Abstract

Objectives

   The Glucokinase Regulatory Protein GKRP, encoded by GCKR, enables acute
   regulation of liver glucokinase to support metabolic demand. The common
   human GCKR rs1260326:Pro446 > Leu variant within a large linkage
   disequilibrium region associates with pleiotropic traits including
   lower Type 2 diabetes risk and raised blood triglycerides and
   cholesterol. Whether the GCKR-P446 > L substitution is causal to the
   raised lipids is unknown. We determined whether mouse GKRP phenocopies
   the human GKRP:P446 > L substitution and studied a GKRP:P446L knockin
   mouse to identify physiological consequences to P446 > L.

Methods

   GKRP-deficient hepatocytes were transfected with adenoviral vectors for
   human or mouse GKRP:446 P or 446 L for cellular comprehensive analysis
   including transcriptomics consequent to P446 > L. Physiological traits
   in the diet-challenged P446L mouse were compared with pleiotropic
   associations at the human rs1260326 locus. Transcriptomics was compared
   in P446L mouse liver with hepatocytes overexpressing glucokinase or
   GKRP:446 P/L.

Results

   1. P446 > L substitution in mouse or human GKRP similarly compromises
   protein expressivity of GKRP:446 L, nuclear sequestration of
   glucokinase and counter-regulation of gene expression. 2. The P446L
   knockin mouse has lower liver glucokinase and GKRP protein similar to
   human liver homozygous for rs1260326-446 L. 3. The diet-challenged
   P446L mouse has lower blood glucose, raised blood cholesterol and
   altered hepatic cholesterol homeostasis consistent with relative
   glucokinase-to-GKRP excess, but not raised blood triglycerides.

Conclusions

   Mouse GKRP phenocopies the human GKRP:P446 > L substitution despite the
   higher affinity for glucokinase of human GKRP. The diet-challenged
   P446L mouse replicates several traits found in association with the
   rs1260326 locus on chromosome 2 including raised blood cholesterol,
   lower blood glucose and lower liver glucokinase and GKRP protein but
   not raised blood triglycerides.

   Keywords: Liver, Glucose metabolism, Glucokinase, Type 2 diabetes,
   Fatty liver, Blood cholesterol

Graphical abstract

   Image 1
   [45]Open in a new tab

Highlights

     * •
       P446 > L substitution in mouse or human GKRP compromises GKRP
       stability and GK binding.
     * •
       The GKRP:P446L mouse has low liver GKRP and glucokinase similar to
       rs1260326-TT liver.
     * •
       The GKRP:P446L mouse manifests low blood glucose and insulin
       similar to rs160326-TT.
     * •
       The diet-challenged GKRP:P446L mouse has high blood cholesterol but
       not triglyceride.
     * •
       The GKRP:P446L mouse replicates the raised blood cholesterol linked
       to rs1260326(C > T).

1. Introduction

   Genome wide studies identified hundreds of gene loci associated with
   Type 2 diabetes or blood lipids [[46]1,[47]2], each harbouring multiple
   variants which are a potential resource for exploring disease
   mechanisms. Several intronic and an exonic (rs1260326^P446>L) variant
   in the GCKR gene, within a large region of linkage disequilibrium on
   chromosome-2, associate with raised blood triglycerides and cholesterol
   and with increased risk for fatty liver disease but with decreased risk
   for Type 2 diabetes [[48][3], [49][4], [50][5], [51][6], [52][7],
   [53][8], [54][9]]. The GCKR gene encodes the glucokinase regulatory
   protein GKRP, which is expressed predominantly in the liver, and
   localized to the hepatocyte nucleus [[55]10,[56]11]. Converse risks for
   Type 2 diabetes and liver disease through the same GCKR variant has
   implications for therapies targeting GKRP or glucokinase for type 2
   diabetes [[57]12].

   The only established function of GKRP is as a negative regulator for
   glucokinase (GK) which catalyses the first reaction in hepatic glucose
   metabolism. GKRP belongs to the sugar isomerase family and has a
   high-affinity binding site for fructose 6-P and fructose 1-P which
   enhance and attenuate, respectively, its affinity for GK
   [[58]13,[59]14]. It is a competitive inhibitor of GK with glucose and
   sequesters GK in the nucleus at basal blood glucose releasing GK to the
   cytoplasm in response to raised blood glucose after a meal. This allows
   a reserve GK pool in the nucleus, which can be rapidly mobilised to
   meet metabolic demand [[60]10,[61]11]. Cytoplasmic GK activity is
   determined by both the affinity of GKRP for GK and by the molar ratio
   of GK-to-GKRP, which is dependent on nutritional state [[62][15],
   [63][16], [64][17]].

