Differential Tissue Specific Expression of Kif23 Alternative Transcripts In

Blood Cells, Molecules and Diseases 85 (2020) 102483

Contents lists available at ScienceDirect

Blood Cells, Molecules and Diseases

journal homepage: www.elsevier.com/locate/bcmd

Diﬀerential tissue speciﬁc expression of Kif23 alternative transcripts in mice with the human mutation causing congenital dyserythropoietic anemia type T III ⁎ Ann-Louise Vikberg, Sandhya Malla, Irina Golovleva

Department of Medical Biosciences, University of Umeå, Umeå, SE 901 87, Sweden

ARTICLE INFO ABSTRACT

Editor: Mohandas Narla Kinesin Family Member 23 (KIF23), a cell cycle regulator, has a key task in cytokinesis. KIF23 over-expression in Keywords: cancer has been associated with tumor growth, progression, and poor prognosis, indicating a potential to be a KIF23 cancer biomarker. A mutation in KIF23 (c.2747C > G, p.P916R) was shown to cause congenital dysery- Srsf3 thropoietic anemia, type III (CDA III). To-date, fifteen KIF23 transcripts have been annotated, but their ex- CDA III pression is poorly investigated. We hypothesized that tissue specific expression of a particular transcript can be Expression critical for CDA III phenotype. In this study, we quantified expression of alternative Kif23 transcripts in a mouse Alternative splicing model with human KIF23 mutation and investigated its association with a regulator of alternative splicing, Droplet digital PCR (ddPCR) serine/arginine-rich splicing factor 3 (Srsf3). We confirmed presence of an additional exon 8 in both human and Knock-in (KI) mice mouse KIF23 transcripts. A transcript lacking exons 17 and 18 was ubiquitously expressed in mice while other isoforms were common in human tissues however in bone marrow of knock-in mice a transcript without exon 18 was prevalent as it was in bone marrow of a CDA III patient. We conclude that the possibility that the tissue specific expression of KIF23 alternative transcripts influence the CDA III phenotype cannot be neglected.

1. Introduction annotated (https://www.ensembl.org) that include full-length (KIF23- FL) and two shorter protein coding transcripts. The shorter KIF23 iso- Kinesin family member 23 (KIF23) also known as Mitotic Kinesin- forms derive from a transcript lacking exon 18 (KIF23-Δ18) and a Like Protein 1 (MKLP1) has a key role in dividing cell cytoplasm at the transcript lacking exons 17 and 18 (KIF23-Δ17–18). Previously, it has end of mitosis and meiosis [1,2]. Due to failure of cytokinesis caused by been shown that KIF23-Δ18 expression was ubiquitous while KIF23-FL KIF23 p.P916R mutation multinucleated erythroblasts appear in the and KIF23-Δ17–18 were expressed in a tissue specific manner [3]. bone marrow of patients with a rare hereditary form of congenital While KIF23 was not detected in normal liver tissue, in hepatocellular dyserythropoietic anemia, type III (CDA III) described in one American carcinomas (HCC) KIF23-FL demonstrated nuclear localization and and one Swedish family [3]. Further study of the Swedish family has KIF23-Δ18 was localized in the cytoplasm [8]. Expression of KIF23-FL shown an increased prevalence of myeloma and monoclonal gammo- was shown to be associated with a longer 5-year survival of HCC pa- pathy of undetermined significance among CDA III patients [4]. Cor- tients, while KIF23-Δ18 protein did not affect 3- and 5-year survival relation of KIF23 over-expression with increased tumor growth and [8]. progression and poor survival of cancer patients shown in several stu- It is known that most human genes are subjected to alternative dies has led to proposal of the potential impact of the protein as a splicing, a process contributing to protein diversity although aberrant molecular biomarker in cancer [5–10]. This hypothesis has further been splicing is associated with many human diseases, including cancers tested in a clinical study where correlation has been seen between a [ 12]. Among numerous factors implicated in alternative splicing, high expression of four genes including KIF23 and poor therapy re- serine/arginine-rich splicing factor 3 (SRSF3) is of particular interest sponse in glioma patients [11]. because it regulates KIF23 splicing by promotion of exon 18 skipping, Despite the important role of KIF23 in cell division and potential resulting in the production of a short KIF23-Δ18 isoform [13,14]. role in cancer little is known about KIF23 expression in normal and The availability of molecular genetic testing instead of bone marrow tumor tissues. In human tissues fifteen KIF23 transcripts have been examination for CDA III diagnosis significantly limited access to

