Intestinal tumor suppression in ApcMin/+ mice by prostaglandin D2 receptor PTGDR

Our earlier work showed that knockout of hematopoietic prostaglandin D synthase (HPGDS, an enzyme that produces prostaglandin D2) caused more adenomas in ApcMin/+ mice. Conversely, highly expressed transgenic HPGDS allowed fewer tumors. Prostaglandin D2 (PGD2) binds to the prostaglandin D2 receptor known as PTGDR (or DP1). PGD2 metabolites bind to peroxisome proliferator-activated receptor γ (PPARG). We hypothesized that Ptgdr or Pparg knockouts may raise numbers of tumors, if these receptors take part in tumor suppression by PGD2. To assess, we produced ApcMin/+ mice with and without Ptgdr knockouts (147 mice). In separate experiments, we produced ApcMin/+ mice expressing transgenic lipocalin-type prostaglandin D synthase (PTGDS), with and without heterozygous Pparg knockouts (104 mice). Homozygous Ptgdr knockouts raised total numbers of tumors by 30–40% at 6 and 14 weeks. Colon tumors were not affected. Heterozygous Pparg knockouts alone did not affect tumor numbers in ApcMin/+ mice. As mentioned above, our Pparg knockout assessment also included mice with highly expressed PTGDS transgenes. ApcMin/+ mice with transgenic PTGDS had fewer large adenomas (63% of control) and lower levels of v-myc avian myelocytomatosis viral oncogene homolog (MYC) mRNA in the colon. Heterozygous Pparg knockouts appeared to blunt the tumor-suppressing effect of transgenic PTGDS. However, tumor suppression by PGD2 was more clearly mediated by receptor PTGDR in our experiments. The suppression mechanism did not appear to involve changes in microvessel density or slower proliferation of tumor cells. The data support a role for PGD2 signals acting through PTGDR in suppression of intestinal tumors.

Lewis lung cancer cells implanted onto the backs of mice lacking the PGD 2 receptor (PTGDR, also known as DP1), grew faster than tumors implanted onto wild-type mice [6]. Furthermore, the PTGDR agonist, BW245C, reduced tumor growth. These results support a role for PGD 2 itself.
Here, we show that knockouts of Ptgdr increased tumor numbers in Apc Min/+ mice, indicating that PGD 2 and PTGDR act to suppress tumors. PPARG had smaller effects in our experiments.

Mice
The protocol and procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Los Angeles Biomedical Research Institute. C57BL/6, FVB/N, and Apc Min/+ (C57BL/6; no. 002020) mice came from Jackson (Bar Harbor, ME), as did mice carrying the Cre transgene controlled by the adenovirus EIIa promoter [Tg(EIIa-Cre) C5379Lmgd/J; FVB/N strain; no. 003314] [9]. Mice in which exon 2 of the Pparg gene is flanked by loxP sites were from F. Gonzalez (Pparg flox/flox FVB/N mice).
Our PTGDS transgenic mice (line B20; FVB/N) overexpress human PTGDS in all tissues [11]. Reported basal brain levels of PGD 2 were 1.5-fold higher than wildtype levels and rose fivefold upon stimulation. PGE 2 levels did not change. The mice had more eosinophilia in a bronchial asthma model, compared to HPGDS transgenic mice [12].

Intestinal histopathology and definitions of tumor sizes
Adenomas were counted histologically at 6 or 14 weeks, without knowing genotypes [5]. We used 24 Swiss roll sections spaced 150 lm apart for PTGDS transgenic mice, Pparg knockout mice, and their controls. We used 10 Swiss roll sections (250 lm apart) for Ptgdr knockout mice and their controls. Tumors sizes were gauged by the number of sections spanned. Small tumors were defined as those seen in only 1 section. Large tumors were those with profiles in multiple sections. Mitotic figures were identified as described [14].

Statistical analyses of tumor data
Tumor data were analyzed by nonparametric methods (Kruskal-Wallis and Mann-Whitney), because numbers of tumors per mouse did not follow a Gaussian distribution. We analyzed total, small, large, and colon tumors. We also calculated ratios of the geometric mean number of tumors in genetically modified mice to the geometric mean number in controls. Ratios were estimated from differences in logarithm-transformed tumor numbers. For the colon, we added 0.5 to all numbers of tumors before taking logarithms, to handle zero values. Data from 6-and 14-week-old mice were analyzed separately. These statistical methods were also used to reanalyze tumor data from Apc Min/+ mice with transgenic HPGDS (and controls) from earlier work [5].