   The assumed hypothesis for the raised blood and liver lipids and lower
   blood glucose associated with the GCKR locus is that the missense
   GKRP:446 L variant impairs GK binding to GKRP [[65][18], [66][19],
   [67][20]] and thereby favours hepatic conversion of glucose to
   triglyceride through the uninhibited GK [[68]21]. Other hypotheses are
   that the raised triglycerides may be due to GKRP functions independent
   of its interaction with GK [[69]14], that the intronic rs780094 variant
   affects gene transcription [[70]22], or other neighbouring gene
   variants at the locus may have synergistic effects [[71]23]. Although
   the first hypothesis is assumed, there has been no supportive evidence
   from animal models that the GCKR P446 > L substitution lowers blood
   glucose or raises blood triglycerides. Whether the rodent is a valid
   model for the human GKRP:P446 > L substitution is contentious [[72]19],
   because of apparent species differences in the P446 > L substitution
   between human and rat GKRP [[73]18,[74]19,[75]24].

   Here we assessed from cellular studies the validity of mouse GKRP to
   model the human GCKR:P446 > L substitution by expressing human or mouse
   GKRP:446 P/L at physiological GKRP-to-GK ratios in GKRP-deficient mouse
   hepatocytes and show that mouse GKRP:P446 > L phenocopies the human
   GKRP:P446 > L substitution despite the higher affinity of human GKRP
   for GK [[76]25]. From comparison of GKRP and GK protein levels in human
   liver homozygous for the rs1260326:P446 > L, with a P446L knock-in
   mouse [[77]26] we show that the mouse model replicates the lower liver
   GK and GKRP protein levels in human liver of 446LL genotype. The P446L
   mouse challenged with a high-fat high-sugar diet manifested lower blood
   glucose and insulin and raised blood cholesterol but not raised
   triglycerides. It thereby replicates some of the metabolic traits
   associated with the Chromosome-2 locus. Transcriptome analysis of P446L
   mouse liver and hepatocytes overexpressing GK demonstrates a link
   between relative GK-to-GKRP protein excess and altered hepatic
   cholesterol homeostasis.

2. Materials and methods

2.1. Animals and ethics statement

   All animal studies were approved by Newcastle Animal Welfare Ethical
   Review Board (AWERB No-532) and covered by UK Home Office Licence
   PC1B78F4. They were conducted in compliance with the Institution's
   ethical (AWERB) and ARRIVE guidelines. Two mouse lines were used: the
   H-GCKR-DEL1262-EMI-B6N generated at MRC Harwell (Gckr^em1(IMPC)H
   [78]mouse.phenotype.org) referred to here as Gckr^+/− which has a
   1262 nt deletion in exons ENSMUSE00000487442 and ENSMUSE00000486780 and
   the GCKR-P446L-EMI-B6N (referred to as here as P446L) generated as
   described previously [[79]26]. Mice were housed in environmentally
   controlled conditions at 20 ± 2 °C, with 12 h light/dark cycle in
   individually ventilated cages, with ad libitum access to water and
   standard rodent diet unless otherwise stated.

2.2. Human liver and ethics statement

   Liver biopsies (formalin-fixed paraffin-embedded needle biopsy, n = 48
   cases; 32 M; 16 F) with confirmed Nonalcoholic fatty liver disease
   (NAFLD) with simple steatosis or NASH (nonalcoholic steatohepatitis)
   were accessed via the European NAFLD Registry [[80]27] from the
   Newcastle upon Tyne Hospitals NHS Trust (n = 36) and University Medical
   Centre Mainz (n = 12). They were selected based on the GCKR rs780094
   genotype from previous analysis [[81]27]. Informed consent was obtained
   to use their surplus tissue. Prospectively recruited patients were over
   18 years of age. Collection and use of biological samples, surplus to
   diagnostic requirement for research was approved by the North East
   -Tyne & Wear South Research Ethics Committee (Fel:15/NE/0150,
   IRAS:178,250, UK) and by Ethikkommission der Landesärztekammer
   Rheinland-Pfalz (Ref: 2018–13269, Germany). Fixed-tissue was genotyped
   for GCKR rs1260326 (ThermoFisher TaqMan, Assay ID C_2862880_1 primers).
   Histopathology was assessed from Haematoxylin and Eosin (H&E) and
   Sirius red fast green (SRFG) stained sections by an expert liver
   pathologist (DT) blinded to genotype and scored for grade of steatosis
   (0–3), hepatocyte ballooning (0–2), lobular inflammation (0–3) and
   NAFLD activity score [[82]28]. From this 40 cases (23CC, 17 TT) were
   used for the study.

