KDM3B SNPs impact radiation therapy toxicity through circular RNA-mediated KDM3B expression and inflammatory responses

Abstract

Purpose:

Genome-wide association studies (GWAS) have identified single nucleotide polymorphisms (SNPs) associated with radiotherapy (RT) toxicities in prostate cancer patients. SNP rs17599026 in intron 21 of KDM3B is significantly associated with the development of late urinary toxicity, specifically in the increase in urinary frequency two years after RT compared with pre-treatment conditions. The present study aimed to provide mechanistic insights for this association.

Methods and Materials:

Using human tissues and cell lines, we examined the protein expression of KDM3B and molecular mechanisms underlying the SNP modulation by variants of KDM3B SNP alleles. In animals with normal and heterozygous expressions of Kdm3b, we examined the relationship between Kdm3b expression and radiation toxicity.

Results:

KDM3B rs17599026 lies in a motif important for circular RNA expression which is responsible for sponging miRNAs to regulate KDM3B expression. Using a murine model with heterozygous deletion of Kdm3b gene, we found that lower Kdm3b expression is associated with altered pattern of urination after bladder irradiation, which is related to differential degrees of tissue inflammation as measured by analyses of gene expression, lymphocyte infiltration, and non-invasive ultrasound imaging.

Conclusions:

KDM3B SNPs can impact its expression through regulating noncoding (nc) RNA expression. Differential KDM3B expression underlies radiation toxicity through tissue inflammation at the molecular and physiological level. Our study outcome offers a foundation for mechanism-based mitigation for radiation toxicity for prostate cancer survivors.

Introduction

Prostate cancer is the most common non-cutaneous cancer in the US, with 1 in 8 men diagnosed in their lifetime [1]. Radiotherapy (RT) is responsible for 40% of all cancer cures [2, 3] and is widely used for the treatment of prostate cancer with approximately 50% patients receiving RT as either a monotherapy or in combination with surgery and/or hormonal therapy [2, 3]. An estimated 3.1 million prostate cancer survivors are currently living in the United States (American Cancer Society. Facts & Figures 2019. American Cancer Society. Atlanta, Ga. 2019), and 10-year survival rates for localized prostate cancer is above 95% [4]. Despite being an effective treatment modality for prostate cancer, RT is associated with toxicities of varying degrees of acute, subacute, and long-term late effects affecting urinary, bowel, and sexual function [5]. Late toxicity resulting from RT is a dose-limiting factor for clinical RT and an enduring issue for the quality-of-life for cancer survivors. Deciphering mechanisms of radiation-induced normal tissue toxicity can yield discoveries for potential interventions to prevent or mitigate late radiation toxicity, which will have a significant impact on the clinical care of millions of prostate cancer patients.

As late radiation toxicity is a result of complex pathophysiological interactions, understanding the mechanistic processes in humans poses a particular challenge to unravel the intricacies underlying the toxicities. Genome-wide association studies (GWAS) of radiation late toxicity have the potential to identify novel biologic targets and guide mechanistic studies of radiotoxicity. A recent GWAS among 1,638 patients treated with RT for prostate cancer identified SNPs associated with increased risk of developing late genitourinary and gastrointestinal toxicities after RT [7]. Among the genome-wide panels of SNPs examined, rs17599026 was significantly (p=4.16×10⁻⁸) associated with the development of late urinary toxicity, specifically an increase in urinary frequency two years after RT relative to pre-treatment conditions. While GWAS revealed valuable genetic associations in prostate cancer patients, the underlying mechanism of such association is completely unknown. This SNP (rs17599026) lies in an intron 23bp downstream of the splicing donor of the exon 20 of the KDM3B gene (MIM 609373, NM_0166042:c4753+23C>T). T/T and C/T genotypes are associated with increased odds of developing toxicity compared to C/C genotype. To date, KDM3B, a histone H3K9 lysine demethylase, has not been reported to play any role in cellular radiation responses, thus this gene may represent a novel radiation response gene or a novel regulator of normal tissue radiation responses.