⁎ Corresponding author. E-mail addresses: [email protected] (A.-L. Vikberg), [email protected] (S. Malla), [email protected] (I. Golovleva). https://doi.org/10.1016/j.bcmd.2020.102483 Received 1 June 2020; Received in revised form 30 July 2020; Accepted 30 July 2020 Available online 31 July 2020 1079-9796/ © 2020 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/). A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483 experimental material in order to understand the uniqueness of CDA III precipitation of liver cells followed by an additional wash with 700 μL phenotype with the presence of multinuclear erythroblasts only in bone after DNase I treatment. Brain tissue formed a huge precipitate; there- marrow [3]. Considered this difficulty and the possibility to study this fore, the interfering proteins were removed by vigorous shaking with rare disease we created a mouse model with knock-in of the human 200 μL phenol-chloroform mix for 15 s followed by incubation for KIF23 p.P916R mutation. Using these mice expression of mouse Kif23 2–3 min. The aqueous phase was transferred to a new tube after cen- transcripts and how it correlates with expression of Srsf3 was in- trifugation at 12000g for 15 min at 4 °C. The following steps of RNA vestigated. We confirmed presence of an additional exon 8 in both extraction were done according to the manufacturer's protocol. The human and mouse KIF23/Kif23 transcripts and revealed significant RNA concentration and quality were estimated as described previously. variations in distribution of KIF23/Kif23 alternative transcripts quantified by reverse transcriptase droplet digital PCR (RT-ddPCR) in 2.4. cDNA synthesis healthy human and murine tissues. 1.0 μg RNA was denatured at 65 °C for 10 min and used in a 20 μL 2. Materials and methods reaction mix with 100 U M-MLW Reverse Transcriptase enzyme, 1× First Strand buffer, 10 mM DTT, 16.5 ng Random primers (all from 2.1. Human and mouse RNA samples Invitrogen™, Thermo Fisher Scientific), 1 mM dNTP (New England BioLabs, Herts, UK) and 16.0 U RNase OUT™ Recombinant Human KIF23 RNA expression was analyzed in Human Total RNA Ribonuclease Inhibitor (Thermo Fisher Scientific). Reverse transcrip- Master Panel II (Clontech, Palo Alto, CA) consisting of 14 tissues where tion was done using a Veriti® 96 Well Thermal Cycler (Applied specimens were pooled from 1 to 24 individuals. Mouse Kif23 RNA Biosystems, Foster City, CA, USA) with following temperature steps: expression was tested in Mouse Total RNA Master Panel (Clontech) 20 °C for 10 min, 42 °C for 45 min and final inactivation of the enzyme consisting of 11 tissues of BALB/c mice where specimens were pooled at 99 °C for 3 min. from 105 to 800 animals. We also analyzed KIF23 expression in peripheral blood (PB) and bone marrow (BM) of anonymized individuals 2.5. Droplet digital PCR (ddPCR) with or without CDA III whose samples were collected from Biobank Norr 472 (Clinical Genetics, Laboratory Medicine, University Hospital Expression of KIF23, Kif23 and Srsf3 was measured by absolute of Umeå). RNA from only one CDA III patient was accessible for this quantification of transcripts in different human and murine tissues. The study. Umeå Regional Ethical Review Board approved the study (M- KIF23-Δ18 and KIF23-Δ17–18 transcripts were multiplexed and mea- 2014/218-31). To avoid freeze/thaw cycles the tissues' RNA was ali- sured simultaneously, while KIF23-FL and Srsf3 were quantified sepa- quoted and each aliquot was used once. Mouse Kif23 RNA expression rately. Each biological sample was measured 2–6 times. Quantification was also studied in 9 tissues of 12 weeks old CDA III mice. KIF23 mu- was done by using EvaGreen® technique that enables detection of tation, c.2747C > G, p.P916R (NC_000015.10) causing CDA III in specific targets based on a size of amplified PCR products. Primers se- humans was knocked-in into C57BL/6 mice. This position corresponds quences and sizes of the PCR products are listed in (Supplemental to mouse Kif23 c.2726C > G, p.P909R (ENSMUSG00000032254), Material S1). 1× QX200™ ddPCR™ EvaGreen® Supermix (Bio-Rad, therefore, we defined wild type mice as Kif23WT/WT, heterozygotes - Hercules, CA, USA) was mixed with 20 nM of each primer in a total Kif23P909R/WT and homozygotes as Kif23P909R/P909R. The model was reaction mix of 22 μL. For quantification of human KIF23 transcripts a created at InGenious Targeting Laboratory, USA. The detailed descrip- volume of cDNA template was 3.3 or 4.4 μL with exception for PB and tion is provided in Supplemental Material S1. We studied Kif23P909R/ BM (1.1 μL) and brain, liver, lung, salivary gland, skeletal muscle and P909R, Kif23P909R/WT and Kif23WT/WT mice with three males and three uterus (5.5 μL). 4.4 μL cDNA was used for Kif23-FL detection and 2.2 μL females in each genotype group (n = 6) (Supplemental Material S1). cDNA was used for the shorter Kif23 transcripts, except for testis and The animals were not randomized, and the experiments were not done thymus where cDNA input was 0.55 μL the same as for Srsf3. in a blinded manner. Droplets were generated in a QX200™ Droplet Generator (Bio-Rad) according to manufacturer's instructions using 20 μL of reaction mix 2.2. RNA extraction from PB and BM and 70 μL of QX200™ Droplet Generation Oil for EvaGreen® (Bio-Rad). PCR reactions were carried out in a Veriti® 96 Well Thermal Cycler BM cells from mice were harvested by flushing BM cells from femur (Applied Biosystems) starting with initial steps at 50 °C for 2 min fol- and tibia with PBS containing 0.8% EDTA. The cells were pelleted and lowing a hot start at 95 °C for 10 min and 30 cycles with denaturation at re-suspended in 200 μL PBS/EDTA. To extract enough RNA equal vo- 95 °C for 10 min and annealing/elongation at 60 °C for 1 min. lumes of PB from 3 mice with the same genotype were pooled. RNA was Annealing/elongation time for Kif23-Δ18 and Kif23-Δ17–18 was 35 s. extracted using TRIzol™ in accordance with manufacturer's instructions Droplets and their EvaGreen signals were detected using QX200™ (Invitrogen, Thermo Fisher Scientific). Finally, the RNA was re-sus- Droplet Reader (Bio-Rad) according to manufacturer's instructions. pended in 10 μL of DEPC pH2O, heated at 60 °C for 10 min, quantified QuantaSoft Software, version 1.7, Regulatory Edition (Bio-Rad) was by Nanodrop 2000 (Thermo Fisher Scientific) and stored in aliquots at used for collection of the raw data. −80 °C. 2.6. Analysis of ddPCR data 2.3. RNA extraction from solid tissues Analysis of cluster generation and absolute quantification of tran- All collected tissues were preserved in RNAlater (Qiagen, Venlo, scripts' copies were estimated by using the Amplitude Multiplex func- Netherlands). RNeasy® Mini Kit (Qiagen) was used for brain, liver, tion, QuantaSoft™ Analysis Pro Software (Bio-Rad). A standard cut-off lungs, and testis while RNA from heart, kidney and spleen was ex- level discriminating between positive and negative droplets was set for tracted using RNeasy® Fibrous Tissue Mini Kit (Qiagen) according to the KIF23-FL, Kif23-FL and Srsf3 amplified separately. This cut-off level was manufacturer's instructions. 15–20 mg of each tissue placed in a 2.0 mL calculated as 40% of the signal intensity difference between negative tube with 5 mm stainless bead (Qiagen) and RLT buffer with 1% β- droplets and the droplets with the highest signal in the target cluster. 3 mercaptoethanol was disrupted at 50 Hz for 4 min using Tissue Lyser LT cut-off levels were applied for the multiplexed targets to distinguish (Qiagen). Spleen had to be disrupted in double amount of RLT buffer between the KIF23-Δ18 and KIF23-Δ17–18 in both human and mouse and incubated with Proteinase K. All tissues were treated with DNase I tissues. The first cut-off was set in the middle between the two target according to the manufacturer's protocol. 50% ethanol was used for clusters. The bottom level cut-off for KIF23-Δ17–18 and the upper cut-

