PKD1 and PKD2. Direct differentiation of the respective knockout pool populations into kidney organoids resulted in the formation of cyst-like structures in the tubular compartment. Our findings demonstrated that we can achieve > 80% editing efficiency in the iPSC pool population which resulted in a reliable 3D organoid model of ADPKD. The described methodology may provide a platform for rapid target validation in the context of disease modeling.

www.nature.com/scientificreports/ identified gene is likely to ameliorate disease phenotype in patients 8,9 . The demand for improved TV platforms grows as the number of identified gene-disease associations increases. Ideally, a TV platform (e.g., an in vitro cellular assay) should functionally validate the mutations in their genomic, cellular and human-specific contexts at high speed and capacity to enable high-throughput programs. Given the cellular complexity of the kidney, the lack of disease models that can be used as reliable TV platforms have remained a challenge for decades. However, advances in direct differentiation of human induced pluripotent stem cells (hiPSCs) into three-dimensional (3D) organoids and RNA-guided gene-targeting technologies hold great promises for overcoming these challenges. An organoid is a complex in vitro 3D multicellular tissue derived from different cell types which resembles the structure and the functionality of the real organ in humans 10 . Remarkable progress has been achieved during the last decade in shifting towards the use of organoids for modeling human diseases 11,12 . Organoid models for many tissues e.g., brain, gastrointestinal tract, breast and liver have now been reported. Likewise, several protocols have been described in the last five years for generation of kidney organoids [13][14][15][16][17] . Notably, our team has previously reported on developing kidney organoids for drug discovery applications 15 . Along with the progress in generation of human 3D models, a revolutionary gene targeting technology, the CRISPR-Cas9 system, has rapidly been established for genome manipulation. CRISPR (Clustered, Regularly Interspaced, Short Palindromic Repeats)-Cas9 system is originally harnessed from bacterial adaptive immune system 18 for applications in human cells 19,20 . This technology encompasses an endonuclease protein (Cas9) and an RNA molecule, designated as short guide-RNA (sgRNA) guiding Cas9 to the specific target locus in the genome. Different genomic sites can be targeted by providing different sgRNA molecules that match the targeted sites. The Cas9-induced DNA breaks often result in the formation of small insertion or deletion (indels) that can subsequently cause frameshifts within the coding sequence, thereby yielding truncated or non-functional proteins. CRISPR-Cas9 unequivocally provides scientists with a tool to study gene functions by comparing wild-type and edited phenotypes.
Applying both 3D kidney differentiation and CRISPR technologies, is opening new opportunities for researchers to study the genetic basis of kidney disorders with an unprecedented translatability. However, notable challenges still exist. One hurdle to increase the throughput of gene knockout and phenotypic studies in kidney organoids is the need for the generation of clonal knockout cell lines followed by their differentiation into organoids. This process is time consuming and labor-intensive, requiring several months for validation of the knockout line. Hence, more work is needed in exploring different routes to improve this lengthy process. Here, we report the successful use of high efficiency knockout pool populations for studying kidney disease phenotypes. This allowed us to skip the generation of clonal cell lines and establish a speedy target validation platform.