The vast majority of SNPs are located in the non-coding regions of the human genome consistent with the fact that protein coding sequences are merely one percent of the whole genome, therefore understanding the molecular basis of these SNPs is particularly challenging as it is not immediately obvious how such SNPs can affect biological processes that are usually regulated by proteins. In reality, SNPs can impact gene expression from millions of bases away through three-dimensional chromatin organization, therefore their immediate genomic locations may not provide sufficient information for their potential functions [8]. On the other hand, even though non-coding (nc) sequences do not encode peptides, it has been increasingly clear that they indeed encode genetic information for roles that are both structural and regulatory in the form of non-coding RNAs (ncRNAs) including microRNAs (miRNAs), long non-coding RNAs (lncRNAs), and circular RNAs (circRNAs) [9–11]. MiRNAs can target 3’UTR of mRNAs through the Argonaute complex, thus repressing mRNA translation or degradation. Other functions of miRNAs have also been reported including activating gene transcription, increasing mRNA translation, and targeting other ncRNAs such as lncRNAs and circRNAs. LncRNAs have more than 200 bases that have been found to function as a scaffold or a decoy for signaling, and as structural components of macromolecules. CircRNAs are another form of ncRNAs that have widespread expressions, and due to their circular nature are generally more stable as a result of enhanced resistance to exonuclease degradation. Their functions not only include sponging miRNAs, but also encoding peptides, binding and competing with RNA binding proteins, regulating gene transcription, as well as serving as a structural component bridging/regulating multi-component complex to impact protein stability as well as ribonucleoprotein (RNP) complex functions.

We report that rs17599026 lies in the motif of RNA binding protein MBNL1 that has previously been implicated in circular RNA genesis [12], and the reference allele of rs17599026 allows a higher expression of circRNA that can impact KDM3B expression through miRNA sponge. Moreover, through a mouse model with heterozygous loss of Kdm3b gene, lower Kdm3b expression is correlated with enhanced tissue inflammatory response upon radiation, thus potentially providing novel mechanistic knowledge underlying the impact of this SNP on the late toxicity of radiation for prostate cancer patients.

Materials and Methods

Materials

Dulbecco’s modified Eagle’s medium (DMEM), high-glucose medium, RPMI 1640 medium, fetal bovine serum (FBS), and 0.25% trypsin were purchased from Gibco. DMEM/nutrient mixture F-12 (DMEM/F12) was purchased from HyClone. Antibodies for KDM3B (Cell Signaling and Abcam), ACTIN (Santa Cruz), Ly-6G (Novus) were purchased commercially.

Cell culture

The human bladder epithelial cells SvHUC, HEK-293T cell line were obtained from ATCC. They were cultured in DMEM high glucose media with 5% CO2 in a humidified cell culture incubator. Immortalized lymphocytes were obtained from Coriell Institute and cultured in RPMI 1640 with 15% fetal bovine serum.

Animals

All experimental animals were housed in specific pathogen-free (SPF) conditions. All animal experiments were approved by the University Committee on Animal Resources. For bladder radiation, C57BL/6 male mice between 8–16 weeks old with wt and heterozygous Kdm3b were anesthetized and treated using an XStrahl Small Animal Radiation Research Platform (SARRP; CT image guided 225 kVp X-irradiator, dose rate of 3.1 Gy min⁻¹). The bladder of a supine mouse was identified and delineated by full-body CT scan after retro-orbital injection of 50μl of contrast agent Visipaque. The bladder was then irradiated to full volume using two opposing beams to minimize GI toxicity.