2 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Fig. 1. Alternative transcripts of human and mouse KIF23/Kif23. (A, B) Schematic representation of KIF23/Kif23 exons 7 and 8 splicing and Sanger sequencing of exon-exon boundaries. Boundaries between exon 7 and exon 8 are shown on the left side. Boundaries between exon 7 and x8 and x8 and exon 8 are shown in the middle and on the right side. (C) Schematic representation of mouse Kif23 exons 16–19 and Sanger sequencing of exon-exon boundaries in alternative transcripts Kif23-Δ18 and Kif23-Δ17–18. Kif23-Δ18 is shown on the left side and Kif23-Δ17–18 is shown on the right side. Dashed lines show conventional splicing and solid lines show alternative splicing. Exons are shown in white (exons 7, 8, 16, 18, 18, 19) and alternative exon 8 (x8) is shown in grey. The primer sequences are provided in Supplemental Material S1. off level for KIF23-Δ18 were set so that the center of each cluster has the 72 °C for 1 min. Other KIF23/Kif23 transcripts were sequenced to same distance to its top and bottom cut-off. In order to calculate the confirm ddPCR specificity and presence of alternative transcripts with absolute concentration of each target in each tissue the output value the primers listed in Supplementary Material S1. PCR mix was prepared from QuantaSoft Software was adjusted in respect to the input of and run the same way as described in Digital Droplet PCR section. PCR template volume using following formula: ((raw data value ∗ reaction products were cut out from the gel and PCR products were separated on volume) / template volume). a 2% agarose gel 1% Meta Phor™ Agarose (Lonza, Switzerland) and 1% Standard Agarose, Type LE (BioNordika, Sweden), cut out from the gel, purified with QIAEX II Agarose Gel Extraction Kit (Qiagen, Germany) 2.7. Sanger sequencing of KIF23/Kif23 transcripts using manufacturer's instructions and sequenced according to geno- typing protocol (Supplementary Material S1). Amplification of Kif23 exon 8 and alternative exon 8 (x8) in mice prior to sequencing was done using 1× KAPA 2G Buffer A, 1× KAPA Enhancer, 0.2 U KAPA 2G Robust HS DNA Polymerase (all from KAPA 2.8. Statistical methods Biosystems Inc., USA), 0.4 mM dNTP (New England BioLabs) and 125 ng cDNA in a total reaction mix of 50 μL. PCR was done using a The ddPCR data was analyzed using statistical functions in SPSS Veriti® 96 Well Thermal Cycler (Applied Biosystems) with initial de- Statistics version 21.0 (IBM Software, USA). Tables, diagrams, con- naturation at 95 °C for 5 min followed by 40 cycles of denaturation at fidence intervals and box plots were created using Excel 365 ProPlus 95 °C for 0.15 min, annealing at 58 °C for 0.15 min and elongation at (Microsoft, USA). Skewness, kurtosis, and their standard errors were 72 °C for 0.3 min. PCR was finalized by an additional elongation at calculated using SPSS Statistics. Then, normality was calculated by