2.3. P446L mouse model

   P446L mice were fed on standard rodent diet unless otherwise indicated.
   For the high-fat diet (HFD) and the high-fat high-sugar diet (HFHSD)
   studies which were run consecutively male mice were used and the food
   pellets (Special Diet Services, Whitham Essex #824018) contained (g%
   w/w): casein 26.5; choline bitartrate 0.296; l-cystine 0.398; lard 18;
   rice starch 18.428; cellulose 6.16; soya oil 4.315; sucrose 20.343;
   mineral mix 4.315; vitamin mix 1.233 and was by energy (kcal): 45% fat
   (36.3%, lard; 8.7%, soya oil); 20% protein; 35% carbohydrate. For the
   HFHSD the drinking water contained 10% glucose and 5% fructose. Food
   intake was monitored weekly. On termination of the studies tissues were
   harvested after isoflurane anaesthesia and blood was collected from the
   heart/thoracic cavity. Liver sections from the left lateral lobe were
   frozen in liquid nitrogen for protein, lipid and RNA analysis and
   formalin fixation.

2.4. Mouse hepatocyte isolation

   Hepatocytes were isolated from mice of the GCKR-DEL1262-EMI-B6N
   (Gckr^−/− genotype) or the GCKR-P446L-EMI-B6N (P446L, PP and LL
   genotypes) strains fed on standard rodent diet, by collagenase
   perfusion [[83]29]. They were cultured in monolayer in Minimum
   Essential Medium (MEM) containing 7% v/v neonatal calf serum, 10 nM
   insulin, 10 nM dexamethasone [[84]29]. After cell attachment (∼2–3 h),
   the medium was replaced by serum-free MEM containing the adenoviral
   vectors. For other experiments it was replaced by serum-free MEM
   containing 10 nM insulin and 10 nM dexamethasone. All experimental
   studies were performed after overnight culture.

2.5. Transfection of human GKRP and mouse GKRP in mouse hepatocytes

   Hepatocytes isolated from either Gckr^−/− mice (male, age 8–24 wk) or
   P446L mice (homs 446LL, 8–24 wk) were used. Male mice were used in all
   studies except for the hepatocyte RNA-sequencing study ([85]Figure 2)
   which was on female P446L mice (n = 3, age 14 wk). Hepatocytes from
   these genotypes do not show nuclear sequestration of endogenous GK.
   After hepatocyte attachment (2–3 h) the serum-containing MEM was
   replaced by serum-free MEM containing the GKRP-adenoviral vectors
   (human GCKR(Adv-209743; Ref-seq [86]BC130481: 446P or 446L) and mouse
   Gckr transcript-1 (Ref seq [87]XM_006503881.3: 446P or 446L);
   Gckr-transcript-2 (Adv:259,984; [88]BC012412: 446P or 446L) and Gckr-X1
   ([89]XM_006503882); Gckr-X2 ([90]XM_006503883: 446P) at final titres
   between 5 × 10^6 and 40 × 10^6 pfu/ml (corresponding to MOI 5–50
   pfu/cell). For the GK-adenoviral vector [[91]30], a titre determined
   empirically to increase GK by 2–4 fold above endogenous was used. After
   incubation with the vectors (4–5 h), the medium was replaced with MEM
   containing 10 nM dexamethasone and 10 nM insulin for overnight culture.

Figure 2.