Nucleic acid analysis: for circRNA overexpression:

exon sequence of KDM3B gene was amplified with high-fidelity polymerase and inserted into a lentiviral vector containing the sequence of intron cassette that favors the production of circRNAs [13]. Construct for miRNA expression: Precursor miRNA was cloned into the pol III promoter PLKO lentiviral vector for lentivirus production and expression. 3’UTR luciferase plasmid construction: The whole sequence of target gene 3’UTR containing wt or mut miRNA-responsive elements were cloned into psiCheck2 construct (Promega) downstream of the Renilla luciferase ORF. CircRNA pulldown assay: 5 million cells were collected after trypsin treatment through centrifugation. The cell pellet was lysed in the lysis buffer: 1%NP40, 50mM Tris pH 7.5, 200mM NaCl, 50mM β-glycerophosphate, 10mM NaF, 0.1mM DTT, 0.5mM EGTA, 1x protease inhibitor cocktail. After centrifugation, the supernatant was mixed with 100pmol of biotin-labeled oligo for the junction region (AACCAGCACTGGCTTCATCATGA) overnight at 4°C with rocking. Next, 20μl streptavidin beads were added and incubated for 1 hour followed by 3x washing buffer (0.1%NP-40, 50mM Tris pH8.0, 150mM NaCl). The beads bound nucleic acid was extracted with Trizol followed by miRNA detection. The control was the sample with the omission of biotin-oligo. CRISPR-prime editing: Cells were infected with pLenti-PE2-BSD [14] followed by selection with 0.5μg/ml blasticidin. The cells were then infected with lenti-PLKO vector encoding sequences for prime editing. Two constructs were used, sharing the guide sequence: TTTCTAGAATCCACTGCTTT. #1 construct of PBS (primer binding sequence)-RT (reverse transcription target): CACCATCCTTGTGCAGTGGATTCTA, #2 construct of PBS-RT: CCTCACCATCCTGCAGTGGATTCTA. The cells were selected with 0.5μg/ml puromycin followed by RNA and protein analysis.

Bladder tissue analysis:

Normal tissue portions of bladder cancer resection samples (male, ages 40–70 years old) embedded in paraffin from Xiangya Hospital of Central South University in Changsha, China were collected and the SNP allele determined by genomic sequencing from the FFPE sections using mass spectrometry and Sanger sequencing. Immunohistochemical (IHC) stains were applied on the samples with the genotype of C/C and C/T alleles through standard procedures for KDM3B antibody. Fluorescence in situ hybridization of circ128 was conducted with standard procedures with biotin-labeled oligonucleotide of antisense sequence to the circular junction. Probe hybridization was conducted at 37°C for 24 hours, and signal amplified by Tyramide Signal Amplification method. Immunohistochemical stains and FISH staining intensity were calculated with ImageJ analysis for the intensity measurement for at least over 25 cells to derive a score for each sample, and then these scores were used to derive an average value for each genotype. For gene expression, the mouse bladder epithelial layer was separated from the muscular layer under the dissecting microscope followed by tissue homogenization in Trizol solution and RNA isolation.

Ultrasound scanning of mouse bladders

Ultrasound imaging was performed following voiding sessions at baseline and post-radiation. Isoflurane (~2% in oxygen) was administered during the session to restrict motion during imaging. Each mouse was comfortably secured to a warm temperature-controlled platform with integrated physiologic monitoring capabilities throughout each imaging session. The VEVO 3100 high resolution ultrasound system (Visual Sonics, Toronto, ON, Canada) with the MX550D transducer probe was used with default settings optimized for small animal abdominal imaging, including respiratory gating to eliminate motion artifacts. Three dimensional Bmode and power Doppler images were obtained sequentially with an axial resolution of 40 microns. The Amira-Avizo software was used to construct the 3 dimensional structure of the bladder with manual segmentation of the bladder wall to obtain the doppler signals in the bladder wall.

Quantitative real-time polymerase chain reaction (RT‒PCR)

RNA was prepared by Trizol solubilization and procedures. For qPCR analysis, standard SYBR quantitative methods were used with primers spanning the intron region to increase the specificity for mRNA. The expression was normalized to Tbp, 18s, GAPDH or ACTIN genes where appropriate with reverse transcriptase reactions from equal amount of total RNA after nucleic acid quantitation by a spectrophotometer. Delta-delta Ct was used to determine the relative expression of genes.