3 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483 dividing the skewness/kurtosis by its standard error. If the quotient was Δ17–18 was evaluated by RT-ddPCR on cDNA derived from 14 human larger than 1.96 the data was considered not normally distributed. and 11 mouse tissues (Fig. 2). Extremely low values in replicate mea- Shapiro-Wilk test was also used for assessing normality. If a vast ma- surements were discarded as outliers due to technical PCR failure. jority of the data was not normally distributed, non-parametric statis- Shapiro-Wilk test showed normal data distribution except for KIF23-FL tical tests based on median values were applied (median test, Mann- in skeletal muscle, for KIF23-Δ18 in liver and for KIF23-Δ17–18 in Whitney U test, Kolmogorov-Smirnov test, and regression analysis). testis. Extensive differences in total KIF23 expression between tissues Parametric statistical tests were used if a large majority of the data was were revealed with the highest total KIF23 expression in testis, an normally distributed. For normally distributed data SPSS Statistics in- average expression in thymus, colon, spinal cord and placenta and low dependent samples t-test was applied using mean values. Two-tailed expression in salivary gland and kidney (Fig. 2A). KIF23-FL transcript significance level was set to p < 0.05. was the most prevalent in testis, spinal cord, kidney, prostate, brain, uterus, skeletal muscle, and salivary gland although the difference be- fi 3. Results tween transcripts was signi cant only in kidney and testis (p < 0.05). For the rest of the tissues the KIF23-Δ18 was dominant with the sig- fi ff 3.1. Human and murine KIF23/Kif23 transcripts ni cant di erence in KIF23 transcripts in colon, placenta, spleen, and thymus (p < 0.05). Previously we identified a novel KIF23 transcript (KIF23-Δ17–18) Total expression of mouse Kif23 was rather equal in all tissues, along with two other isoforms, full-length KIF23-FL except for testis and thymus, where the Kif23 transcripts accounted for μ μ (ENST00000260363.9) and KIF23-Δ18, lacking exon 18 3221 copies/ L and 928 copies/ L, respectively (Fig. 2B). The Kif23- Δ – (ENST00000352331.8) [3]. In the present study, we reviewed currently 17 18 transcript was the most dominant (p < 0.05) in all tissues ff fi annotated human KIF23 transcripts and found that transcript except testis and the di erence was signi cant in all tissues except for Δ KIF23–207 (ENST00000559279.6) lacks exons 17 and 18, as KIF23- thymus (p > 0.05). The expression of Kif23-FL and Kif23- 18 was low Δ17–18 does, and retains partial sequence of intron 7 in all tissues except for the testis and thymus (Fig. 2B). Kif23-FL was fi Δ (chr15:69,425,282-69,425,323) coding for alternative exon 8 (x8). RT- signi cantly higher than Kif23- 18 in brain, kidney, spinal cord, and PCR on cDNA extracted from testis and thymus followed by Sanger testis, while the opposite was seen in salivary gland (p < 0.05). No fi ff Δ sequencing revealed exon-exon boundaries and thus, confirmed pre- signi cant di erence in expression of Kif23-FL and Kif23- 18 was ob- sence of alternative x8 in human KIF23 (Fig. 1A). We also detected served in eye, liver, lung, spleen, thymus or 15-day embryo (p > 0.05). Kif23-Δ18, Kif23-Δ17–18 and a transcript with an alternative x8 in mice fi (Fig. 1B, C). In our experiments we define human and mice transcripts 3.3. Quanti cation of KIF23 in human PB and BM without exons 17 and 18 as KIF23-Δ17–18 or Kif23-Δ17–18. A motif search for comparing protein binding features between the isoforms Analysis of KIF23 expression in PB and BM from a CDA III patient revealed that KIF23-FL, KIF23-Δ18 and KIF23-Δ17–18 has 72, 74 and and a healthy control showed that total expression was higher in BM μ μ 79 domains or motifs, respectively with different matching scores and PB of control individual (1226 copies/ L in BM and 66 copies/ Lin μ μ ranging from low to high (Supplementary Material S2). KIF23-FL and PB) compared to 356 copies/ L in BM and 16 copies/ L in PB from the Δ KIF23-Δ18 share 8 motifs that KIF23-Δ17–18 does not have. KIF23-Δ18 CDA III patient (p < 0.05) (Fig. 3). KIF23- 18 was the most prevalent and KIF23-Δ17–18 share 3 motifs that KIF23-FL does not have. KIF23- in PB and BM of both control and CDA III (Fig. 3). FL has 2 unique motifs, KIF23-Δ18 has 1 unique motif and KIF23- fi Δ17–18 has 14 unique motifs that none of the other transcripts share. 3.4. Quanti cation of Kif23 expression in a CDA III mouse model The search results are summarized in Table 1. Since mouse transcript containing alternative exon 8 (Ensembl, To understand CDA III molecular mechanisms we investigated Kif23 ff ENSMUST00000214295.1) is not completely annotated missing the C- expression in di erent tissues of knock-in (KI) mice with human KIF23 terminal end, a comparison to Kif23-FL lacking alternative exon 8 was p.P916R mutation that corresponds to p.P909R in mouse homologue done including only the first 500 amino acids covering the region of (Supplemental Material S1, Materials and Methods). Both, hetero- P909R/WT P909R/P909R alternative exon and the domain it belongs to. These two proteins share zygous (Kif23 ) and homozygous (Kif23 ) mice were all 30 motifs found in this region. Alternative exon 8 resulted in slightly viable, active, lacked characteristic features of human CDA III and grew weaker matching scores for the most kinesin domain motifs which old without any signs of any disease. A bone marrow examination might affect the it's interaction ability. showed no signs of erythroid population with several nuclei in bone marrow smears from Kif23P909R/WT and Kif23P909R/P909R mice. Despite the absence of CDA III phenotype the effect of the mutation on splicing 3.2. Quantification of KIF23 and Kif23 expression in healthy human and remained to be a target of investigation. For analysis of the Kif23 ex- mouse tissues pression we collected 9 different tissues from the mice with different genotypes Kif23P909R/WT, Kif23P909R/P909R and Kif23WT/WT. Normal KIF23 expression represented by KIF23-FL, KIF23-Δ18 and KIF23- distribution of the data was observed in 80% of the datasets. The rest of the datasets were slightly right skewed, making only non-parametric Table 1 statistical tests applicable. Due to small group size (n = 6, 3 females Summary of motif search for KIF23/Kif23 isoformsa. and 3 males) non-parametric tests could not generate enough power to Motifs Shared Shared Shared Unique reach significance with p < 0.05; therefore, all estimations were made found motifs motifs motifs motifs from visual observations and all figures are based on median values. ff KIF23-FL 68 62 8 – 2 Kif23 expression did not show any di erence regarding gender KIF23-Δ18 74 3 1 (Supplemental Material S3) and in our further analyses we combined KIF23-Δ17–18 79 – 14 the data from both, male and female animals. Since only cases with Kif23-FL 30 30 –– 0 heterozygous P916R mutation in KIF23 have been associated with CDA Kif23-x8 30 –– 0 III and no homozygotes in any of the two affected CDA III families were a Search of the motifs for each KIF23 isoform was done via MOTIF Search described, we decided to only compare heterozygous mutation carriers P909R/WT WT/WT (GenomeNet, Institute for Chemical Research, Kyoto University, Japan) using mice (Kif23 ) with wild type (WT) mice (Kif23 ). The data default settings. The results include matches from Pfam, Prosite and NCBI-CDD on expression of Kif23 alternative transcripts in the mice of all geno- databases. types is provided in Supplemental Material S3. Based on median values

4 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Fig. 2. Absolute quantification of KIF23/Kif23 expression in human and mouse tissues. RT-ddPCR was performed on total RNA extracted from 14 healthy human tissues (A) and from 11 healthy tissues from mice (B). Expression of three KIF23/Kif23 transcripts was quantified as described in Materials and Methods. KIF23/Kif23- FL (black circle) is a full-length transcript, KIF23/Kif23-Δ18 (black triangle) is a transcript lacking exon 18 and KIF23/Kif23-Δ17–18 (white square) is a transcript without exons 17 and 18. The values including confidence intervals represent a mean of each transcript, measured 2 to 6 times in each tissue. The differences in expression are significant (p < 0.05) for KIF23-FL in kidney and testis and KIF23-Δ18 in colon, placenta, spleen, and thymus. Expression of Kif23-Δ17–18 was significantly higher than all other transcripts except for testis (p < 0.05). Expression of Kif23-FL was significantly higher than Kif23-Δ18 in all tissues except brain, kidney, spinal cord, and testis, while the opposite is seen in salivary gland (p < .05). Total KIF23/Kif23 expression including all transcripts is shown above each tissue as a bold black line. Expression of KIF23/Kif23-FL in testis was higher than in any other tissue, which is shown on a left Y-axis.

