Dipeptidyl Peptidase 9 (DPP9) Depletion From Hepatocytes in Mice Retards Liver Tumour Growth and Increases Intrahepatic Caspase-1 Activation

1. Introduction

Liver cancer accounts for 8.2% of cancer-associated mortality [1] and is the second leading cause of cancer related death in male patients [2]. In Australia, liver cancer is associated with a death-to-incidence ratio of 0.98 [3]. Over 90% of liver cancer is hepatocellular carcinoma (HCC), which has a median survival rate of 6.1 to 10.3 months [4,5]. The few therapeutic medicines for liver cancer patients include the multi-kinase inhibitors sorafenib, lenvatinib and brivanib which can upregulate tumour cell apoptosis and restrain angiogenesis [6,7], and immune checkpoint and angiogenesis inhibitors [8,9,10]. Drug-associated adverse events are common [8,9,11]. Therefore, there is a need to investigate novel medicines. Metabolic liver disease is the most rapidly rising risk factor for HCC, alongside rising global obesity and metabolic syndrome [12]. Therefore, our HCC mouse model incorporates a high-fat, high-sucrose, high-cholesterol diet.

A malignant tumour in the liver has distinct cytological and architectural features. Trabecular HCC hepatic cord width is above three nuclei while normal hepatic cord width is under two nuclei [13]. In humans, the histopathological differences between HCC and normal liver, and the variety of HCC nodules, have been established. There are two types of unusual liver lesions: dysplastic foci (< 1 mm) and dysplastic nodules (> 1 mm). The former can be divided into large cell change and small cell change, while the latter can be classified as low-grade dysplastic nodules and high-grade dysplastic nodules. Small cell dysplastic foci are regarded as premalignant [14]. Initial stages of HCC are marked by stromal invasion. In mouse and human tumours, there are many similarities among lesion patterns and histopathological details. Therefore, it is effective to classify liver lesion similarly in animal models as is done in human pathology [14,15,16,17].

The Dipeptidyl Peptidase 4 (DPP4) protease family is involved in cancers, inflammation and collagen turnover [18,19,20,21]. The DPP4 family is composed of six proteins, four of them (FAP, DPP4, DPP8 and DPP9) have enzymatic activities. Of these enzymes, FAP and DPP4 are on the cell surface and released as soluble proteins [22,23,24]. DPP8 and DPP9 are intracellular [25,26,27,28,29,30,31]. The enzymatic activities of DPP8 and DPP9 are very similar to DPP4 and the expression of these three proteases is ubiquitous in immune, endothelial, and epithelial cells. The differentiation of the complex roles of DPP4 with the more recently discovered DPP8 and DPP9 is essential [10,26,28,32].

Gene expression profiling of HCC has shown that DPP9 is upregulated in HCC and tumour-bearing liver tissue compared with normal liver tissue [33]. Moreover, protein expression of DPP9 is upregulated in chronically injured liver [34]. Overexpression of DPP9 is independently associated with lower 5-year survival in non-small lung cancer (NSCLC) [35]. Conversely, DPP9 inhibition has been associated with slower tumour growth, enhanced intratumoral immune response and increased macrophage pyroptosis [36,37].

DPP9 is a unique intracellular serine protease with diverse roles that encompass tumour growth, monocyte/macrophage death, inflammation, DNA repair, mitochondrial function and metabolism [38,39,40,41,42,43]. We previously found that pan-DPP inhibition in mice can reduce the volume and number of primary liver tumours [44]. However, the roles and functions of DPP9 in HCC and in pan-DPP inhibition are not understood. DPP9 suppresses inflammasome activation in epithelial cells [45,46]. This study specifically depleted DPP9 from a crucial and abundant liver epithelial cell, the hepatocyte, to examine whether that action upregulates inflammation markers and decreases tumour burden in a HCC model. This study also asked whether DPP9 is a dominant target of pan-DPP inhibition in mouse HCC.

With the development of DPP9 selective inhibitory compounds as potential therapeutics [47], the extent to which mammals can tolerate DPP9 depletion becomes a more important question. We discovered that loss of DPP9 is neonate lethal in mice [48,49] and that DPP9 Loss-of-Function SNPs is not tolerated in humans[49,50]. The present study showed that loss of DPP9 from hepatocytes is benign.

