Absence of myeloid Klf4 reduces prostate cancer growth with pro-atherosclerotic activation of tumor myeloid cells and infiltration of CD8 T cells

The microenvironment of prostate cancer often includes abundant tumor-associated macrophages (TAMs), with their acquisition of an M2 phenotype correlating with local aggressiveness and metastasis. Tumor-derived M-CSF contributes to TAM M2 polarization, and M-CSF receptor inhibition slows prostate cancer growth in model systems. As additional cytokines can direct TAM M2 polarization, targeting downstream transcription factors could avoid resistance. Klf4 and C/EBPβ each contribute to monocyte development, and reduced expression of macrophage Klf4 or C/EBPβ favors their adoption of a pro-inflammatory M1 state. We find that a Hi-Myc C57BL/6 prostate cancer line grows more slowly in syngeneic Klf4(f/f);Lys-Cre compared with Klf4(f/f) mice when inoculated subcutaneously, but grows equally rapidly in C/EBPβ(f/f);Lys-Cre and C/EBPβ(f/f) hosts. In the absence of myeloid Klf4, TAMs have reduced expression of surface mannose receptor and Fizz1 mRNA, both M2 markers. Global gene expression analysis further revealed activation of pro-inflammatory, pro-atherosclerotic pathways. Analysis of tumor-infiltrating lymphocytes (TILs) demonstrated markedly increased activated CD8 T cell numbers, and CD8 T cell depletion obviated the inhibitory effect of myeloid Klf4 deletion on prostate cancer growth. These findings suggest that reducing expression or activity of the Klf4 transcription factor in tumor myeloid cells may contribute to prostate cancer therapy.


Introduction
The micro-environment of prostate cancer (PCa) often includes abundant tumor-associated macrophages (TAMs). In a study of 131 PCa patients, increased TAMs correlated with PSA >50 and increasing Gleason score or T stage. In a multivariate analysis of these data that included PSA, Gleason score, extra-capsular extension, lymph nodes metastasis, and distant

Mice and ethics statement
Wild-type (WT) C57BL/6 (B6) mice were obtained from Charles River Laboratories. Lysozyme-Cre (Lys-Cre) B6 mice were obtained from Jackson Laboratory (#004781). Klf4(f/f) and C/EBPβ(f/f) B6 mice were obtained from the Mutant Mouse Regional Resource Center (#29877 and #34760). This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol (M013M116) was approved by the Johns Hopkins University Animal Care and Use Committee. All efforts were made to minimize suffering. Euthanasia was carried out by carbon dioxide inhalation followed by verification of death.

B6 Hi-Myc prostate cancer cells
After back-crossing FVB Hi-Myc mice [20] into the B6 background for >10 generations, a metastatic lesion was identified in a prostate draining lymph node from an 17-month-old mouse. Cells from this lesion were dissociated and passaged serially by subcutaneous (SQ) inoculation into the flanks of B6 mice as a B6 Hi-Myc PCa line. Hematoxylin and eosin staining of paraffin embedded tumors revealed dysplastic epithelial cells, and immunohistochemistry confirmed their expression of c-Myc and AR (S1 Fig) and CK8 (not shown). Androgensensitivity was confirmed by the rapid tumor regression evident upon orchiectomy (not shown).

Tumor cell inoculation
When tumors reached 1.2-1.5 cm, mice were euthanized and tumor tissue was collected, minced with a razor blade, washed with phosphate-buffered saline (PBS), resuspended in 5mL DMEM/F12 media containing 10% FBS, 500 μL Collagenase/Hyaluronidase (Stem Cell Technologies), 2.5U/mL Dispase and 0.05mg/ml Dnase I, and then incubated at 37˚C for 1 hr with occasional mixing. Tumor tissue was vigorously triturated and passed through a 40 μM cell strainer with the aid of a syringe plunger. Cells were then pelleted at 350 g x 5 min and resuspended in PBS. Live cells were then enumerated using Trypan Blue dye and a hemocytometer, and 2E6 viable cells in 100 μL PBS were injected SQ into the shaved flank of mice anesthetized with isofluorane. Tumor growth was monitored using caliper measurements of length (L), width (W) and height (H), with volume estimated from the ellipsoid volume formula: V = L Ã W Ã H Ã π/6 [21]. Mice were euthanized when tumors reached 2.0 cm in largest diameter or volume greater than 1000 mm 3 . Paraffin-embedded tumors were subjected to hematoxylineosin staining following by imaging at 400X using a Nikon E400 microscope, with attached CCD camera.