In situ hybridization
Digoxigenin-labeled probes were prepared by in vitro transcription from a linearized plasmid vector containing the mouse PTGDR cDNA (DIG RNA labeling kit; Roche; Indianapolis, IN). T7 RNA polymerase was used to make anti-sense probes. SP6 RNA polymerase was used to prepare control sense probes [16].

mRNA analyses by reverse transcription and real-time PCR (RT-PCR)
Primers, probes, and procedures for preparing RNA and determining copy numbers of RNA transcripts are in Supporting Information. Assays for v-myc avian myelocytomatosis viral oncogene homolog (MYC), GAPDH, and vascular endothelial growth factor A (VEGFA) were performed with kits (Applied Biosystems; Grand Island, NY; Mm00487803_m1, Mm99999915_g1, Mm00437304_m1, respectively).

Tumor scoring
We histologically examined >35,000 tumors in Swiss roll sections ( Fig. 1A-E), including 9837 tumors from 147 mice in Ptgdr knockout experiments, 21,763 tumors from 104 mice in experiments on PTGDS transgenic and Pparg knockout mice, and 3431 tumors reexamined from 39 HPGDS transgenic mice and controls from earlier work [5].
Intravillar tumors progressed by enlarging, forming adjoining cysts (Figs. S1C and S2), or erupting through the villus surface into the bowel lumen (Figs. S1B and S3). Although early tumors arise from crypts [17,22], we found only a few examples of out-pouching of cysts from crypts ( Fig. S4). Serial sections from two tumors (75 sections each) showed that early tumors may have no crypt connection (Fig. S5) [21]. Examples of early colon tumors are shown in Fig. S6.
To obtain data on occurrence of the earliest tumors, we also scored tumors in ten 3-week-old mice: six Apc Min/+ mice (3-8 tumors each); three Apc Min/+ mice with heterozygous Ptgdr knockouts (5-11 tumors each); and one Apc Min/+ mouse with homozygous Ptgdr knockouts (11 tumors). However, data from these 10 mice were not included in statistical analyses, because of the age difference.
In situ hybridization for PTGDR mRNA showed consistent, but weak, staining of inflammatory cells in the mucosal stroma (lymphocytes or monocytes, or both; Fig. 1F-G). There was no detectable staining in epithelial cells of crypts or villi. Staining with PTGDR antibodies was not conclusive (not shown).

Expression of transgenic PTGDS
Human PTGDS transgenes were highly expressed in the intestines, as measured RT-PCR. Specifically, we found 1.61 9 10 5 and 8.13 9 10 5 copies of human PTGDS transcripts per nanogram of total RNA in two transgenic mice (geometric mean, 3.6 9 10 5 copies). These values were comparable to levels for HPGDS transgenes in earlier work (7.5 9 10 5 copies-a 375-fold increase in expression of transgenic HPGDS over endogenous mouse Hpgds) [5]. Immunohistochemistry showed heavy staining of transgenic PTGDS in all intestinal cells ( Fig. 1H-I). Endogenous mouse PTGDS mRNA was not detectable in the colon.

Transgenic PTGDS and large tumors
With 104 Apc Min/+ mice, we scored intestinal tumors in relation to transgenic PTGDS, with and without heterozygous Pparg knockouts. Among mice without Pparg knockouts, only large tumors were reduced in number by transgenic PTGDS (medians were 52 vs. 83 for controls; Table S3). Tumor suppression was also reflected by the ratio of the geometric mean number of large tumors in PTGDS transgenic mice to the geometric mean number in controls (ratio = 0.56 for large tumors; 95% confidence interval 0.34-0.92; Table S3C). Large tumors were >150-300 lm in diameter, based on the spacing between sections.
We measured colon mRNA levels for VEGFA and MYC, relative to endogenous GAPDH transcript levels. PTGDS transgenes lowered median levels of MYC and VEGFA transcripts by 50% in Apc Min/+ mice (Fig. S7).  Heterozygotic Pparg knockouts and transgenic PTGDS Without PTGDS transgenes, the numbers of tumors in heterozygotic Pparg knockout mice were comparable to numbers in mice without Pparg knockouts ( Fig. 3; Table  S3; see "Control"). Thus, heterozygous Pparg knockouts alone did not increase tumors in Apc Min/+ mice.
On the other hand, Apc Min/+ mice with both transgenic PTGDS and heterozygotic Pparg knockouts had intermediate numbers of large tumors. Specifically, going by medians, there were 52 large tumors in mice with PTGDS transgenes alone, 88 in mice with heterozygotic Pparg knockouts alone, and 70 in mice with both mutations (Table S3C). Similarly, the ratio of the mean number of large tumors in PTGDS transgenic mice to the mean number in controls was 0.56 for mice without heterozygotic Pparg knockouts (95% confidence interval, 0.34-0.92), compared to 0.78 for mice with heterozygotic Pparg knockouts (95% confidence interval, 0.48-1.26).