2. Materials and Methods

Mouse handling and maintenance

All mice were maintained in the Centenary Institute animal facility under specific, pathogen free conditions, co-housed with ad libitum food and water, filtered air and exposed to a 12 h light-dark cycle. C57BL/6 mice were purchased from either Animal Resource Centre (Perth, WA, Australia) or Australian Bio Resources (Moss Vale, NSW, Australia). All experiments were approved and monitored by animal ethics committees of The University of Sydney and Sydney Local Health District (animal welfare approvals 2013/030 and 2017/030) and conducted in accordance with applicable laws and regulations.

Generation of hepatocyte-specific DPP9 knockout primary HCC mouse model

The Dpp9^fl/fl allele was provided by the European Mouse Mutant Archive (C57BL/6N-Dpp9^{tm1a(EUCOMM)Hmgu/Cnrm}; EMMA ID EM:04611). The FRT-flanked LacZ reporter/neomycin resistance cassette was removed by FLPe recombinase transgenic mice [51]. Consequently, Dpp9 exons 5-7 were flanked by loxP sites and ready for deletion after breeding with mice expressing a Cre-recombinase such as Alb-Cre.

Female B6.Cg-Speer6-ps1^{Tg(Alb-cre)21Mgn}/J (Alb-Cre) mice (wt/wt Cre^+/+) (JAX stock #003547) [31], were crossed with male DPP9^fl/fl (Cre ^-/-) to generate hepatocyte-specific DPP9 knockout mice (fl/fl Cre^+/-, DPP9-KO). The littermates were used as controls (fl/fl Cre ^-/-, DPP9-WT). DPP9 depletion was validated in hepatocyte purification and in the liver at mRNA level.

In order to generate primary HCC, DEN was injected intraperitoneally at 25 mg/kg body weight at 12 days of age into male DPP9-KO (n = 17) and DPP9-WT mice (n = 13). At weaning (approximately 3 weeks old), mice were fed a High Fat Diet (HFD). The HFD was 45% kcal fat, 20% kcal protein, 35% kcal carbohydrate (Table 1). At 8 weeks of age, mice were given the hepatotoxin thioacetamide (TAA) (Alfa Aesar, Shanghai, China; catalogue number A12926) at 200 mg/L in drinking water, twice a week. Mice were euthansed and organs harvested at 28 weeks of age (Figure 1).

High fat diet (HFD)

The HFD was prepared in a fume hood and based on rodent diet no. D12451 [52]. The HFD ingredients are presented in Table 1. In a clean container, 182.88 g casein, 156.96 g sucrose, 150.48 g starch, 35.72 g AIN mineral mix, 39.92 g bran, 2.4 g methionine, 16.12 g gelatine, 3.24 g choline bitartrate, 10.36 g AIN vitamins, 4.2 g cholesterol were mixed thoroughly. Following that, 175.16 g of room temperature lard, 20.8 mL safflower oil and 2 drops of strawberry flavouring were added and mixed together to form a dough like texture. The mixture was stored under nitrogen at 4 $℃$ for up to two weeks.

Glucose Tolerance Test (GTT)

The intraperitoneal GTT was performed as previously described [53]. All mice were fasted for 6 hours and then administered α-D-glucose at 5 g/kg (Gibco TM, Auckland, NZ) via intraperitoneal injection, respectively. Glucose levels were measured in blood from the tail vein before (0 min) and at 15, 30, 45, 60, 90 and 120 min after the glucose bolus. All glucose concentrations were measured using an Accu-Chek Glucometer (Roche, Diagnostics GmbH, Mannheim, Germany).

Caspase-1 assay

The activity of caspase-1 in liver lysates was determined using the Caspase-1 Assay Kit (Fluorometric) (Abcam, ab39412) according to the manufacturer’s protocols. Fluorescence values were measured with a fluorescence microplate reader at excitation (400 nm) and emission (505 nm). The fold change in caspase-1 activity was determined by comparing the readings of induced samples with the results of the non-induced control.