T cell depletion
To deplete CD8 T cells, mice were inoculated with 200 μg rat-anti-CD8 antibody (Bio X Cell, BE0117) intraperitoneally on days -7, -5, -3 and 0 relative to Hi-Myc PCa cell inoculation (day 0). Peripheral blood obtained on d -1 from mice was subjected to flow cytometry analysis to assess depletion.

RNA analysis
Tumor CD11b + cells were isolated using an anti-CD11b immunomagnetic isolation kit (Miltenyi), and RNA was isolated using NucleoSpin RNA II kit (Machery-Nagel). First-strand cDNA was prepared using AMV reverse transcriptase (Promega) and oligodT primer. Quantitative real-time PCR (qRT-PCR) was carried out using Lo-Rox SYBR Green (Alkali Scientific). Primer pairs are provided (S1 Table). For global gene expression analysis, total RNA was prepared from tumor CD11b + cells using the NucleoSpin RNA II kit. RNA quality was confirmed via Nanodrop-1000 spectrometric analysis and via a Bioanalyzer (Agilent Technologies). 500 ng total RNA from each sample was amplified and labeled using the Illumina TotalPrep RNA Amplification Kit with oligo(dT) priming (Ambion). 750 ng biotin-labeled cRNA was combined with hybridization buffer and hybridized to the Agilent Mouse GE 4 x 44K v2 Microarray at 58˚C for 16-20 hours. After hybridization, the array was washed with buffer at 55˚C and blocked at room temperature. Bound, biotinylated cRNA was stained with streptavidin-Cy3 and then washed. Dried arrays were scanned with the iScan System, and data with quantile normalization and background correction to 100 was exported from GenomeStudio v2011.1 Gene Expression Module 1.9.0 and imported to GeneSpring. Differentially expressed genes with >1.4-fold mean change were subjected to pathway analysis using Ingenuity Systems software with false discovery rate (FDR) = 0.05 as threshold.

Statistics
Tumor volumes, tumor growth rates, myeloid and T cell subsets, and RNA expression values were compared using the Student t test. Tumor growth rates were estimated by fitting exponential curves to volume data (Excel) for each tumor according to the equation V = V o e bt (V = volume, V o = initial volume, t = time in days and b = exponential slope). The slope or growth rate was estimated as the value 'b' from the above equation because ln(V) = ln(V o ) + bt. Means and standard errors (SE) are shown.

Hi-Myc prostate cancer growth is slowed in the absence of myeloid Klf4
Klf4 mRNA was reduced 11-fold in Klf4(f/f);Lys-Cre compared with Klf4(f/f) peritoneal macrophages (S2A Fig), consistent with prior findings with this model [19]. B6 Hi-Myc PCa cells were inoculated into the flanks of B6 Klf4(f/f) and Klf4(f/f);Lys-Cre mice and tumor volumes were monitored every 2-3 days beginning on day 22 (Fig 1A). Initial growth was uniformly slower in the absence of myeloid Klf4, with mean tumor volumes on day 29, with 6.4-fold lower mean tumor volumes in the Klf4(f/f);Lys-Cre recipients ( Fig 1B). Plotting the estimated tumor volumes subsequent to day 22 on a logarithmic scale allows fitting later growth rates to best fit lines ( Fig 1C), with the slopes of these lines on average~2-fold lower in the Klf4(f/f); Lys-Cre cohort ( Fig 1D). The average, estimated growth rates predict a 7-fold difference in tumor volume on day 29, similar to what was found by caliper measurements. In additional mice sacrificed on days 24-29 for tumor myeloid or T cell analyses, the tumor volumes were also significantly lower in the absence of myeloid Klf4 (S3 Fig).