5 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Fig. 3. Absolute quantification of KIF23 expression in peripheral blood (PB) and bone marrow (BM) of healthy control and a CDA III patient. RT-ddPCR was performed on total RNA extracted from PB and BM. Expression of three KIF23 transcripts was quantified as described in Materials and Methods. KIF23-FL (black circle) is a full-length transcript, KIF23-Δ18 (black triangle) is a transcript lacking exon 18 and KIF23-Δ17–18 (white square) is a transcript without exons 17 and 18. The values including confidence intervals represent a mean of each transcript, measured 2 to 6 times in each tissue. Total KIF23 expression including all transcripts is shown above each tissue as a bold black line. The values are from one individual each and represent a mean of several measurements, including their confidence intervals. Total KIF23 expression in control BM was higher than in PB, which is shown on a left Y-axis. Total KIF23 is significantly higher in control than in CDAIII patient in blood and bone marrow (p < 0.05).

Kif23-Δ17–18 was the most prevalent transcript in brain, heart, liver, 3.5. Quantification of Srsf3 expression in CDA III mouse model and lung. Kif23-FL dominated in kidney and testis (Fig. 4). In spleen of Kif23P909R/WT mice a slight shift from Kif23-Δ18 to Kif23-Δ17–18 was Recently it has been shown that a serine/arginine-rich splicing noted though the change was too small to be significant in this dataset. factor 3 (SRSF3) regulates alternative splicing by the promotion of exon Analysis of Kif23 expression in PB and BM demonstrated the highest 18 skipping in KIF23, thus resulting in KIF23-Δ18 isoform [15]. Lack of level of total Kif23 expression in BM (Fig. 5). Both, in BM and PB the human CDA III material did not allow studies of SRSF3 expression, most prevalent transcript was Kif23-Δ18 (Fig. 5), in contrast to the therefore, we examined Srsf3 expression only in our mouse model. No tissues from BALB/c mice with prevalent Kif23-Δ17–18 (Fig. 1B). significant difference in Srsf3 expression was observed between genotypes, however the Srsf3 expression tended to be higher in all tissues

Fig. 4. Kif23 tissue specific expression in Kif23P909R/WT and Kif23WT/WT (C57BL/6) mice is shown as box and whisker chart. Kif23-FL-is shown as a dark grey box; Kif23-Δ18-light grey box; Kif23-Δ17–18-transcript white box; bold line indicates total expression of all Kif23 transcripts based on median values. Differences in expression levels between groups within this dataset does not reach significance (p > 0.05).

6 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Fig. 5. Box and whisker plot analysis of Kif23 expression in peripheral blood (PB) and bone marrow (BM) of Kif23P909R/WT and Kif23WT/WT mice (C57BL/ 6). Kif23-FL-is shown as a dark grey box; Kif23-Δ18- light grey box; Kif23-Δ17–18–white box; bold line indicates total expression of all Kif23 transcripts. Peripheral blood from 3 Kif23P909R/WT and Kif23WT/ WT mice, respectively was pooled before measurement, resulting in only one measure point for each transcript. Differences in expression of different transcripts within this dataset does not reach significance (p > 0.05).

from Kif23WT/WT mice compare to Kif23P909R/WT (Fig. 6A). To in- of Kif23-Δ17–18 detected in this study. The functional heterogeneity of vestigate variations in Srsf3 expression in the Kif23P909R/WT mice we KIF23 (MKLP1/CHO1) due to alternative splicing was earlier demon- performed regression analysis using regression curves (Fig. 6B). Mod- strated by interaction of F-actin with the sequence encoded by exon 18 erate correlation between total amount of Srsf3 and Kif23 expression (KIF23-FL) implying association with both microtubules and actin cy- was seen in BM for both Kif23WT/W (R2 = 0.65) and Kif23P909R/WT toskeletons [16]. Moreover, it was shown that the antibodies specific (R2 = 0.75) (Fig. 6B). Using regression analysis, we also tested whether for exon 18 did affect only the terminal phase of cytokinesis resulting in Srsf3 expression correlates with expression of different Kif23 transcripts bi-nucleated cells, the exact result seen in HeLa cells in rescue experi- regarding Kif23 genotype. The association (R2) and direction of the ments with KIF23 p.P916R [3]. The likely role of protein sequences association revealed an overrepresentation of positive association in encoded by exon 18 and probably exon 17, is direct or indirect parti- expression between Kif23 and Srsf3 (Table 2). The correlations seemed cipation in the membrane events necessary for completion of the to be independent of the transcript. The associations vary from very low terminal phase of cytokinesis that is defective in CDA III. Interestingly, to high, and no pattern was seen that could suggest selection of any that the binding motifs identified through MOTIF (GenomeNet, In- Kif23 transcript by splicing factor Srsf3. stitute for Chemical Research, Kyoto University, Japan) in human KIF23 isoforms were different with some domains being gained or lost regarding the full-length protein. KIF23-FL and KIF23-Δ18 has similar 4. Discussion matches to motifs with a few differences, while KIF23-Δ17–18 has a more unique motif profile. In general, KIF23-Δ17–18 with its alter- KIF23 was recently proposed to function as a molecular biomarker in cancer [5–10]. It had also been shown that full-length protein native exon (x8) inside the kinesin motor domain has slightly weaker match score than KIF23-Δ18 and KIF23-FL. It is possible that KIF23 (KIF23-FL), localized in nucleus was associated with a longer survival of patients with hepatocellular carcinomas compared to patients with a binding to microtubules might be impaired although functional studies should confirm if this sequence influences protein binding via the motor shorter protein, encoded by KIF23-Δ18, detected in the cytoplasm [8]. Currently, 5 human and 2 mouse protein-coding transcripts are anno- domain. In mice the prediction tool showed that the alternative exon 8 does not introduce or exclude any motifs as it does in human KIF23. tated in the genome browser of vertebrate genomes ENSEMBL (Ensembl fl release 100, April 2020) however the role of alternative KIF23 tran- This suggests that the alternative exon 8 might in uence the phenotype more in human than in mice. Further it was noticed that the loss of exon scripts has been poorly studied. In congenital dyserthropoietic anemia (CDA III) a tissue-specific KIF23 expression was observed, revealing 17 and 18 creates new motifs and that shared motifs are located between 500 and 650 amino acids residues in KIF23 sequence. Most of the transcript lacking exons 17 and 18 (KIF23-Δ17–18), mainly present in peripheral blood, bone marrow and spleen when analyzed by qualita- motifs shared between the three KIF23 isoforms get a slightly higher match score in KIF23-Δ17–18 compared to KIF23-FL or KIF23-Δ18. A tive RT-PCR [3]. Notably, currently annotated human transcript motif present in KIF23-Δ17–18 and absent in KIF23-FL and KIF23-Δ18 without exons 17 and 18 has an additional exon 8 (x8) (KIF23–207, is PF11559 in ADIP (afadin- and alpha -actinin-binding). Little is known ENST00000559279.6) that was confirmed in this study. We also de- about ADIP, though it is found in cell-to-cell adherence junctions in tected sequence corresponding to x8 with identity of 83% in mouse human cells [17], and in fungi it participates in cytokinesis, promoting Kif23, that might represent annotated Kif23–204 transcript. 3′-sequence a correct spindle anchoring. Knockdown of ADIP was shown to cause of this transcript is not complete (Ensembl release 100, April 2020) and faulty chromosome segregation [18]. A motif present in KIF23-FL and lacks exons 17 to 23 making impossible to conclude whether x8 is part