PTGDS versus HPGDS
As mentioned above, RT-PCR showed similar expression of transgenic PTGDS, compared to transgenic HPGDS (as measured in our previous work) [5]. Also, immunohistochemistry showed high levels of PTGDS and HPGDS (Fig. 1H-K). Both experiments scored tumors in the same way (24 Swiss roll sections; 150 lm between sections). Therefore, we reanalyzed slides from HPGDS transgenic mice from our first report [5] to directly compare PTGDS to HPGDS (Fig. S8; Table S4). Ratios of the mean total number of tumors in transgenic mice to the mean total in controls were 0.70 for PTGDS, compared to 0.28 for HPGDS (Tables S3A and S4A). Thus, HPGDS may be two times stronger than PTGDS in suppressing tumors.
We assessed tumor cell proliferation in relation to HPGDS transgenes, by use of immunohistochemistry with anti-PCNA antibodies. Again, we used slides from our earlier work on Apc Min/+ mice with HPGDS transgenes [5]. We focused on intravillar tumors, because they are fairly uniform in size. There was no difference in PCNA staining in intravillar tumors in HPGDS transgenic versus nontransgenic Apc Min/+ mice (Fig. 1L-M Immunohistochemistry with anti-CD31 antibodies showed no consistent difference in microvessel staining between HPGDS transgenic and nontransgenic tumors (Fig. 1N-O). Thus, microvessel growth does not appear to explain occurrence of fewer tumors with PGD 2 .

Tumors in eight mutants with altered PGD 2 synthesis or binding
We have now analyzed tumors in eight different Apc Min/+ mouse mutants that have altered PGD 2 production or binding, due to knockouts or transgenes. Some experiments used different procedures for cutting sections. For example, we used up to 24 Swiss roll sections for scoring tumors in our first report [5] and in the PTGDS and PPARG experiments shown here. Alternatively, we used 10 sections per Swiss roll in the PTGDR experiments, because reanalysis of earlier data showed that the same conclusions can be reached with 8-10 sections.
To compare data across experiments, we converted the total number of tumors for each mouse to a "multiple of the median" value. Specifically, we divided the total number of tumors for each mouse by the median number of tumors for that mouse's controls. By this analysis, the most tumor-promoting mutations were Hpgds knockouts and homozygous Ptgdr knockouts-raising tumor numbers 40% above control values ( Fig. 4; all mice were analyzed at 14 weeks). In contrast, HPGDS transgenes were the most tumor-suppressing mutations-reducing tumor numbers to 20-30% of the control value.

Female versus male Apc Min/+ mice
To assess female-male differences in tumor numbers at 14 weeks, we used current data and two earlier reports [5,23] (for a total of 61 female and 75 male Apc Min/+ mice; Fig. S9). Males and females had similar numbers of intestinal tumors (ratio of tumors in males vs. females, 0.82; P = 0.069). However, males had more colon tumors (ratio of colon tumors in males vs. females, 1.6; P = 0.0002). Results are consistent with McAlpine et al. [24].