Hepatocytes perfusion

Buffers used are: A: 50 mL HBSS (pH 7.4), B: 49.55 mL HBSS with 0.05 mL 0.5 mM EDTA, C: 50 mL HBSS with 0.25% (v/v) 5 mM CaCl₂ and 0.05% (w/v) Collagenase, D: 90 mL isotonic Percoll dissolved in 10 mL 10 X PBS, E: 250 mL RPMI with 5 mL FBS and 2.5 mL P/S, and F: 6 mL isotonic Percoll dissolved in 10 mL 1X PBS. Water bath was prewarmed to 40 $℃$ and the pump tube was prewashed by HBSS to minimise bubbles. Mice were killed by CO₂ asphyxiation and dissected. A blunt needle was inserted into the IVC and clamped down inside the vessel. The pump rate was set at 50 and circulated buffers A-B-A-C. The liver was removed and placed into buffer E, chopped and kept on ice. The liver mixture was passed through a 70 $\mu$M cell strainer and centrifuged at 50 x g for 3 minutes at 4 $℃$. The supernatant was collected as non-hepatocyte suspension and the pellet was resuspended into buffer E as hepatocyte suspension [54].

Following, the hepatocyte suspension was centrifuged at 50 x g for 3 minutes at 4 $℃$. The pellet was resuspended into buffer D and centrifuged at 50 x g for 10 minutes at 4 $℃$. The pellet was resuspended into buffer E to perform cell counting. After that, the cell suspension was centrifuged at 50 x g for 3 minutes at 4 $℃$ and washed with 1X PBS.

10⁷ cells were resuspended into 1 mL lysis buffer and stored at -20 $℃$ for future enzyme assay or Western blot. 5 x 10⁶ cells were stored as snap frozen cell pellet in RNAase-free tubes at -80 $℃$ for future qPCR use.

Western blot

Snap frozen mouse liver pieces were lysed in lysis buffer containing 50 mM Tris-HCl pH 7.6, 1 mM EDTA, 10% glycerol, 1% Triton-X100 and complete mini inhibitor cocktail. Protein concentrations were determined using the Micro BCA^TM Protein Assay Kit (ThermoFisher Scientific, Rockford, 23235) following manufactures instructions.

Following separation of proteins by SDS-PAGE, samples were transferred to PVDF membranes. Following the transfer, PVDF membranes were stained with Ponceau S to visualise proteins and subsequently de-stained in successive PBST washes. PVDF membranes were blocked in 5% (w/v) skim milk in PBST for an hour at room temperature. After blocking, PVDF membranes were incubated with primary antibodies overnight at 4 $\mathrm{℃}$ on a roller (Table 2). The following day, PVDF membranes were washed in PBST for 5 minutes 3 times, then incubated with secondary antibodies (Table 2) conjugated to horse-radish peroxidase (HRP) in 5% skim milk in PBST for 1-2 hours at room temperature. Proteins were then visualised using Immobilon^® Forte Western HRP in Chemi Doc MP imaging system.

qPCR

Total liver RNA was isolated using PureLink RNA Mini kit (ThermoFisher, 12183018A), and reverse transcribed to cDNA using Superscript VILO cDNA synthesis kit (Invitrogen, 11756050) following manufacturer’s instructions. The expression of genes was measured by qPCR using custom TaqMan array cards (format 384-well microfluidic card, Applied Biosystems, Foster City, CA), which were pre-spotted with custom designed, dried-down TaqManTM probes including housekeeping control Hprt1, as listed (Table 3). Real time qPCR used the QuantStudioTM 12K Flex Real-Time PCR System (Applied Biosystems) and Expression Suite v1.0 (Applied Biosystems), utilizing the comparative Cτ (ΔΔCτ) method for data analyses. Gene expressions were as a percentage to the housekeeping control.

Histology

Tissue samples were collected and fixed in 10% neutral buffered formalin overnight, then stored in 70% ethanol. Samples were then processed by the Histopathology Core Facility, Charles Perkins Centre, The University of Sydney.

5 $\mu$M sections were cut and incubated at 65 °C for 1 hour prior to Haematoxylin and Eosin (H&E), immunohistochemistry and immunofluorescence staining (Table 2 as described (Gall et al. 2013). Bright-field imaging was performed at 20x magnification on a Leica DM6000B microscope and analysed on the Mosaic software (Leica, Wetzlar, Germany) to calculate % area of total tissue stained.