7 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Fig. 6. (A) Tissue specific expression of Srsf3 gene in Kif23WT/WT (C57BL/6) (grey boxes) and Kif23P909R/WT mice, heterozygotes for human KIF23 c.2747C > G mutation (white boxes) is presented as box and whisker plot. Peripheral blood was pooled before measurement resulting in only one value. Differences in expression levels between groups within this dataset does not reach significance (p > 0.05). Regression analysis shows correlation between Kif23 and Srsf3 expression in bone marrow (BM) (B). Correlation curves are shown for Kif23P909R/WT and Kif23WT/WT. Correlation is presented as R2 values where R2 = 0 means absence of correlation and R2 = 1.00 means 100% correlation.

KIF23-Δ18 but absent in KIF23-Δ17–18 is STAT5-CCD, known as a showed that both human KIF23 and mouse Kif23 were expressed in all mediator of common and cell-type-specific erythropoiesis regulation tissues with considerable diversity of expression level in different tis- [19] and also as a driver in the development of leukemia and T-cell sues. With our large dataset from the mouse model we also showed that lymphoma [20]. Future studies are required to uncover cell-specific levels of both total Kif23 and each transcript included in this study interaction between KIF23 transcripts and STAT5 and its role in CDA III differ a lot between individuals which is the reason why we could not pathogenesis as well as in other blood malignancies. Considering that reach significance when comparing groups within the dataset. the disease resides mainly in erythroblasts of bone marrow demon- Human KIF23-Δ18 was present in all tissues while KIF23-FL and strating presence of multiple nuclei we hypothesized that tissue specific KIF23-Δ17–18 were expressed in tissue-specific manner. In this study alternative KIF23 splicing might possibly explain a higher incidence of we took advantage of ddPCR that allows absolute quantification of any blood malignancies in CDA III patients. In this study we found 2 murine target. Our results on total KIF23 expression agree with the BioGPS data transcripts currently not annotated, Kif23-Δ18 lacking exon 18 and (http://biogps.org) except for the highest expression level in testis de- Kif23-Δ17–18 lacking exon 17 and 18. Confirmed sequences of those monstrated in this study. In control murine tissues the highest level of transcripts covered exons 16 and 19 and their exon-exon junctions. We Kif23 expression was also observed in testis due to dominating Kif23-FL

8 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Table 2 Association of Kif23 transcripts expression and Srsf3 expression in BM in KI mice with diﬀerent c.2726C > G genotype.

Bone marrow (BM)

2 R Direction

Kif23WT/WT Kif23P909R/WT Kif23P909R/P909R Kif23WT/WT Kif23P909R/WT Kif23P909R/P909R

KIF23-FL 0.516 0.655 0.499 + + + KIF23-Δ18 0.877 0.441 0.152 + + n/a KIF23-Δ17–18 0.732 0.874 0.158 − + n/a KIF23-total 0.006 0.7518 0.499 n/a + +