   [92]Figure 2
   [93]Open in a new tab

   Functional assessment of human and mouse GKRP:P446 > L by
   glucose-dependent translocation and transcriptome analysis. A-C)
   Effects of glucose concentration (5–35 mM) on nuclear sequestration
   (N/C: nuclear/cytoplasmic ratio) of GKRP (A) and GK (B),(C) in
   hepatocytes from GKRP-deficient mice transfected (10 × 10^6 pfu/ml)
   with human or mouse GKRP:446 P/L (hP, hL, mP, mL) and cultured for
   24 h. A) Nuclear sequestration of GKRP is independent of glucose
   concentration. B) Translocation of GK at 35 mM glucose in hepatocytes
   expressing human GKRP (hP or hL). C) Translocation of GK at ≥20 mM
   glucose and ≥10 mM glucose in hepatocytes expressing mouse GKRP mP or
   mL, respectively, n = 3 hepatocyte preparations, ∗P < 0.05 relative to
   5 mM glucose. D-H) Gene expression by unbiased RNA-transcriptome
   analysis in hepatocytes that were either untreated (none) or
   transfected (10 × 10^6 pfu/ml) with human (hP, hL) or mouse (mP, mL)
   GKRP and either without (GK-End) or with GK overexpression (GKOE) and
   cultured for 24 h followed by 4 h incubation with 25 mM glucose before
   RNA extraction and processing for RNA-Sequencing, n = 3 hepatocyte
   preparations (D). E) Gene counts for mouse Gck and Gckr showing
   expression by the adenoviral vectors. F) Numbers of significantly
   up-regulated or down-regulated genes by GKOE relative to the 5 combined
   groups of GK-End (none, hP, hL, mP, mL). G) Venn diagram showing genes
   down-regulated by GKRP with GKOE relative to GKOE alone. H) Gene counts
   for 14 of the genes up-regulated by GK and down-regulated by GKRP and
   for Hmgcr which is conversely regulated by GK and GKRP showing
   heterogeneity of response (hP, mP, hL, mL).

2.6. Hepatocyte incubations for glucose metabolism and cell metabolite
analysis

   Glucose phosphorylation and glycolysis in hepatocytes were determined
   after 24 h culture by incubation with [2–^3H]glucose or [3-^3H]glucose,
   respectively as described [[94]17]. Cell ATP was determined by a
   Luciferase coupled luminometric assay (Sigma FLAA) and glucose 6-P and
   glycerol 3-P were determined by fluorimetric assays [[95]29].

2.7. Immunostaining for GKRP and GK in hepatocyte monolayers

   Hepatocyte monolayers were fixed for 30 min in 4% (w/v)
   paraformaldehyde in PBS, treated with sodium borohydride (1 mg/ml in
   PBS, 10min) followed by Triton X100 (0.2% w/v) in PBS (10 min) and
   stained separately for GKRP (SC-6340, GCKR-N19, 4 μg/ml) or GK
   (SC-7908, GCK-H88 at 4 μg/ml or PT15629 at 1:200 dilution) with
   Triton-X100 (0.1% w/v), BSA (1%) in PBS (2 h). For GKRP immunostaining
   the secondary antibody was Alexa-Fluor-488 anti-goat-IgG and for GK
   Alexa-Fluor-546 anti-rabbit-IgG (1:50 dilution) with 1% Triton-X100,
   BSA (1%) in PBS (1 h). Nuclei were counterstained with DAPI. Cells were
   imaged using a Nikon E400 microscope (60X). For each coverslip 5 fields
   were selected based on DAPI staining representing 6–20 nuclei/field.
   The mean pixel intensity for the nucleus and cytoplasm was analysed
   using ImageJ and the nuclear-to-cytoplasmic (N/C) ratio was calculated
   for each coverslip from the 5 fields averaging 40–50 nuclei. With the
   exception of the data in [96]Figure 1K–N which represents a single
   hepatocyte preparation (representative of 2) other data
   ([97]Figure 1G–J) represent individual hepatocyte experiments.

Figure 1.