Results and discussion
In this study, we selected a doxycycline-inducible Cas9 expressing hiPSC line 21 to conduct gene-editing and differentiation experiments. The reason behind selecting the Cas9-inducible cell line in our study was to try and achieve as high a rate of editing as possible, as each cell would express Cas9 upon doxycycline induction and so only guides would need to be introduced to the cells to enable efficient editing, simplifying the transfection step. This cell line contains a Cas9 expressing cassette integrated into the AAVS1 locus using the ObLiGaRe system 21,22 .
For organoid development, we optimized a previously published kidney organoid differentiation protocol 15 on this cell line to validate it's competency for differentiation to kidney organoids. This protocol consisted of two stages. In the first stage cells were cultured in conventional plates for 10 days; and media were supplemented with CHIR99021 and FGF9 to initiate differentiation. This resulted in the generation of kidney progenitor cells. In the second stage, cells were transferred into aggrewell plates to allow formation of organoids. Finally, organoids were transferred into suspension plates for further maturation until day 31 when they were harvested for phenotyping studies. A schematic presentation of our kidney organoid protocol is presented in Fig. 1A.
Next, we used Cas9-mediated gene editing to introduce mutations in PKD1 and PKD2. These two genes were selected because mutations in these genes cause ADPKD and can therefore serve as positive controls (i.e., expecting the disease phenotype) for evaluation of our methodology. ADPKD is an inherited monogenic renal disease characterized by the accumulation of clusters of fluid-filled cysts in the kidneys. Previous studies have confirmed that knockouts of PKD1 or PKD2 result in formation of a cyst-like phenotype in kidney organoids. For example, Freedman and colleagues reported an in vitro model for polycystic kidney disease (PKD) by introducing biallelic mutations in PKD1 or PKD2 using CRISPR-Cas9 and generating kidney organoids from a clonal knockout iPSC line. Those knockout organoids formed cyst-like structures in kidney tubules, recapitulating some characteristic features of the disease phenotype in vitro 23,24 . In our study, we evaluated the activity of the previously described PKD1 and PKD2 sgRNAs (PKD1-sgRNA1 and PKD2-sgRNA4 respectively), in addition to new sgRNAs designed in this study for PKD1 (PKD1-sgRNA2 and PKD2-sgRNA3) and PKD2 (PKD1-sgRNA5 and PKD2-sgRNA6) ( Fig. 2A). In brief, to test the activity of the sgRNAs, we first transfected the sgRNA-expressing plasmids into the Cas9-induced cells and performed deep sequencing on the targeted loci at 72 h post transfection. All sgRNAs showed activity in all three different human cell lines (HEK293T, A549, hiPSCs) tested in our study (Fig. 2B). We could detect mutations introduced at the targeted sites using deep targeted sequencing, with efficiencies ranging from approximately 15-75% and of these mutated sequences, around 70-80% of these mutations caused frameshifts (Fig. 2C). www.nature.com/scientificreports/ The editing efficiency of the tested sgRNAs were higher in HEK293T and A549 cell lines than in hiPSCs, presumably due to the different transfection efficiencies. However, the normalized percentage of the frameshifting mutations were consistent among different cell types regardless of the variations in the achieved editing efficiencies (Fig. 2C). In hiPSCs, the highest editing efficiencies were about 50% and 30% for PKD1 and PKD2 genes, respectively. These numbers represent the number of mutated alleles in the pool of cells and do not represent the percentages of the cells that contain mutations on both alleles. To achieve a higher genome-editing efficiency, we rationalized that paired sgRNAs would result in a higher indel formation than single guides alone. This approach is based on the proximity of two sgRNAs introducing two simultaneous DNA breaks. It has been previously shown that the repair at these DNA breaks is mediated by Non-Homologous End Joining (NHEJ) repair pathway and results in the precise deletion of the DNA sequence between the two cut sites 25 . Additionaly, we tested synthetic-sgRNA transfection into the Cas9-induced hiPSCs instead of plasmid DNA transfection (Fig. 3A). Introducing these changes resulted in achieving a remarkable > 80% editing efficiencies in both PKD1 and PKD2 using deep targeted sequencing (Fig. 3B). Next, we examined the off-target sites for all tested sgRNApairs by performing deep targeted sequencing. Prospective off-target sites were identified using an in-house developed software identifying sites with up to a maximum of four mismatches. Our results showed no Cas9 off-target activity at any of these evaluated loci (Fig. 3C). Investigating Cas9 off-target effects is an important aspect of performing gene editing experiments as previous reports have shown that some off-target sites can be edited at a similar efficiency as the on-target site 26 .
We further examined the PKD knockout pools (PKD KOp) for pluripotency and differentiation capability after CRISPR-Cas9 gene editing. PKD KOp iPSCs expressed expected pluripotency markers such as octamerbinding transcription factor 4 (OCT4) and NANOG ( Supplementary Fig. S1a). The tri-lineage differentiation potential was assessed using immunofluorescence analysis of markers for ectoderm (PAX6 and Nestin) (Supplementary Fig. S1b), endoderm (SOX17 and CXCR4) ( Supplementary Fig. S1c) and mesoderm (Brachyruy and CXCR4) ( Supplementary Fig. S1d) along with DAPI for nucleus staining in 2-dimensional cultures. As expected, our results showed that PKD1 and PKD2 are dispensable for pluripotency and differentiation potential. These observations are in line with the previous report 27 .
Next, we confirmed that mutations in iPSC cells could be maintained across multiple passages and during differentiation. To achieve this, samples were collected at different time points and the editing frequency analyzed on the pool of cells (Supplementary Fig. S1e). We observed similar editing frequency (80-90%) across all collected samples, indicating that the edited cells were not depleted during passages.
We then evaluated whether PKD knockout pools (PKD KOp) could be differentiated into kidney organoids to obtain the previously observed cystic phenotype 23,24 . Both PKD KOp hiPSCs and isogenic unmodified cells (control) were differentiated into kidney organoids (Fig. 4A). We could detect the expected formation of fluidfilled cyst-like structures in about 50% of the PKD KOp organoids but at a very low rate (less than 1%) in our control organoids (Fig. 4B-D). Achieving an efficient gene-editing frequency (> 80%) therefore eliminates the need for laborious clonal cell line isolation. This is an advantage for target validation studies where large sets of genetic factors need to be experimentally tested.
Taken together, we described a highly tractable 3D kidney organoid model that consistently recapitulated ADPKD cystogenesis in vitro. Going forward, this organoid model of ADPKD may provide a high-throughput opportunity to study human cyst initiation and modification in a disease-relevant multicellular model.