Western blotting

Total cell lysates were collected in RIPA buffer followed by BCA protein concentration measurement. The lysate was then prepared with sample loading buffer and heat denaturation. Equal amounts of protein, usually 25μg, were loaded in each lane and analyzed by SDS-PAGE of 8% gel followed by transferring to the PVDF membrane, which was placed into the TBST blocking with overnight primary antibody incubation. Further washing and incubation of appropriate secondary antibody were followed by Biorad gel doc chemiluminescence analysis. Primary antibody for KDM3B (Cell Signaling, C6D12), and loading control with ACTIN or GAPDH where appropriate was used.

Statistical analysis

Data were expressed as mean ± standard deviation (SD) with appropriate replications. Statistical analyses were conducted with the Student’s t test with GraphPad Prism 5 (GraphPad Software, Inc.). P <0.05 is considered statistically significant.

Results

KDM3B rs17599026 C/C genotype has been reported to be associated with increased KDM3B expression at the RNA level in multiple tissue types in the GTEx repository [15], suggesting that this SNP might have a tissue specific impact on KDM3B expression. However, the GTEx repository lacks sufficient bladder tissue samples for mRNA expression analysis. To examine the potential relationship between rs17599026 and KDM3B expression at the protein level in urinary bladder, we obtained normal bladder tissue from surgical specimens from 50 male East Asian patients. Through genotyping and sequencing of rs17599026 in a collection of formalin-fixed tissues, we found 42 were C/C alleles, 8 were C/T alleles, and none were T/T alleles, which is consistent with the frequency of the T allele in the Asian population at approximately 9% (https://www.ncbi.nlm.nih.gov/snp/rs17599026). From these sequenced samples, we selected 10 of the reference alleles and 5 of the heterozygous alleles for IHC analyses using the anti-KDM3B-specific antibody. We found that the expression of KDM3B based on the intensity of staining is higher in samples with the C/C alleles than in the samples of C/T alleles (Fig. 1A). More specifically, the differential KDM3B expression is most pronounced in the bladder epithelial layer without a clear difference in the stroma of the bladder (enlarged portion of Fig. 1A). To further examine the protein expression in the association with three different genotypes, we turned to the immortalized lymphocyte cell lines (GM06984, GM11891, GM12827) from Coriell Institute. Each cell line bore one of the 3 different genotypes of rs17599026 (C/C, C/T, and T/T), and we examined the KDM3B expression by immunoblot analysis. We found that the C/C genotype has the highest expression of KDM3B compared to the other two genotypes (Fig. 1B) in these cells notwithstanding of other background genetic differences. Taken together, these results support a correlation between rs17599026 and KDM3B protein expression. Particularly of interest to us is the expression in the epithelial cell layer of urinary bladder samples from East Asian men.

Figure 1. KDM3B SNP rs17599026 is correlated with KDM3B protein expression.

(A) KDM3B expressions in epithelial layer of normal male human bladder tissues is higher in those of the reference allele (C/C) than the heterozygous alleles (C/T). Normal human bladder tissues from surgical samples of the bladder cancer preserved in formalin were genotyped by mass spectrometry-based genotyping as well as standard Sanger sequencing, and stained with anti-KDM3B antibody with standard IHC procedures. The intensity of the staining per positive cell was measured with ImageJ apps for semi-quantitative quantitation of IHC staining. 10 of the references alleles (C/C) samples and 5 of the heterozygous (C/T) samples were quantitated and plotted. (B) Human immortalized lymphocytes with KDM3B SNP variants have different KDM3B protein expression with the reference allele expressing higher KDM3B protein while the rare variant with the lowest expression. The graph shows the relative protein expression normalized against ACTIN.

To dissect the potential mechanism linking rs17599026 with the KDM3B protein expression, we performed bioinformatic analyses. Among several possibilities such as cryptic splicing on this SNP as well as possible retained intron/lncRNA of KDM3B, we found that rs17599026 is located in the middle of motif (UGCUUU) responsible for binding to MBNL1 RNA binding protein, and likely the change of C to T will abolish the binding of MBNL1 to the canonical sequence upon a change of nucleotide in the consensus binding motif (http://rbpmap.technion.ac.il/index.html). This finding also leads to the natural connection of the role of MBNL1 in the expression of circRNAs, as MBNL1-mediated circular RNA formation is one of the major pathways for circRNA expression [12]. Therefore, it is plausible that this SNP is correlated with circRNA expression, resulting in differential KDM3B protein expression. To test this hypothesis, we examined the expression of 5 circular RNAs that share the SNP-containing intron with different splicing donors in the immortalized lymphocytes. As shown in Figure 2B, expression of circ128 and circ129 is correlated with the KDM3B SNP; specifically, the C/C genotype leads to higher expression, while the T/T genotype results in lowest expression, consistent with the likely role of the MBNL1 binding motif for the expression of circ RNA.