Haematoxylin and Eosin (H&E) staining

Paraffin-embedded liver tissue was sectioned at 5 µm thickness. Briefly, slides were de-paraffinised with histolene and rehydrated. Sections were then stained with Harris Haematoxylin for 1.5 min, washed in ddH₂0 and 0.3% acetic alcohol. Slides were then incubated in Scott’s bluing solution for 3 mins, washed again in ddH₂0 and stained with Eosin Y for 1 min. Slides were dehydrated with ethanol, cleared with histolene and mounted with Eukitt, as previously described [44,55]. Histology was assessed by a certified pathologist.

Immunohistochemistry (IHC)

IHC was performed as described previously [44,56]. Briefly, paraffin-embedded liver tissue was sectioned at 5 $\mu$M. Slides were then deparaffinised with histolene, rehydrated and then antigen retrieved for 20 minutes using a pressure cooker and Sodium Citrate Buffer (pH 6.0). Sections were incubated with 0.3% H₂O₂ for 10 mins to inhibit endogenous peroxidases and rinsed with PBS. Sections were then incubated for 1 hour at room temperature with blocking solution (10% BSA, 10% in normal serum in PBST) and then incubated in primary antibody with 1% BSA in PBST at 4 °C overnight. After washing thoroughly in PBST, sections were incubated for 1.5 hours with secondary antibody conjugated to HRP, washed again thoroughly in PBS then incubated with 3,3- diaminobenzidine (DAB) dissolved in triple distilled water (TDW). Bright-field imaging was performed using a Leica DM6000B microscope.

Image analysis

Bright-field imaging was performed at 20x magnification on a DM6000B microscope and analysed using Mosaic software (Leica, Wetzlar, Germany) to calculate % area of total tissue stained. Entire tissue sections were scanned and individual tiles analysed using Leica application suite v4.8.0 (Leica, Wetzlar, Germany). Tiles with damage or artefacts were excluded from the final analysis. Measuring the immunostained area used a section exposed only to isotype-control immunoglobulin and the secondary antibody as negative control. The thresholds for total tissue were set with all non-tissue areas of the section including blood vessels excluded, at the following thresholds; H:3-5, S:48-255, I:0-243. The image tiles were then automatically analysed by the software and % tissue stain was determined using the following formula, as described [56]: % tissue stained = total stain area 𝑥 100 / total tissue area. Analysis of Sirius red staining was similar, as described previously [56].

To derive steatosis and inflammation scores, multiple photomicrograph tiles of each H&E stained slide, were scored, blinded, by two experienced researchers, using scoring criteria described elsewhere [52]. Lesions were categorised as either HCC, high grade dysplasia or low grade dysplasia [52].

Statistics and Data analysis

A two-way analysis of variance (ANOVA) with Tukey’s multiple comparison test, Kruskal-Wallis comparison test, or one or two-tailed Mann Whitney U test was used to compare data between groups. Data was plotted on GraphPad Prism (GraphPad v. 9.9, San Diego, CA, USA). Significance was assigned to p values; * = 0.05, ** = 0.01, *** = 0.001, **** = 0.0001.

3. Results

3.1. Hepatocyte-Specific DPP9 Knockout Primary HCC Mouse Model Validation

To validate the hepatocyte-specific DPP9 knockout mice, RT-qPCR was performed. As shown in Figure 2, DPP9 mRNA expression in the whole liver was greatly reduced, but not in spleen and kidney. qPCR analyses of RNA extracted from the perfused hepatocytes further validated that the reduced DPP9 mRNA expression occured in the hepatocytes. The results showed that DPP9 was successfully and specifically depleted in the hepatocytes in DPP9-KO mice.

In our mouse model of primary HCC, the body weight of mice was recorded every week until organs were collected at 28 weeks of age. Compared to DPP9-WT mice, DPP9-KO mice appeared to gain less weight (p = 0.053) (Figure 3B). DPP9-KO mice had significantly less liver and subcutaneous white adipose tissue (WAT) to body weight ratios at harvest (Figure 3A).