PTGDR and intestinal tumors
Homozygous deletion of the gene for PGD 2 receptor PTGDR led to 30-40% more intestinal tumors in Apc Min/+ mice. The result supports an interpretation that PTGDR mediates tumor inhibition by PGD 2 in these mice. We now have data on eight different Apc Min/+ mouse mutants, each with a different alteration in PGD 2 production or binding. Homozygous Ptgdr and homozygous or heterozygous Hpgds knockout mutations are the most pro-tumorigenic. On the other hand, HPGDS transgenes are the most tumor-suppressive mutations-lowering numbers of tumors by 70-80% (Fig. 4). There was no detectable staining of PTGDR mRNA in the epithelium of intestinal crypts or villi. However, PTGDR mRNA was consistently detected in inflammatory cells in the mucosal stroma (Fig. 1F-G). Tissue-specific gene knockouts will be needed to more conclusively identify the cells that respond to PGD 2 .
Mutoh et al. [25] treated homozygous Ptgdr knockout mice with azoxymethane starting at 7 weeks and examined colons at 12 weeks. They did not find more aberrant crypt foci in the colons of knockout mice, compared to controls. Our results are consistent with Mutoh et al., because we did not observe more colon tumors at 6 or 14 weeks with Ptgdr knockouts ( Fig. 2A-B). However, a role for PTGDR in colon tumor growth is supported by human data. Gustafsson et al. [26] found fivefold lower expression of PTGDR in colorectal cancers, compared to normal tissues (62 tumors and 43 normal tissues, from 62 patients). Galamb et al. [27] showed a trend toward decreased PTGDR expression going from normal tissues, to adenomas, to early cancers, and to advanced cancers.

Comparison of PTGDS and HPGDS effects
Transgenic PTGDS in Apc Min/+ mice reduced numbers of large adenomas (>150-300 lm; Fig. 3C; Table S3C). In this way, PTGDS had a tumor blocking effect. However, transgenic PTGDS was less effective than transgenic HPGDS in suppressing tumors (Fig. 4). Reasons are unknown. A difference between PTGDS and HPGDS is secretion of PTGDS into body fluids, whereas HPGDS stays in the cytosol [28]. We recognize that our comparison between PTGDS and HPGDS is based on only one transgenic mouse line for each mutant. But these lines had comparable numbers of PTGDS or HPGDS mRNA transcripts in the intestines (3.6 9 10 5 and 7.5 9 10 5 copies, respectively).
Transgenic PTGDS was associated with lower colon expression of MYC (Fig. S7). MYC is a major part of WNT signaling following Apc loss [29]. Moreover, disruption of Myc restores the normal appearance of intestinal crypts in mice with intestine-specific Apc knockouts [30]. Thus, lower intestinal levels of MYC may be part of the tumor preventive mechanism of PGD 2 .
Reduced levels of VEGFA mRNA were also seen in PTGDS transgenic mice (Fig. S7). The finding is consistent with VEGFA effects in Apc mice [31]. However, we did not see a decrease in microvessel density in large tumors in Apc Min/+ mice with transgenic HPGDS (Fig. 1N-O). Thus, tumor suppression by PGD 2 did not appear to involve antiangiogenesis in our experiments [32], at a level detectable by anti-CD31 immunohistochemistry.
Transgenic HPGDS did not reduce PCNA immunostaining in early tumors in Apc Min/+ mice (Fig. 1L-M). PCNA is a marker of intestinal cell proliferation and belongs to the family of sliding DNA clamps that bind factors at replication forks [33]. Similarly, transgenic HPGDS did not lower numbers of mitotic figures in early tumors. Thus, PGD 2 does not appear to suppress tumors by lowering rates of tumor cell division.
A possible explanation is increased tumor cell death with PGD 2 and PTGDR, as shown by Lewis lung cancer cells implanted onto the backs of mice [6]. Alternatively, PGD 2 may prevent tumors by slowing initiation.

PPARG and intestinal tumors
Heterozygous Pparg knockouts alone did not increase the numbers of tumors in our Apc Min/+ mice. The result is consistent with earlier reports [24,34]. However, McAlpine et al. [24] found~30% more tumors in male Apc Min/+ mice with heterozygous or homozygous intestine-specific Pparg deletions.
In our transgenic mice with PTGDS overproduction and reduced adenoma occurrence, the decrease in numbers of large tumors caused by PTGDS appeared blunted in heterozygous Pparg knockout mice (Fig. 3C and Table S3C). Such blunting could be compatible with tumor suppression by PGD 2 metabolites bound to PPARG [35], when PGD 2 production is increased. A limitation in our experiments with heterozygous Pparg knockouts and PTGDS transgenes was the use of mice with mixed C57BL/6-FVB/N backgrounds (all 50% C57BL/6, but not all F 1 ). However, fairly large numbers of mice were used in the Pparg experiments (104 in total). The 147 mice in the Ptgdr knockout experiments were all 100% C57BL/6.