   [98]Figure 1
   [99]Open in a new tab

   P446 > L substitution in human or mouse GKRP compromises protein
   expressivity and GK nuclear sequestration. Hepatocytes isolated from
   GKRP-deficient mice were treated with adenoviral vectors (at 5 × 10^6
   or 10 × 10^6 pfu/ml) to express human or mouse GKRP:446 P/L (hP, hL,
   mP, mL) and where indicated with a vector to overexpress GK (−/+ GKOE:
   D,F,I,J,L-N) and cultured for 24–30 h. Inhibitors (CX, AGK2: C-E) were
   added for the last 6 h culture. A) GKRP-deficient hepatocytes
   transfected with human GKRP:446 P at 5 × 10^6 or 10 × 10^6 have
   comparable GKRP immunoactivity, lower endogenous GK and higher GKRP/GK
   immunoactivity ratios than wild-type hepatocytes. Means ± SEM, n = 5,
   ∗P < 0.05. B) Lower expressivity of GKRP:446 L variant (hL, mL vs hP,
   mP) in transfections at 5 × 10^6 or 10 × 10^6 pfu/ml is more prominent
   at the higher titre, n = 4, #P < 0.05 446 L vs 446 P. C) Lower protein
   stability of human compared with mouse GKRP after incubation with
   cycloheximide (CX). Pooled data of transfections with 5 × 10^6 and
   10 × 10^6 pfu/ml and cultured for 24 h followed by 6 h culture −/+
   10 μM CX, n = 5, ∗P < 0.05 effect of CX; #P < 0.05, 446 L vs 446 P. D)
   GK overexpression by ∼2-fold relative to endogenous GK promotes
   stabilization of human GKRP:446 P/L and mouse GKRP:446 L. Means ± SEM,
   n = 5 hepatocyte preparations; pooled data of 5 × 10^6 and
   10 × 10^6 pfu/ml transfections. ∗P < 0.05 effect of GKOE; #P < 0.05,
   446 L vs 446 P. E) Stabilization of GKRP:446 L by a Sirt2 inhibitor
   (10 μM AGK2) in transfections at 5 × 10^6 pfu/ml. Means ± SEM, n = 3,
   ∗P < 0.05 effect of AGK2; #P < 0.05, 446 L vs 446 P. F) GKRP and GK
   immunostaining in transfections with human (hP, hL) or mouse (mP, mL)
   GKRP at 10 × 10^6 pfu/ml at endogenous GK (upper panels) or with GK
   overexpression (+GKOE). G) Lower nuclear sequestration (N/C,
   nuclear/cytoplasmic ratio) of GKRP:446 L in hepatocytes expressing
   endogenous GK, n = 15–16 hepatocyte experiments. H) Greater nuclear
   sequestration of endogenous GK with human compared with mouse 446 P.
   n = 7; #P < 0.05, 446 L vs 446 P; $P < 0.05 human vs mouse (G,H). I)
   Nuclear sequestration of GKRP:446 P or 446 L is increased by GKOE,
   n = 7. J) Nuclear sequestration of GK during GKOE is lower with
   GKRP:446 L, n = 6; #P < 0.05, 446 L vs 446 P; ∗P < 0.05 effect of GKOE
   (I,J). K-M) Comparison of GKRP expressivity (N + C intensity) and
   nuclear GKRP sequestration (N/C ratio), in a representative experiment
   (−/+ GKOE), with data points representing individual hepatocytes (K,L)
   or mean value of fields (M,N). $ P < 0.05 human vs mouse; #P < 0.05,
   446 L vs 446 P; ∗P < 0.05, 10 × 10^6 vs 5 × 10^6 pfu/ml; ∼P < 0.05
   effect of GKOE (M,N). K) At endogenous GK, with GKRP expressed at
   5 × 10^6 or 10 × 10^6 pfu/ml, expressivity of GKRP:446 L (N + C), was
   compromised at 10 × 10^6 pfu/ml, but nuclear sequestration (N/C) of
   446 L was lower at both 5 × 10^6 or 10 × 10^6 pfu/ml. L) With GKOE,
   GKRP:446 L expressivity was not compromised at 10 × 10^6 pfu/ml and
   GKRP:446 L N/C was lower at both titres. M) GKRP N/C summary for
   (K)–(L): showing greater N/C for human vs mouse GKRP ($) and lower N/C
   for 446 L (#). N) GK N/C summary showing greater N/C for human vs mouse
   and lower N/C for 446 L vs 446 P (#) with GKOE.

2.8. GK activity in mouse liver and hepatocyte monolayers

   Liver samples (40–80 mg) were homogenized in 15 volumes of 100 mM KCl,
   50 mM HEPES, 1 mM EDTA, 2.5 mM DTT, pH 7.5 and centrifuged at 100,000 g
   for 45min [[100]31]. The microsomal pellet and floating lipid layer
   were discarded and assays for low-K[m] hexokinase and total hexokinase
   activity were performed on 5 μl of 100,000 g supernatant in an assay
   containing 100 mM KCl, 50 mM HEPES, 2 mM MgCl[2], 5 mM ATP-Mg^2+,
   0.5 mM NAD, 2.5 mM DTT, 8U/ml glucose 6-phosphate dehydrogenase (Sigma,
   G8404), pH 7.8 and 0.5 mM or 100 mM glucose (340 nm, Spectramax,
   Molecular Devices). GK activity (Units/g, representing μmol/min per g)
   was determined by subtraction of low-K[m] hexokinase [[101]31].
   Hepatocytes were extracted and assayed as in [[102]16,[103]17].