After mice were treated to generate HCC, depletion of DPP9 was retested. At the mRNA level, DPP9 was successfully depleted in mice livers, without compensation from other DPP4 family genes (Figure 4A). At the protein level, DPP9 depletion in mice livers was also shown upon western blotting (Figure 4B,C). Moreover, IHC showed dense brown stain in the nuclei for DPP9 in DPP9-WT mice, which was absent from DPP9-KO mice, both inside and separate from lesions (Figure 4D–G).

m. DPP9-WT n = 13; DPP9-KO n = 17. Individual data presented with mean. Mann-Whitney statistical test; ****p < 0.0001.

3.2. DPP9 and Glucose Metabolism

The metabolic effects of DPP9 deficiency in the liver at 22 weeks of age were investigated by GTT. The results showed that there was no difference between genotypes in glucose clearance and glucose AUC (Figure 5A, B). However, DPP9-KO mice had significantly lower Fasting Plasma Glucose (FPG) levels compared to DPP9-WT mice (Figure 5C). Interestingly, FPG and body weight presented a positive correlation in DPP9-KO mice (rho = 0.627, p < 0.01), while glucose AUC and body weight presented a moderately positive correlation in DPP9-KO mice (rho = 0.45, p < 0.05), indicating a possible function of DPP9 in glucose metabolism that requires further investigation (Figure 5D, E).

3.3. DPP9 and Cancer Burden

Mice livers from the primary HCC model were harvested at 28 weeks of age. Macroscopic nodules were observed on the surfaces of the mouse livers (Figure 6A). Hepatocyte-specific depletion of DPP9 in the primary HCC model caused fewer macroscopic nodules in DPP9-KO mice but without statistical difference compared with DPP9-WT mice (Figure 6B). These nodules were then classified by size, either as ≤ 3 mm and > 3 mm, whereby significantly fewer small macroscopic nodules (≤ 3 mm) were found in the DPP9-KO livers (Figure 6C). There was no significant difference in large macroscopic nodule numbers between the two genotypes (Figure 6D). These livers were sectioned, stained with H&E and histology assessed by a trained pathologist (Figure 7). There were no significant differences observed in the number of HCC, high grade dysplasia and low grade dysplasia, as well as inflammation and steatosis scores, between DPP9-KO and DPP9-WT mice (Figure 8). As a fibrosis measurement, Sirius red staining of the HCC bearing livers showed no difference in crosslinked collagen between DPP9-KO and DPP9-WT mice (Figure 9).

3.4. DPP9 and Inflammasomes Regulation

Caspase-1 is the hallmark of canonical NLRP1 inflammasome pathway which responds to the inhibition of the DPP9 [57,58,59]. Pro-inflammatory cytokines such as IL1β and Il18 can be cleaved by activated caspase-1 and leads to pyroptosis in cells [60,61,62]. There was a significant increase in the activated caspase-1 (p20) in DPP9-KO mice, revealing potential roles of DPP9 in tumour-bearing livers via the inflammasome activation pathway (Figure 10A,B). NLRP3 is an inflammasome that is highly involved in HCC but not regulated by DPP9. NLRP3 expression is dynamic during HCC development, with a low expression in normal liver tissue, higher expression during liver injury and downregulated after HCC developed [63]. Exported data suggested that the protein expression of NLRP3 remained the same between DPP9-KO and DPP9-WT mice in the HCC livers (Figure 10C, D). The mRNA levels of Nlrp1b, Caspase-1, Il18 and Il1β between DPP9-KO and DPP9-WT mice showed no difference (Figure 11A–D). Correlation analysis revealed that the mRNA expression of Caspase-1 is moderately positive associated with the mRNA expression of NLRP1b (rho = 0.5810, p < 0.05) and IL18 (rho = 0.5980, p < 0.05) in DPP9-KO mice livers (Figure 11E,F).