Hi-Myc prostate cancer has increased TAMs with reduced MR and increased CD11c in the absence of myeloid Klf4
To evaluate how Klf4 influences myeloid composition and macrophage polarization in Hi-Myc tumors, we analyzed a panel of surface markers on dissociated tumor cells by multi-color flow cytometry (Fig 2A). CD45 + CD11b + myeloid cells represented~20% of live tumor cells isolated on day 21 from Klf4(f/f) and~29% of Klf4(f/f);Lys-Cre Hi-Myc PCa recipients, p = 0.07 (Fig 2B, right). The majority of tumor myeloid cells were F4/80 hi Ly6C lo/mid TAMs, and these increased approximately 1.5-fold in the absence of myeloid Klf4 (Fig 2B, left). The day 21 TAM population was further characterized by analysis of surface MR, MHCII/CD86, and CD11c (Fig 2C). MR hi cells, representing M2-biased TAMs, were reduced 1.8-fold, and MHCII + CD86 + cells, representing M1-biased TAMs, were unchanged. The mean fluorescence

Hi-Myc prostate cancer growth and tumor myeloid cells are unaffected by absence of myeloid C/EBPβ
C/EBPβ is another transcription factor reported to promote M2 polarization [10]. To test whether tumor myeloid cells utilize C/EBPβ to drive an M2 phenotype in prostate cancer, we Tumor growth rates were similar in the two cohorts ( Fig 3A). Tumor cell FC analysis found no differences in the proportions of total myeloid cells, TAMs, monocytes, granulocytes, MR hi TAMs, or MHCII + CD86 + TAMs on day 21 in control or C/EBPβ-deleted prostate cancer recipients ( Fig 3B). Neither the proportion of TAMs expressing CD11c nor their CD11c MFI was affected by absence of C/EBPβ (not shown). These data demonstrate that myeloid expression of C/EBPβ does not affect TAM polarization or Hi-Myc tumor growth.

Hi-Myc prostate cancer myeloid cells lacking Klf4 express RNAs associated with pro-inflammatory pathways
CD11b + myeloid cells were isolated from Hi-Myc prostate cancers 21 days after inoculation into Klf4(f/f) or Klf4(f/f);Lys-Cre recipients. RNA from these cells was analyzed for a panel of M1 and M2 macrophage markers by qRT-PCR (Fig 4). Most were not significantly changed, though the M1 marker Ccl2 was increased and the M2 marker Fizz1 decreased in the absence of myeloid Klf4. IL-1b and IL-6 were also not significantly different between the two groups (not shown). Consistent with our findings with peritoneal macrophages, TAM Klf4 mRNA was reduced 10-fold in the presence of Lys-Cre (S4 Fig).
To further evaluate the influence of Klf4 on TAM polarization, RNAs isolated from CD11b + tumor myeloid cells from two tumors in each host were subjected to global gene expression analysis. mRNAs corresponding to 277 genes were up-regulated and 119 down-regulated 1.4-fold or greater in the absence of myeloid Klf4 (S2 Table). Ingenuity Pathway analysis revealed activation of pathways involved in cellular movement, inflammation and adhesion. Genes increased or decreased within these pathways in our data set are listed (Table 1). Because "Atherosclerosis Signaling" was the top regulated pathway from our analysis we compared our microarray data to a recent proteomic analysis of atherosclerotic plaques from human patients that underwent endarterectomy [24]. 20 genes that were differentially regulated in CD11b + cells overlapped with 146 extracellular matrix (ECM) or ECM-related proteins identified in atherogenic plaques. Of these 20 genes, 19 were upregulated in tumors that formed in Lys-Cre;Klf4(f/f) mice (Table 2). These results suggest that myeloid deletion of Klf4 promotes an atherogenic-like microenvironment in Hi-Myc prostate cancer.