Conclusion
In this study we demonstrated that efficient genome editing could be achieved using an inducible CRISPR-Cas9 expressing iPSC cell line to speed up generation of genetically-modified kidney organoids. Our results are useful in an industrial settings where large scale experiments are desired to screen reliable phenotypes for identification of targets for initiating drug discovery programs. In conventional settings, the need for generation of the clonal lines is a major bottleneck in scalability of the target identification and target validation studies. By using knockout pools instead of generating isogenic clonal lines, our approach would enable faster turnaround time by using knockout pools instead of isogenic clonal lines. Moreover, our work presents a novel disease model for ADPKD target validation.

Experimental procedures
Cell culture. All experiments presented in this study used the previously described doxycycline-inducible Cas9 expressing hiPSC line 21 . Undifferentiated hiPSCs were maintained and transfected in the feeder-free and chemically defined culture system DEF-CS 500 (Cellartis by Takara Bio Europe). Cells were passaged every three to four days using TrypLE Select (Gibco) and single cells reseeded onto fresh DEF-coated plate using DEF-CS media with 10 µM Rock inhibitor (Tocris bioscience). Quality control of the cells was based on morphology, kar-

CRISPR-Cas9 design and generation.
Deep targeted amplicon sequencing. Genomic DNA was isolated either using Puregene cell and tissue kit (Qiagen) or QuickExtract DNA Extraction Solution (Lucigen, Epicenter) according to the manufacturer's recommended protocols. Gene specific primers were designed using Primer-BLAST tool 29 . Targeted genomic sites were first amplified using gene-specific primers linked with the Illumina Nextera adapters (listed in Supplementary Table S1) using PhusionFlash High-Fidelity PCR MasterMix (Thermo). Amplifications were performed either using 100 ng of genomic DNAs extracted with PureGene kit or using 2 µl of the cell lysate in 20 µl PCR containing 0.25 µM final concentration of each forward and reverse primers. The following thermal cycling condition was used in the first PCR: 1 cycle initial denaturation (2 min at 98 °C), 35 amplification cycles (10 s at 98 °C, 10 s at 60 °C, 5 s at 72 °C) and 1 cycle final extension (5 min at 72 °C). Amplicons then were purified with Agencourt AMPure XP beads (Beckman Coulter) according to the manufacturer's recommended protocol and concentrations were measured using Fragment Analyzer (Agilient's Advanced Analytical Technologies). A www.nature.com/scientificreports/ secondary indexing PCR was performed using Illumina Indexes followed by a purification and concentration measurement as previous described. The indexed libraries were pooled and subjected to paired-end sequencing using mid-throughput 300 cycles (2 × 150 cycles) on an Illumina NextSeq500 instrument. Sequencing reads were analyzed as described previously 30 . Briefly, sequencing reads were quality controlled and merged using FLASH software 31 . Reads were mapped to the amplicon reference sequence using bwa software 32 . Variants (excluding the single nucleotide variations) were called with a minimum allele frequency of 0.1% and a minimum base quality of 25 Phred-33 scores. The resulting variants were analyzed using RIMA 30 . Variants not overlapping of the cutting window (± 2 base pairs of the cut-site) were excluded. Finally, the editing efficiencies were measured as the percentage of modified reads in mapped reads.