The enzyme levels of IL-1β and IL-18 were not significantly different between DPP9-WT and DPP9-KO livers, suggesting other drivers of inflammation (Figure 12A,C). Further, IL-1β and IL-18 did not correlate with activated caspase-1 (Spearman’s rho = -0.22, p = 0.399 for IL-1β in DPP9-KO; Spearman’s rho = -0.22, p = 0.470 for IL-1β in DPP9-WT; Spearman’s rho = -0.06, p = 0.809 for IL-18 in DPP9-KO; Spearman’s rho = 0.09, p = 0.765 for IL-18 in DPP9-WT) (Figure 12B,D).

3.5. Intrahepatic Gene Expression Other than Inflammasome Related Genes

Intrahepatic expression of genes associated with HCC, fibrosis, immune response, macrophages and TLR pathway was also assessed. There was no significant difference in the mRNA expression of HCC associated genes: Afp, Gpc3, Birc5, Braf, Trp5 and Ccnd1; immune response associated genes: Tnf, Itgam, Nfkbib, Ccl5, Ccl2, Cx3cr1, Il-6, Cxcl10, Ccr2 and Itgax; inflammasome associated genes: Nlrp3, Caspase-3 and Gasdermin D; macrophage associated genes: CD163, CD47 and CD68; TLR genes: Tlr7, Tlr8 and Tlr9; and extracellular matrix associated genes: Col1a2 and Col3a1. Conversely, there was significantly greater expression of the immune associated genes: Nfkbia and Cxcr3 and macrophage associated gene CD64 in the livers of DPP9-KO mice compare to DPP9-WT mice (Figure 13).

Tumour infiltrating CD8+ cells in the HCC bearing livers showed no difference between DPP9-WT and DPP9-KO mice (Figure 14).

3.6. Autophagy and DNA Repair

Beclin1 is a marker for autophagy [64,65]. Beclin1 protein showed significant upregulation in DPP9-KO mice compared to DPP9-WT mice (Figure 15A, B). The tumour suppressor protein p53 can arrest cell cycle progression in cellular stress (Soussi 2000). Here, protein levels of p53 were elevated in the livers of DPP9-KO mice compared to their DPP9-WT counterparts (Figure 15C, D).

4. Discussion

This study successfully established non-lethal hepatocyte-specific knockout of the DPP9 protease in an epithelial cell type using Cre recombinase driven by the albumin promoter. This approach permits studies of DPP9 knockdown following our previous discovery that universal knockout of DPP9 is neonate lethal [48]. Applying our multi-insult DEN/TAA/HFD primary liver cancer model [66] to this novel mouse strain revealed that DPP9-KO mice had improved energy metabolism, evidenced by smaller subcutaneous adipose tissues and livers and better glucose control. Our observation of fewer small liver lesions suggests depressed levels of tumour initiation. The DPP9-KO mice exhibited more intrahepatic cleaved caspase-1 and a few inflammation markers but not increases in cellular inflammation or downstream pyroptosis markers, so NLRP1 inflammasome activation was not a clear mechanism in this model. However, p53 and Beclin-1 levels were increased compared with littermate control livers. These data are consistent with known roles of DPP9 in adipogenesis and in NLRP1 inflammasome activation as well as with previously unappreciated roles in glucose metabolism, tumour suppression and autophagy (Figure 16).

Validation of DPP9-KO Model

Cre enzyme expression in the Alb-Cre transgenic mouse strain that we used is controlled by the albumin promoter and so is restricted to hepatocytes [31,67]. DPP9 mRNA and protein expression was greatly depressed in our DPP9-KO mice in whole liver and in isolated hepatocytes, thus validating the model.

DPP9 and Liver Inflammation

Our data is consistent with the well characterised role of DPP9 in suppressing NLRP1 activation [68]. NLRP1 inflammasome activation in response to intracellular stress signals results in self-cleavage activation of caspase-1. Activated caspase-1 (p20) can convert the precursor of inflammatory cytokines IL-1β and IL-18 into mature and active cytokines and release them by inducing cell apoptosis [69,70,71]. DPP9 is ubiquitous and is prominent in both epithelial and leucocytic cells [72,73]. Showing that caspase-1 cleavage was greater in DPP9-KO mice, and thus in liver epithelial cells, concords with a recent finding that DPP9 suppresses NLRP1 in corneal epithelial cells [45]. So, this action of DPP9 is common to both epithelia and the monocyte/macrophage lineage. Our observation of increased intrahepatic caspase-1 cleavage in mice treated with an inhibitor of DPP9 confers additional support for this conclusion [44]. In contrast to that previous study, in which inflammation score and expression of some inflammatory markers was greater with DPP9 inhibition, lifelong DPP9 depletion from hepatocytes did not alter inflammation score, but did increase several inflammation markers, NFkbia, NFkbib, CCL5 and CXCL10 (IP10). Our data supports the already expanded view of an anti-inflammatory role for DPP9: that is, establishment of the pro-inflammatory effects of DPP9 down-regulation derived from hepatocytes and other epithelial cell types, in addition to the monocyte/macrophage lineage [45,62,68,74].