The correlation is presented as R2 values where 1.0 means 100% correlation and 0.0 means absence of correlation. The direction is either positive (+) corresponding to increased expression as number of Kif23 mutated alleles (0, 1 or 2) increases or negative (−) when expression decreases as number of mutated alleles increases. Genotype of Kif23 c.2726C > G is shown as Kif23WT/WT for wild type homozygotes; Kif23P909R/WT for heterozygotes; and Kif23P909R/P909R for mutant homozygotes. BM-bone marrow, n/a means there is very low correlation therefore no direction or the correlation. transcript. In all other tissues Kif23 expression was almost equal and utilization of SRSF3-binding motifs, CCAGC(G)C and A(G)CAGCA pro- Kif23-Δ17–18 was prevalent, unlike the human tissues. ducing a short KIF23 transcript without exon 18 (KIF23-Δ18) [15]. In To analyze if any of KIF23 transcripts had alternate expression in our study we analyzed tissue specific Kif23 alternative splicing in the cells derived from a CDA III patients and to understand why the CDA III mice with human KIF23 mutation and how Kif23 genotype affects ex- phenotype resides in precursors of red blood cells, we examined spli- pression of Srsf3. The data concerning the total Kif23 expression and cing outcome in peripheral blood and bone marrow. Total KIF23 ex- prevalence of a transcript lacking exons 17 and 18 (Kif23-Δ17 –18) in pression was higher in bone marrow than in peripheral blood and the most tissues was consistent with the data from control mice. No sig- transcripts distribution with prevalent KIF23-Δ18 was similar in bone nificant differences in Kif23 expression were observed in homozygous marrow of control individual and the CDA III patient. To investigate or heterozygous knock-in animals. The individual variations in ex- whether KIF23 c.2747C > G variant, causing CDA III, somehow pro- pression of each Kif23 transcript were so substantial, that the sig- motes exon skipping the expression studies in erythropoietic precursors nificance could not be reached at comparison of different genotypes. No are warranted though access to bone marrow of CDA III patients is significance could be reached regarding Srsf3 expression levels, how- extremely limited. Previously, morphological assessment of bone ever if human KIF23 mutation leads to decrease of Srsf3 expression, we marrow and biochemical analysis of hemolytic anemia markers in might be able to observe more of transcripts with skipped exons in the serum was the way to diagnose CDA III, that has now been replaced by future studies using erythropoietic precursors. molecular genetic testing of peripheral blood, making bone marrow aspiration unnecessary and collection of bone marrow from CDA III 5. Conclusions patients became a real challenge. Due to inability to study any other tissues of CDA III patients we Differential tissue-specific expression of three KIF23 transcripts was used a mouse model with the human KIF23 mutation. Considering demonstrated in human and mouse tissues. The mice with the human KIF23 central role in cell division we were surprised that the mice CDA III mutation did not reveal any signs of the disease. The human homozygous for the human homologue mutation were viable KIF23-Δ18 was present in all tissues while KIF23-FL and KIF23-Δ17–18 P909R/P909R (Kif23 ). A microscopic examination of their bone marrows were expressed in tissue-specific manner. Mouse Kif23 was almost did not reveal CDA III signs such as bi- or multinucleated erythroblasts. equally expressed in all tissues with prevalence of Kif23-Δ17–18. It Notably, a disease model created for another type of congenital dys- remains to investigate if KIF23 c.2747C > G, p.P916R causes shift in erythropoietic anemia, type II, also failed to reproduce a disease spe- splicing in erythropoietic precursors cells. The possibility that the tissue fi ci c phenotype [21]. The mouse model of recessive CDA II with specific KIF23 expression and distribution of alternative transcripts fi SEC23B de ciency died perinatally, while in conditional mice where influence phenotype cannot be neglected and requires further in- the expression of the null allele was restricted to the hematopoietic vestigations. cells, animals survived without any signs of CDA II [21]. The absence of Supplementary data to this article can be found online at https:// a CDA II, and in our study CDA III phenotype, in mice could be ex- doi.org/10.1016/j.bcmd.2020.102483. plained by uniqueness of human SEC23B and KIF23 or evolutionary changes of their function, expression pattern or interaction network important for erythropoiesis [21,22]. Regarding the expression, the Funding different patterns of SEC23A/B expression were shown at different stages of maturation of human and murine erythroid progenitors Financial support was obtained through regional agreement be- [23,24] but, unfortunately, KIF23 was not part of this study. In a model tween Umeå University and Västerbotten County Council on coopera- fi of CDA I, usually caused by bi-allelic missense mutations in CDAN1 tion in the eld of Medicine, Odontology and Health (ALF), grant RV- (Codanin 1) gene, embryos homozygous for null allele died at early 583221 and the Cancer Research Foundation in Northern Sweden stages of development while heterozygous carriers demonstrated (Cancerforskningsfonden Norrland /Lions Cancerforskningsfond), grant widespread expression of codanin-1 [25]. By another words, despite AMP 20-994. Ann-Louise Vikberg received Kempestifterlserna fellow- critical roles of all three genes causing CDAs none of mice models were ship for this study. successful and the question why only erythropoiesis is affected is still open. CRediT authorship contribution statement Alternative RNA splicing and tissue-specific expression is re- sponsible for proteomic diversity and aberrant splicing is associated Ann-Louise Vikberg: Conceptualization, Methodology, Software, with human diseases and cancers [12]. Recently in has been shown that Validation, Formal analysis, Investigation, Writing - original draft, KIF23 splicing is regulated by a serine/arginine-rich splicing factor 3 Writing - review & editing, Funding acquisition. Sandhya Malla: (SRSF3) [15]. SRSF3 knockdown induces exon skipping due to Methodology, Formal analysis, Writing - review & editing. Irina Golovleva: Conceptualization, Validation, Investigation, Data curation,