Figure 2. KDM3B SNP rs17599026 is correlated with circular RNA expression.

(A) KDM3B SNP is located in the middle of MBNL1 binding motif that is potentially related to the expression of five circular RNAs of KDM3B that is likely impacted by this motif. (B) Expression of these five circRNAs were measured in the immortalized lymphocytes with the defined KDM3B SNP alleles. The expression of each circRNA in the C/C alleles was defined as unit 1 while all others were calculated against this level after normalization against 18s level. (C) KDM3B protein and circ128 and circ129 were determined in a pair of SvHUC cells (C/C genotype) after destruction of the MBNL1 binding sites through removal of 3 nucleotides through CRISPR-dCas9 prime editing to mimic the KDM3B TT alleles. (D) KDM3B circRNA128 expression was higher in C/C samples than in C/T samples through fluorescence in situ hybridization using the probe that is specific for the junction of the circRNA sequence. The fluorescence signals per positive cell from 4 samples of the two genotypes were plotted.

To directly implicate the MBNL1 binding motif for the expression of circ128 and circ129, we turned to the immortalized male human bladder epithelial cell line SvHUC. Sequencing of the genomic DNA of SvHUC cells indicates that they have the C/C genotype, which is expected given that this is the most common of the three possible genotypes. We utilized CRISPR-dCas9 prime editing [16] to modify the MBNL1 motif with a deletion of 3 uridines to mimic the T/T SNP as the PAM sequence precluded the precise change of C to T transition. Indeed, in cells with the expression of prime-editing RNA for the destruction of MBNL1 binding site, there is a corresponding decrease of both circ128 and circ129 with a simultaneous decrease of KDM3B protein expression (Fig. 2C). Two sequences were used for this prime editing analysis to increase the specificity of the genome engineering, both of which showed similar impact on the circRNA and KDM3B expression. We chose the #2 construct for the remainder of the experiments, as circ128 has a predicted MBNL1 binding site on the potential donor splicing position, thus likely contributing to the complex formation of MBNL1 to promote the circRNA genesis. Therefore, we will focus on this circRNA only. Indeed, fluorescence in situ hybridization with the probe specific for the circRNA junction sequence resulted in higher signals in C/C samples compared with the C/T samples (Fig. 2D), supporting the mechanistic connection between circular RNA expression and KDM3B protein expression. Taken together, these results indicate that rs17599026 regulates KDM3B protein expression through regulating circRNA expression via the binding motif of RBP MBNL1.

To investigate the molecular mechanisms responsible for the regulation of KDM3B protein expression by the circ128, we turned to the miRNA sponge mechanism from among many possibilities including regulating gene transcription and encoding peptides. In particular, several reports have characterized miRNAs that can be sponged by circRNAs to impact the host gene protein expression [17]. We tested this possibility by searching potential miRNAs that could bind to circ128 as well as 3’UTR of KDM3B gene. The potential miRNAs would likely bind to circ128 that could be identified through pulldown of circ128 by biotin-conjugated oligonucleotide specific for the junction sequence of the circRNA to reduce non-specific interactions. Among the 8 candidate miRNAs, miRNA-760 and −378g showed increased binding to the streptavidin beads than to the control pulldown (Fig. 3A). Indeed, when we reintroduced circ128 into SvHUC cells containing T/T-like genotype, KDM3B protein increased to the level in control cells, and more significantly a mutant circ128 (mcirc128) deficient in binding to two miRNAs lacked the ability to rescue the impact of T/T genotype for KDM3B protein expression (Fig. 3B), supporting the role of miRNA-760 and −378g in mediating the effect of circ128 expression in regulating KDM3B protein expression. Moreover, we verified the role of these two miRNAs through exogenous expression of both miRNAs. Consistent with the role of miRNA in suppressing mRNA translation, expression of miRNA-760 or −378g significantly reduced the KDM3B expression in normal bladder epithelial cells (Fig. 3C). To directly implicate the role of miRNA-760 and −378g in regulating KDM3B protein expression, we also generated two reporter constructs bearing the wildtype KDM3B 3’UTR and a deletion mutant lacking the miRNA binding sequences due to their close localization in the 3’UTR. The reporter expression showed a significant reduction in response to both miRNA-760 and −378g with the wildtype sequence, but with a significantly reduced impact on the deletion mutant (Fig. 3D). These experimental results established a direct connection between miRNA-760 and −378g in regulating KDM3B protein expression as well as their reduced capacity in doing so after being sponged by circ128 (Fig. 3E).