The transcription factor NF-κB can mediate the expression of cytokines and inflammatory mediators [75,76]. Both Nfkbia and Nfkbib are inhibitors of NF-κB via IκB [77,78]. When NF-κB is activated, it enters the cell nucleus and binds to the promoters of chemokines for promoting their transcription [79,80]. In this study, all measured cytokines exhibited no change in expression between DPP9-KO and DPP9-WT mice, however, two important chemokines were upregulated. CCL5 and CXCL10 induce migration, mainly of monocytes and lymphocytes, into injured tissues [81,82,83]. However, in contrast to systemic administration of a DPP9 inhibitor to hepatocyte-specific depletion in this liver cancer model [44], we did not observe significant differences in CD8+ cell numbers or locations in the livers.

DPP9 and Adipose Tissue

Knowledge of a relationship between DPP9 and adipose tissue is limited. Global depletion of DPP9 enzyme activity influences metabolic gene expression levels in newborn mouse liver, including increased AMPK [84], which could increase fat utilization [85]. DPP9 downregulation in vitro prevents preadipocyte differentiation in 3T3-L1 cells and downregulates expression of the nuclear receptor PPARγ [86], a regulator of adipocyte differentiation and lipid metabolism [87]. Therefore, the lower liver:body weight and subcutaneous white adipose tissue (WAT):body weight ratios in DPP9-KO mice suggest that DPP9 has an important role in the normal development and accumulation of adipose tissue and possibly in intrahepatic lipid deposition. Despite no difference in steatosis scores, the DPP9-KO livers were smaller and therefore stored less total lipid. Inhibition of the DPP9 related protease DPP4 lowers plasma glucose, and protects against diet-induced liver and adipose tissue inflammation through decreasing the expression of fatty acid synthase [88]. However, there is no rationale for suggesting that the intracellular molecule DPP9 could act similarly to extracellular DPP4.

DPP9 and Glucose Metabolism

DPP9 is not an established glucose metabolism regulator, so the improved FPG in DPP9-KO mice was unexpected. Our finding is possibly a consequence of reduced WAT, or a consequence of the same, unknown mechanism that caused reduced WAT. The FPG effect was significant but small and was insufficient to significantly affect glucose tolerance. Nevertheless, DPP9 has been associated with metabolic processes in the only study of gut and liver gene expression in DPP9 deficient mice [40]. That study observed associations of DPP9 deficiency with long chain fatty acid uptake, lipoprotein metabolism, mitochondrial function, adipokine transport and gluconeogenesis in vivo, and, in vitro, showed a link with AMPK and that insulin and palmitate cause altered DPP9 levels [40]. Furthermore, proteomics has shown that DPP8/DPP9 inhibition regulates glycolysis in the THP-1 cell line [39]. Thus, a variety of data implicate DPP9 in energy storage and metabolism. Our new data supports that role of DPP9 in hepatocytes in vivo.

The impact of DPP9 on FPG could be mediated by an unidentified binding activity, but would more probably be mediated by DPP9 enzyme activity on a substrate. The only potential substrate that has been identified is nucleobindin-2/nesfatin-1 [89], which influences appetite and is regulated by PPARγ [90]. Other metabolism/appetite – associated DPP4 substrates, such as NPY, PYY, GlP-1 and GLP-1 can be cut by DPP9 [91,92,93] but are extracellular, so their exposure to DPP9 is very unlikely.