9 A.-L. Vikberg, et al. Blood Cells, Molecules and Diseases 85 (2020) 102483

Writing - review & editing, Supervision, Funding acquisition. variant 1 expression and relevance as a novel prognostic factor in patients with hepatocellular carcinoma, BMC Cancer 15 (2015) 961. [9] K. Valk, T. Vooder, R. Kolde, M.A. Reintam, C. Petzold, J. Vilo, A. Metspalu, Gene Declaration of competing interest expression profiles of non-small cell lung cancer: survival prediction and new bio- markers, Oncology 79 (2010) 283–292. fl [10] J.X. Zou, Z. Duan, J. Wang, A. Sokolov, J. Xu, C.Z. Chen, J.J. Li, H.W. Chen, Kinesin The authors declare no con ict of interest. family deregulation coordinated by bromodomain protein ANCCA and histone methyltransferase MLL for breast cancer cell growth, survival, and tamoxifen re- Acknowledgments sistance, Mol. Cancer Res. 12 (2014) 539–549. [11] Q. Wang, Z. He, Y. Chen, Comprehensive analysis reveals a 4-gene signature in predicting response to Temozolomide in low-grade glioma patients, Cancer Control We acknowledge Dr. Liljeholm and Prof. Wahlin for providing in- 26 (2019) (1073274819855118). formation and collecting biological specimen from CDA III patient and [12] E.T. Wang, R. Sandberg, S. Luo, I. Khrebtukova, L. Zhang, C. Mayr, S.F. Kingsmore, Dr. Huldtin for microscopic examination of bone marrow smears of G.P. Schroth, C.B. Burge, Alternative isoform regulation in human tissue tran- scriptomes, Nature 456 (2008) 470–476. CDA III model mice. [13] J.L. Manley, A.R. Krainer, A rational nomenclature for serine/arginine-rich protein splicing factors (SR proteins), Genes Dev. 24 (2010) 1073–1074. Ethics approval and consent to participate [14] V. Majerciak, M. Lu, X. Li, Z.M. Zheng, Attenuation of the suppressive activity of cellular splicing factor SRSF3 by Kaposi sarcoma-associated herpesvirus ORF57 protein is required for RNA splicing, RNA 20 (2014) 1747–1758. A study with use of human DNA from CDA III patients was approved [15] M. Ajiro, R. Jia, Y. Yang, J. Zhu, Z.M. Zheng, A genome landscape of SRSF3- by Umeå Regional Ethical Review Board (M-2014/218-31). Studies regulated splicing events and gene expression in human osteosarcoma U2OS cells, Nucleic Acids Res. 44 (2016) 1854–1870. involving transgenic mice were approved by Regional Animal Research [16] R. Kuriyama, C. Gustus, Y. Terada, Y. Uetake, J. Matuliene, CHO1, a mammalian Ethics Committee (A81-14). kinesin-like protein, interacts with F-actin and is involved in the terminal phase of cytokinesis, J. Cell Biol. 156 (2002) 783–790. [17] M. Asada, K. Irie, K. Morimoto, A. Yamada, W. Ikeda, M. Takeuchi, Y. Takai, ADIP, Availability of data and materials a novel Afadin- and alpha-actinin-binding protein localized at cell-cell adherens junctions, J. Biol. Chem. 278 (2003) 4103–4111. The authors declare that all data supporting the findings of this [18] M. Toya, M. Sato, U. Haselmann, K. Asakawa, D. Brunner, C. Antony, T. Toda, study are available within the article and its supplementary information Gamma-tubulin complex-mediated anchoring of spindle microtubules to spindle- pole bodies requires Msd1 in fission yeast, Nat. Cell Biol. 9 (2007) 646–653. files. The datasets generated, used, and analyzed during this study are [19] E. Porpiglia, D. Hidalgo, M. Koulnis, A.R. Tzafriri, M. Socolovsky, Stat5 signaling available from the corresponding author on reasonable request. specifies basal versus stress erythropoietic responses through distinct binary and graded dynamic modalities, PLoS Biol. 10 (2012) e1001383. [20] B. Maurer, H. Nivarthi, B. Wingelhofer, H.T.T. Pham, M. Schlederer, T. Suske, References R. Grausenburger, A.I. Schiefer, M. Prchal-Murphy, D. Chen, S. Winkler, O. Merkel, C. Kornauth, M. Hofbauer, B. Hochgatterer, G. Hoermann, A. Hoelbl-Kovacic, [1] C. Nislow, V.A. Lombillo, R. Kuriyama, J.R. McIntosh, A plus-end-directed motor J. Prochazkova, C. Lobello, A.A. Cumaraswamy, J. Latzka, M. Kitzwogerer, enzyme that moves antiparallel microtubules in vitro localizes to the interzone of A. Chott, A. Janikova, S. Pospisilova, J.I. Loizou, S. Kubicek, P. Valent, T. Kolbe, mitotic spindles, Nature 359 (1992) 543–547. F. Grebien, L. Kenner, P.T. Gunning, R. Kralovics, M. Herling, M. Muller, T. Rulicke, [2] R. Neef, U.R. Klein, R. Kopajtich, F.A. Barr, Cooperation between mitotic kinesins V. Sexl, R. Moriggl, High activation of STAT5A drives peripheral T-cell lymphoma – controls the late stages of cytokinesis, Curr. Biol. 16 (2006) 301–307. and leukemia, Haematologica 105 (2) (2020) 435 447. [3] M. Liljeholm, A.F. Irvine, A.L. Vikberg, A. Norberg, S. Month, H. Sandstrom, [21] R. Khoriaty, M.P. Vasievich, M. Jones, L. Everett, J. Chase, J. Tao, D. Siemieniak, A. Wahlin, M. Mishima, I. Golovleva, Congenital dyserythropoietic anemia type III B. Zhang, I. Maillard, D. Ginsburg, Absence of a red blood cell phenotype in mice fi – (CDA III) is caused by a mutation in kinesin family member, KIF23, Blood 121 with hematopoietic de ciency of SEC23B, Mol. Cell. Biol. 34 (2014) 3721 3734. (2013) 4791–4799. [22] K. Schwarz, A. Iolascon, F. Verissimo, N.S. Trede, W. Horsley, W. Chen, B.H. Paw, [4] H. Sandstrom, A. Wahlin, M. Eriksson, I. Bergstrom, S.N. Wickramasinghe, K.P. Hopfner, K. Holzmann, R. Russo, M.R. Esposito, D. Spano, L. De Falco, Intravascular haemolysis and increased prevalence of myeloma and monoclonal K. Heinrich, B. Joggerst, M.T. Rojewski, S. Perrotta, J. Denecke, U. Pannicke, ff gammopathy in congenital dyserythropoietic anaemia, type III, Eur J Haematol 52 J. Delaunay, R. Pepperkok, H. Heimpel, Mutations a ecting the secretory COPII (1994) 42–46. coat component SEC23B cause congenital dyserythropoietic anemia type II, Nat. – [5] T. Kato, H. Wada, P. Patel, H.P. Hu, D. Lee, H. Ujiie, K. Hirohashi, T. Nakajima, Genet. 41 (2009) 936 940. M. Sato, M. Kaji, K. Kaga, Y. Matsui, M.S. Tsao, K. Yasufuku, Overexpression of [23] X. An, V.P. Schulz, J. Li, K. Wu, J. Liu, F. Xue, J. Hu, N. Mohandas, P.G. Gallagher, ff KIF23 predicts clinical outcome in primary lung cancer patients, Lung Cancer 92 Global transcriptome analyses of human and murine terminal erythroid di er- – (2016) 53–61. entiation, Blood 123 (2014) 3466 3477. [6] H. Murakami, S. Ito, H. Tanaka, E. Kondo, Y. Kodera, H. Nakanishi, Establishment [24] N. Pishesha, P. Thiru, J. Shi, J.C. Eng, V.G. Sankaran, H.F. Lodish, Transcriptional of new intraperitoneal paclitaxel-resistant gastric cancer cell lines and compre- divergence and conservation of human and mouse erythropoiesis, Proc. Natl. Acad. – hensive gene expression analysis, Anticancer Res. 33 (2013) 4299–4307. Sci. U. S. A. 111 (2014) 4103 4108. [7] S. Takahashi, N. Fusaki, S. Ohta, Y. Iwahori, Y. Iizuka, K. Inagawa, Y. Kawakami, [25] R. Renella, N.A. Roberts, J.M. Brown, M. De Gobbi, L.E. Bird, T. Hassanali, K. Yoshida, M. Toda, Downregulation of KIF23 suppresses glioma proliferation, J. J.A. Sharpe, J. Sloane-Stanley, D.J. Ferguson, J. Cordell, V.J. Buckle, D.R. Higgs, Neuro-Oncol. 106 (2012) 519–529. W.G. Wood, Codanin-1 mutations in congenital dyserythropoietic anemia type 1 ff – [8] X. Sun, Z. Jin, X. Song, J. Wang, Y. Li, X. Qian, Y. zhang, Y. Yin, Evaluation of KIF23 a ect HP1{alpha} localization in erythroblasts, Blood 117 (2011) 6928 6938.