Figure 3. KDM3B circRNA128 can sponge miRNAs to impact KDM3B protein expression.

(A) CircRNA128 likely sponges miRNA-378g and miRNA-760. MiRNAs that likely bind to both circRNA128 and KDM3B 3’UTR were measured in the circ128 pulled down by the oligonucleotide specific for the junction sequence of the circ128 followed by qPCR analysis. (B) Overexpression of circ128 can rescue the reduced KDM3B expression in SvHUC cells with the disruption of the MBNL1 binding site to mimic the KDM3B SNP while a mutant circ128 that do not bind to the two miRNAs fails to do so. (C) Overexpression of miRNA-378g and miRNA-760 suppress KDM3B expression in bladder epithelial cells. The overexpression of the two miRNAs is demonstrated in the plot on the right. (D) MiRNA-378g and miRNA-760 can suppress KDM3B 3’UTR reporters containing the wildtype binding sequence but much less so in the mutant reporter lacking the corresponding target sequence in SvHUC cells. The normalized luciferase activity is reduced upon expression of the two miRNAs. (E) A mechanistic diagram illustrates the different amount of KDM3B circ128 that likely impacts the KDM3B protein expression through miRNA-mediated translation suppression.

To characterize the role of KDM3B protein in mediating radiation toxicity, we turned to an animal model with a heterozygous deletion of the Kdm3b gene [18] to mimic the likely reduced KDM3B expression in patients with the T/T genotype. Indeed Kdm3B expression in the bladder was reduced in the heterozygous mouse (supplemental Fig. 1). As rs17599026 was associated with bladder toxicity in the clinical GWAS, as proof of principle we focally delivered a single dose of X-rays to 100% full volume of the mouse bladder under CT-image guidance using two opposing collimated beams on a small animal radiation research platform (Xstrahl, SARRP). Whole bladder irradiation allowed bladder toxicities to be measured in our animal model. We remain cognizant of the effects of bladder irradiation compared with prostate targeting, e.g., the human prostate surrounds the urethra, therefore directly impacting urination, while the mouse prostate is anatomically separate from the urethra and bladder and thus much less impacting urination. On the other hand, RT for prostate cancer patients invariably induces off-target effects in the bladder due to their close anatomical proximity. We therefore monitored urination in mice before and after bladder radiation. Mice were acclimated to reverse lighting cycle to ensure a more active daily period during urine collection, then urination was evaluated over a four-hour period by assessing urine spots on absorbent filter paper placed underneath circular wire holding cages. Each mouse was monitored three times before radiation to establish an individualized urination baseline followed by different time points after the bladder irradiation. As shown in Fig. 4A, we found that Kdm3b heterozygous (het) mice had increased urination frequency compared to the wildtype mice after 24Gy radiation 7 days after treatment. No differences in urination were seen after 8Gy bladder irradiation over the 28 day time frame (data not shown). Therefore, the differential effect appears to be dose dependent, or a longer time frame is required to observe a difference. At the same time, as wildtype mice and Kdm3b het mice have a significant (2 fold) difference in Kdm3b expression, a clear difference in urine frequency at day 7 post radiation is possible, modeling the late toxicity in terms of urine frequency in humans with KDM3B SNPs echoing the finding that acute radiation toxicity is significantly predictive of late sequelae [19].