Absence of myeloid Klf4 increases activated CD8 + T-cells in Hi-Myc prostate cancer
Induction of pro-inflammatory pathways in tumor myeloid cells by Klf4 deletion might induce immune cell influx and activation. Prostate cancers from Klf4(f/f) or Klf4(f/f);Lys-Cre mice were analyzed for CD3, CD4 and CD8 tumor-infiltrating lymphocytes (TIL) on day 21 ( Fig 5A). CD3 + cells, representing total T cells, were increased 2-fold, and CD8 TIL were increased 4-fold in the absence of myeloid Klf4, whereas CD4 T cell numbers were unchanged. In absolute terms, CD8 T cells represented~8% of viable tumor cells. Activation of CD8 T cells was assessed by staining for intracellular IFNγ or surface CD69 (Fig 5B). The frequency of CD8 T cells expressing these markers were markedly increased in Klf4(f/f);Lys-Cre hosts, representing 1.5% or 3% of viable tumor cells. The proportion of CD3 + CD4 + CD25 + Foxp3 + Tregs was not changed (Fig 5C), indicating an overall increase in the CD8:Treg ratio.
To determine the contribution of CD8 T cells to the slower growth of Hi-Myc PCa calls in Klf4(f/f);Lys-Cre mice, we used antibody to deplete CD8 T cells. Klf4(f/f) and Klf4(f/f);Lys-Cre mice received 4 doses of CD8 antibody 7, 5, 3 and 0 days prior to PCa cell implantation, and depletion was confirmed by FC analysis of peripheral blood 1 day prior to implantation ( Fig  6A and 6B). Hi-Myc PCa growth rates were similar in both cohorts (Fig 6C-6E). Thus, CD8 depletion eliminated the difference in growth rates between Hi-Myc tumors grown in Klf4 (f/ f) and Klf4(f/f); Lys-Cre mice shown in Fig 1. These results suggest that CD8 T cells are critical for suppressing Hi-Myc prostate tumor growth in Klf4(f/f); Lys-Cre mice.