PGD 2 and inflammation
Mechanisms for tumor suppression by PGD 2 in the intestines have not been proven, but useful information is available. For example, in the skin [36] and lung [37], PGD 2 delays migration of dendritic cells to draining lymph nodes, where T cells are primed. PGD 2 also reduces the ability of dendritic cells to stimulate na€ ıve T cells [38,39]. In the intestinal mucosa, dendritic cells produce IL-23, to stimulate release of IL-22 by immune cells (innate lymphoid cells [40,41], T H 17 cells [42], and T H 22 cells [43]). In turn, IL-22 induces proliferation of epithelial cells, production of inflammatory mediators, and release of antimicrobial proteins, to guard against invaders [44]. This cytokine can be neutralized by IL-22-binding protein, a soluble receptor also made by dendritic cells in the colon. Huber et al. [45] showed that IL-22 gene knockouts allowed fewer tumors in Apc Min/+ mice, whereas knockouts of IL-22-binding protein caused more (in the colon). Further work is needed to determine if these functions of dendritic cells explain tumor suppression by PGD 2 . Identification of mechanisms involving PGD 2 and PTGDR may suggest molecular targets for tumor prevention studies.

Conclusions
By scoring tumors in Apc Min/+ mice histologically at 6 and 14 weeks, we showed that homozygous knockouts of the gene for the PGD 2 receptor, PTGDR, raised median numbers of tumors by 30-40%. The results support an interpretation that PGD 2 is a tumor-suppressing molecule, acting through PTGDR. Heterozygous Pparg knockouts had smaller effects in our experiments. The observation that PGD 2 and PTGDR can affect tumorigenesis may have impact for prevention.    Figure S4. Three examples of intravillar tumors that show a connection to a normal crypt. Scale bar, 100 lm. Figure S5. Sections of an intravillar neoplasm in the small bowel of an Apc Min/+ mouse, showing a uniglandular, intravillar lesion with a simple cystic structure. Although tumors arise from crypt cells, we did not observe a connection between the cystic structure and the crypt for this tumor. Thus, early tumors may become fully enclosed or "sealed off." All mounted sections containing profiles for this tumor are shown here. Scale bar, 100 lm. Figure S6. Examples of colon tumors seen at 6 weeks. Tumors at this age are typically small and lie below the mucosal surface. They would be overlooked without histological examination. The inset in A shows a higher magnification view of the tumor. Scale bar, 100 lm (applies to all panels, except the inset). Figure S7. Lower expression of VEGFA and MYC in the colon of Apc Min/+ mice with PTGDS transgenes (TG) and without (WT). mRNA was prepared from colon tissue, and expression levels for VEGFA and MYC were quantitated relative to endogenous mouse GAPDH. Plotted points are averages of triplicate measurements in different mice. VEGFA expression in PTGDS transgenic mice was approximately 50% of expression in controls (P = 0.022, Mann-Whitney; P = 0.012, t-test). MYC expression was also 50% lower in PTGDS transgenic mice (P = 0.041, Mann-Whitney; P = 0.050, t-test). Filled symbols: females. Open symbols: males. Horizontal bars show medians. *P < 0.05. Figure S8. Numbers of adenomas in Apc Min/+ mice with HPGDS transgenes (TG) and without (WT). Transgenic HPGDS was associated with statistically significant reductions in the numbers of tumors in all size categories. See Table S4 for median values, ranges, numbers of mice, Pvalues, and ratios of numbers of tumors in HPGDS transgenic mice to numbers in controls. Filled symbols: females. Open symbols: males. Horizontal bars indicate medians. *P < 0.05. Figure S9. Tumor development in female and male Apc Min/+ mice at 14 weeks. We combined data from the current experiments with data from two earlier reports [5,23]. For each mouse, numbers of tumors (in the entire intestine and in the colon) were normalized to the median number among females in the same experiment.
For colon tumors, we added 0.5 to the number of tumors before taking the median. Horizontal bars indicate medians. The dotted horizontal lines indicate 1.0 (which is the median value for females). Males (77 mice) and females (61 mice) tended to have similar numbers of total tumors throughout the intestine (median ratio for males to females = 0.82; P = 0.069; A), but males had roughly 60% more colon tumors, compared to females (median ratio for males to females = 1.6; P = 0.0002; B). Table S1. Adenomas at 6 weeks in Apc Min/+ mice with Ptgdr knockouts. Table S2. Adenomas at 14 weeks in Apc Min/+ mice with Ptgdr knockouts. Table S3. Adenomas in Apc Min/+ mice with PTGDS transgenes, with and without heterozygous Pparg knockouts. Table S4. Adenomas in Apc Min/+ mice with HPGDS transgenes.