Figure 4. Mice with wildtype and heterozygous Kdm3b alleles have different responses upon bladder radiation.

(A) Increased urination frequency in heterozygous male mice upon stereotactic radiation to the bladder through SARRP. The urination pattern of individual mouse (n=3/group) was recorded in a circular filter paper in a span of 4 hours in a circular cage and was analyzed by the ImageJ Void Whizzard app. * P<0.05. (B) Expression of groups of genes related to inflammation was measured in the bladder epithelial layer of the mouse bladder. The inner bladder epithelial layer was dissected under microscope away from muscular layer followed by Trizol RNA preparation. The values were calculated based on the combined biological replicates of the qPCR reactions from 2 wildtype mice and 2 heterozygous mice. (C) Increased neutrophil filtration in the bladder stromal area in the Kdm3b heterozygous mouse 7 days after bladder radiation of 24Gy. N=2 for each group with each sample having two sections thus 4 samples for each group. **P<0.01. (D) Ultrasound analysis with the doppler signal detection reflecting the blood flow in the microvessels in the bladder walls. (E) Quantitative analysis of the doppler signal intensity of the bladder walls of the mice before and after 24Gy radiation to the bladder. N=4 for each group. * P<0.05.

As radiation damage to the bladder is complex, involving processes besides cell death, we examined the other known effects of radiation, such as inflammatory responses induced by ionizing radiation. In particular, we measured a panel of genes that have been well established to be connected with inflammation by various stress signals including ionizing radiation. As shown in Figure 4B, the gene expression pattern in some of them did not show much difference between wildtype and heterozygous deletion mice while expression of Gmcsf, a cytokine that responds to stress such as radiation, showed a dramatic increase in the heterozygous mice compared with wildtype mice, suggesting that a differential inflammatory response between these mice might contribute to the difference of urination phenotype between wildtype and heterozygous animals. Indeed, IHC analysis of the bladder demonstrated a clear increase of Ly6G staining, a marker for neutrophils as well as other lymphocytes, in the bladder stroma of heterozygous mice compared to that in the wildtype mice (Fig. 4C), indicative of a positive inflammatory response. To characterize tissue inflammation in a non-invasive manner, we also used ultrasound imaging with doppler signal analysis for fluid movement in the blood vessels as a gauge of inflammation-induced vasodilation [20]. Through 3D reconstruction of ultrasound imaging (Fig. 4D), a precise measurement of overall doppler signals demonstrated an increase of these signals in the bladder wall of heterozygous mice than in the wildtype mice. Quantitation of these signals in a times series before and after radiation indicated that this increase has rapid onset in the heterozygous mice while wildtype mice did not show any increase in doppler signals (Fig. 4D). It is worth noting that there is close temporal correlation between urination frequency and doppler signals-indicated tissue inflammatory response, suggesting their likely causal relationship between these two phenotypes. These results thus established that Kdm3b heterozygous deletion animals have a more pronounced radiation-induced bladder tissue inflammatory response, likely leading to a change of urination frequency, a significant aspect of radiation toxicity associated with prostate cancer RT.

Discussion

RT is a widely used cancer treatment modality. The efficacy of RT depends on tumor sensitivity and normal tissue tolerance, which can cause acute and late toxicities [21]. The therapeutic ratio of RT is increased by enhancing tumor sensitivity while reducing normal tissue toxicity, which may be achieved by adjusting radiation type, dose, frequency, and volume. Recent literature suggests that RT can activate the body’s immune system to induce cytotoxicity, which contributes to its efficacy [22, 23]. Normal tissue toxicities can also be modulated by immune response towards dead cell debris and cytokine secretions, leading to fibrosis and other complications [24, 25]. Understanding the molecular processes related to radiation toxicity can help develop effective therapeutic interventions that protect normal tissue function without compromising radiotherapy efficacy. Proper protection of normal tissue function can increase the dose of radiation and make radiotherapy more effective.