Another possible mechanism for energy regulation by DPP9 may be through mitochondrial protein degradation. Recently, Finger et al. (2020) identified a novel pathway in which cytosolic DPP8 and DPP9 have a quality control function to prevent abnormal cytosolic accumulation of mitochondrial precursors [38]. Notably, DPP8/9 activity ensures stable processing of intracellular AK2 [38]. Attenuation of DPP9, but not DPP8, leads to excess cellular levels of AK2 [38]. AK2 is associated with oxidative phosphorylation [94] and metabolic signaling [95]. Therefore, these studies suggest a possible role for DPP9 in energy regulation by processing mitochondrial proteins.

Taken together, our data and the literature suggest an under-recognized role for DPP9 in energy regulation.

DPP9 and Beclin-1

The upregulation of intrahepatic Beclin1 protein levels in DPP9-KO mice is a novel observation suggesting that DPP9 has a role in autophagy. This finding concords with the very recent discovery that DPP9 downregulation in the MCF-7 breast cancer cell line increases autophagy [96]. Beclin1 regulates autophagy by forming complexes through interactions with other proteins and participating in various stages of autophagy [97,98,99,100]. Beclin1 can inhibit autophagy by binding to Bcl-2 family proteins to regulate cell growth and apoptosis [101,102]. In addition, Beclin-1 up-regulates USP10 and USP13, which stabilize p53 by de-ubiquitination [103]. p53 is an important tumor suppressor that plays crucial roles in cell cycle regulation, DNA damage repair, apoptosis, and metabolic regulation [104]. Upon cellular stress, p53 can halt cell cycle progression, promote cellular senescence or promote apoptosis (reviewed in: [105]). Mutations in p53 are present in almost 50% of all cancers, making it an attractive target in cancer therapeutics and diagnostics [105]. DPP9 downregulation has been found to increase p53 protein levels in human non-small cell lung cancer cells [106], analogous with our observation that DPP9 depletion promoted p53 protein expression in mouse hepatocytes. Concordantly, tumoral overexpression of DPP9 is linked to poor survival in humans [35,50]. These findings suggest that inhibition of DPP9 may stabilize or enhance mammalian p53 expression, suggesting a possible avenue for tumour therapeutics.

Therefore, our data suggests that DPP9 might both regulate hepatocyte autophagy and influence tumor suppressor p53 levels via beclin-1.

DPP9 and Liver Tumor Growth

Tumourigenesis was assessed by histopathology as well as quantifying macroscopic liver nodules. DPP9-KO mice showed no difference in total macroscopic nodule numbers or tumor burden, but had fewer small liver nodules of < 3 mm diameter compared with their DPP9-WT littermates. This HCC mouse model uses DEN, HFD and TAA, which cause continuous chronic hepatic inflammation, chronic hepatocyte cell death and metabolic dysregulation. The high fat, high sucrose, high cholesterol diet (HFD) causes expansion of adipose tissue and lipid deposition in hepatocytes whereas TAA causes some weight loss and lipid depletion from hepatocytes, but, as again seen here, the TAA in this model does not reverse all the HFD induced lipid accumulation [44,66]. Effects upon inflammation markers, beclin-1, lipid storage and blood glucose were seen in this model when DPP9 was depleted from hepatocytes, and all four parameters could have influenced tumorigenesis and nodule development. The reduced number of small nodules may reflect a reduction in liver cancer initiation in the mice with downregulated hepatocyte DPP9, while progression into larger liver tumors was unaltered. The latter was unexpected and suggests that derepressing NLRP1 was not the dominant outcome of DPP9 depletion in hepatocytes. These findings are also consistent with an overall interpretation that inhibition of DPP9 stabilizes or enhances mammalian p53 expression.

5. Conclusions

We established a hepatocyte-specific DPP9 knockout mouse model to investigate functions of DPP9 in this crucial epithelial cell type. We observed effects on WAT and liver mass, blood glucose, hepatocellular carcinoma development, inflammation markers, p53 and beclin-1. These results suggest that DPP9 exerts effects in hepatocytes in multiple aspects in vivo, rather than just suppressing activation of the NLRP1 inflammasome. This study initiates a deeper and broader understanding of the functions and mechanisms of DPP9 in epithelial cells, in the liver, and in liver cancer.