Discussion
Klf4 and C/EBPβ are transcription factors that play a role in monocyte development and favor macrophage M2 polarization. The Lys-Cre transgene deletes floxed alleles in macrophages, activated monocytes, and granulocytes [25]. We found that B6 Hi-Myc PCa tumors grew more  Limitations of our study include lack of evaluation in the orthotopic prostate environment, which can be challenging due to early ureteral obstruction and difficulty with monitoring tumor growth, and lack of evaluation of the effects of myeloid KLF4 deletion using additional prostate cancer models.
Amongst tumor myeloid cells, TAMs predominated over monocytes and granulocytes in this model, and F4/80 + TAMs were increased 2-fold in the absence of Klf4. TAMs lacking Klf4 manifested substantial reduction in surface MR and Fizz1 mRNA expression (M2 markers) and a gene expression signature demonstrating activation of multiple pro-inflammatory pathways. CD11b + cells from tumors grown in Klf4(f/f);Lys-Cre mice displayed increased expression of several interferon-regulated genes (Ifi27l2a, Isg20, Oasl1, Irf7 and Mx2), indicating an activated innate immune cell phenotype. However, several classic M1 genes, including Ifnb, Tnf, Cxcl9, and genes encoding MHC II subunits showed higher expression in Klf4(f/f) control mice. This suggests that deletion of Klf4 influences specific aspects of M1 differentiation and represses others in the Hi-Myc PCa microenvironment and perhaps less than that seen upon culture of Klf4(f/f) versus Klf4(f/f);Lys-Cre BMDM in IFNγ (M1 inducer) or IL-4 (M2 inducer) [19].
Notably, "Atherosclerosis Signaling" was the top pathway associated with the gene expression pattern in Klf4-deleted CD11b + myeloid cells isolated from Hi-Myc tumors. Atherosclerosis is a progressive disease that narrows the vascular lumen through a buildup of arterial plaque, an amalgam of lipids, macrophages and other leukocytes, smooth muscle cells, endothelial cells (EC), necrotic regions, and extracellular proteins. Nascent atherosclerotic plaques activate the vascular endothelium by increasing shear stress and by releasing proinflammatory cytokines, promoting recruitment and invasion of monocytes and other blood cells [26,27]. Klf4 has been identified as a critical molecule involved in the pathogenesis of atherosclerosis in macrophages and ECs. Exposure of peritoneal macrophages to an inflammatory lipid reduces Klf4 and increases M1 mRNAs, the latter induced to a greater extent in Klf4(f/f);Lys-Cre macrophage, and absence of myeloid Klf4 augments atherosclerosis in ApoE-/-mice [28]. Exposure of ECs to pro-inflammatory cytokines or shear stress reduces their expression of Klf4, deletion of endothelial Klf4 augments atherosclerosis, and transgenic EC-Klf4 reduces atherosclerosis [29,30]. Notably, we found that 20 differentially regulated genes from our microarray analysis matched proteomic analysis of extracellular proteins from atherosclerotic plaques [24], with a striking 19 out of these 20 mRNAs having higher expression in myeloid cells obtained from prostate cancers growing in Lys-Cre;Klf4(f/ f) compared with the Klf4(f/f) hosts.
In addition, we observed a nearly 2-fold increase in the frequency of tumor CD3 + T cells and a 4-fold increase in tumor CD8 + T cells in response to myeloid Klf4 deletion. Amongst CD8 T-cells, there was no change in the percentage of cells expressing IFNγ or CD69, indicating that myeloid Klf4 deletion did not augment CD8 lymphocyte activation. We found that CD8 T cells were critical for the growth defect caused by absence of myeloid Klf4 because depletion of CD8 T cells eliminated the growth deficit observed in Hi-Myc tumors. As Treg numbers were unchanged, the CD8 effector:Treg ratio was similarly increased. Notably, the CD8:Treg ratio has been shown to correlate with favorable prognosis in prostate cancer [31,32]. Cxcl9, a widely reported T-cell chemoattractant, was down-regulated in Klf4(f/f);Lys-Cre mice and our data did not reveal up-regulation of genes involved in T cell chemotaxis. These findings suggest that upregulation of chemokines is not likely to explain the rise in tumor T cell frequency in the Klf4(f/f);Lys-Cre mice. In co-cultures of Klf4(f/f) or Klf4(f/f); Lys-Cre CD11b + tumor cells with splenic T cells stimulated with anti-CD3 and anti-CD28, we did not observe a difference in CD8 T cell proliferation (not shown), suggesting that increased proliferation also does not explain the observed increase in T cell frequency. Increased T cell numbers in Hi-Myc tumors forming in Klf4(f/f);Lys-Cre mice is potentially related to a change in the composition of the ECM generated by CD11b + tumor myeloid cells lacking Klf4. The expression of α 5 , β 1 and β 3 integrins promotes retention and motility of activated T cells in inflamed non-lymphoid tissues dependent upon binding to fibronectin and collagenous fibers of the ECM [33][34][35][36]. In atherosclerotic lesions, T cells accumulate within the fibrous cap, where the ECM is composed of types I and III collagen with small patches of fibronectin [37]. CD8 T cells enter the arterial wall of atherogenic plaques and represent the dominant lymphocyte during advanced, high risk plaques, which coincides with active degradation of the plaque ECM by infiltrating macrophages [38,39]. We observed an increase in types I and III collagen, Pcolce (a collagen endopeptidase required for maturation of pro-collagen), and Dcn (a proteoglycan involved in the staggered assembly of collagen molecules in myeloid cells from Klf4(f/f);Lys-Cre vs Klf4(f/f) tumors. Low density ECM has the potential to suppress tumorigenicity via T cell recruitment; for example, in explants from human lung tumors and lung tumor xenografts T cells are motile along low density ECM fibers and excluded from areas of high ECM density [40,41]. Interestingly, myeloid Fizz1/Relmα expression, which was substantially decreased in Klf4(f/f);Lys-Cre tumor myeloid cells, was shown to Absence of myeloid Klf4 slows prostate cancer growth promote collagen-cross linking, increase collagen fibril density and facilitate wound healing in a model of cutaneous injury [42], further suggesting presence of low-density collagen fibers in Hi-Myc PCa tumors in Klf4(f/f);Lys-Cre recipients.
In conclusion, our findings implicate myeloid Klf4 in prostate cancer progression. Targeting myeloid Klf4 may provide therapeutic benefit to PCa patients via induction of a proinflammatory myeloid program and increased density of activated CD8 T cells and CD8:Treg ratio, potentially through effects on macrophage-mediated ECM remodeling.