A GWAS study identified SNPs tagging KDM3B associated with late bladder toxicity, specifically increased urinary frequency, following radiation to treat prostate cancer [7]. Two fundamental questions need to be addressed before mechanism-based intervention can be established. The first is if and how this genomic locus affects KDM3B expression, and the second is how molecular mechanisms involving KDM3B affect the physiological process of urination. Our experimental results provide evidence demonstrating that KDM3B SNP through regulating ncRNA, specifically in regulating circRNA expression via sponging miRNAs, can impact KDM3B protein expression in cells, thus establishing a direct connection between this SNP (rs17599026) with KDM3B protein expression. Moreover, different KDM3B expression levels might lead to different degrees of radiation-induced tissue inflammatory response, therefore leading to varying degrees of physiological impact on urination frequency. Our study also provided a potential missing link between genetic association and regulatory function by expanding the connection to the post-transcriptional regulation of the affected gene beyond the conventional linkage at the transcriptional level [26].

One caveat inherent in our analysis is that the animal model we used, namely the pair of mice with normal and heterozygous Kdm3b genes, might not represent the exact scenario of human KDM3B expression influenced by the SNP, an intrinsic limitation of any animal model. It is anticipated that in humans the differences in KDM3B expression will have more nuances and the impact will be less than the loss of half of the gene expression. On the other hand, as circRNAs likely have additional functions that remain to be characterized, there might be functions associated not with KDM3B per se, but through functions of circRNA with graded expression. The exact details of these additional phenotypes remain to be characterized. The other potential difference is the temporal span of the radiation toxicity in patients vs. animal models. While in humans there is universal impact on genitourinary functions during the course of RT, the late toxicities did reveal individual differences associated with SNPs from GWAS studies, lending the genetic influence on these phenotypes. In contrast, mice of wildtype and heterozygous Kdm3b genes showed a more rapid onset of differences in urinary frequency, inflammation-related cytokine gene expression, and vasodilation after irradiation. These differences likely were the result of combined differences between human clinical situation and animal modeling. One aspect in particular is the anatomical relationship between prostate and bladder in that the prostate surrounds the urethra in humans, but there is no physical integration between these two organs in mice. In addition, the total dosage of radiation is different in that human patients receive a much higher fractionated total radiation dose for prostate cancer, while in animals the experimental radiation is targeted at the bladder with a single dose. Moreover, in humans late toxicity likely reflects a difference in recovery from radiation-induced stress/inflammation influenced by different genetic background, while in mice the genetic background is more homogeneous. In mice, there appeared to be a clear difference in urinary frequency at 7 days after irradiation but less difference between wildtype and heterozygous mice beyond 2 weeks. We also acknowledge that more biological replicates would provide stronger statistical power for the experimental observation. Our findings are consistent with previous reports that normal mice did not show an immediate response toward radiation on the bladder while genetic background could have a profound impact on mouse urination behavior [27, 28]. Nevertheless, the findings in the our experimental animals showing a rapid difference between the two genotypes are relevant for understanding the urinary toxicity in humans as radiation-induced inflammation can have a lasting impact [29]. Overall, these experimental data provided the evidence that a SNP-regulated differential KDM3B protein expression might mediate radiation-induced tissue inflammatory response.

Mechanistic understanding of complex human pathophysiological processes of normal tissue response can reveal molecular targets for precision medicine in alleviating clinical symptoms. Identification of SNPs through GWAS analyses in humans offers unique opportunities to examine its potential functional role in cancer patients and provide advantages to the understanding of the potential molecular mechanisms as the basis for pharmaceutical interventions. In the case of the KDM3B SNP for late urinary toxicity after RT, our work provides the first mechanistic investigation in connecting the SNP rs17599206 of GWAS to the function of KDM3B, and correlates with the manifestation of urinary bladder injuries by RT. Our findings suggest that enhancing the protein function of KDM3B through its expression or activity may result in reducing adverse urinary symptoms for prostate cancer patients by minimizing symptoms from radiation bladder injuries after RT in the future.