2.9. GKRP and GK immunostaining in human and mouse liver

   Human liver and P446L mouse liver were immunostained using the Ventana
   automated system (Roche, BenchMark XT) with the standard retrieval
   protocol. For GKRP immunostaining a goat antibody (SC-6340 at 1:800
   dilution) with anti-goat-HRP OMNIMAP (Roche, 6607233001) and for GK
   immunostaining either of two rabbit polyclonal antibodies (SC-7908,
   1:100; PT-159629, 1:800) were used with anti-rabbit-HRP OMNIMAP (Roche,
   5269679001). HMGCR immunostaining was with a rabbit antibody (PT
   13533-1-AP, 1:500). Negative controls were with goat IgG (Sigma, 5256)
   or rabbit IgG (Sigma, I5006).

   The slides were scanned x20 magnification with Aperio Image Software
   (Leica Biosystems). The scans were annotated for stained regions of
   interest. Quantitative image analysis of the annotated regions was
   performed using Aperio Brightfield Image Analysis Toolbox Software
   (Leica Biosystems) for nuclear H-scores and cytoplasmic H-scores (GKRP
   and GK) or cytoplasmic H-scores (HMGCR). For GKRP and GK the Aperio
   Software default thresholds were used for nuclear staining (200, 160,
   120 pixel intensity) and cytoplasmic staining (210, 180, 150 pixel
   intensity). For HMGCR the cytoplasmic thresholds were set at 235, 230,
   225. The data was exported into MicroSoft Excel for further analysis.
   The quantitative analysis data exported comprised the total numbers of
   hepatocyte nuclei and surrounding cytoplasmic regions within the
   annotated regions and the percentages corresponding to the 3 intensity
   thresholds. Nuclear and cytoplasmic H-scores were determined according
   to the formula: ([% weak staining] + [% moderate staining x 2] + [%
   strong staining x 3]) yielding a range of 0–300.

2.10. Histopathology scoring of mouse liver

   Mouse liver sections were stained with H&E and SFRG. Histopathology was
   assessed for steatosis, lobular inflammation, hepatocyte ballooning and
   fibrosis according to the NASH score [[104]28] by an expert liver
   pathologist (DT), blinded to genotype. Other histological features
   including portal inflammation and lipogranulomas were recorded.
   Semi-quantitative evaluation of sinusoidal fibrosis was based on a
   3-tiered system (0–2) where 0 is none, 1 is focal (zone 3), and 2 is
   extensive (zone 3 and beyond).

2.11. Western blot GKRP and GK analysis in mouse liver and hepatocytes

   Frozen liver or mouse hepatocytes were lysed in 100 mM NaCl, 25 mM NaF,
   2 mM EDTA, 0.1 mM Na[3]VO[4], 100 mM Tris–HCl pH 7.4, 0.1% Triton
   X-100, 1 mM benzamidine, protease inhibitors (Sigma P8340), sonicated
   (MSE-Soniprep 150), centrifuged (14,000 g, 10 min) and supernatant
   protein determined (Bio-Rad #5000006). Proteins (10–40 μg) were
   resolved by SDS-PAGE on 10% SDS or 4–12% SDS(Bio-Rad 456-8093) gels,
   electrotransferred to PVDF membrane and immunoblotted (GK, Proteintech
   PT15629; GKRP, rabbit-AZ680) followed by horse radish peroxidase
   conjugated anti-rabbit IgG (1:5000 in 5% BSA/TBST) and developed with
   Enhanced Chemiluminescence (Perkin Elmer) and exposed to X-ray film.
   Band densitometry was determined using ImageJ.

2.12. RNA extraction and RT-qPCR

   Frozen liver or mouse hepatocytes were extracted in Trizol and the RNA
   was DNase treated (Sigma, 04716728001). cDNA was synthesized from total
   RNA (1 μg) using MMLV (Promega M1705) and amplified using GoTaq Probe
   qPCR Master Mix (Promega A6002) with the primers indicated
   ([105]Supplementary Table-1). Relative gene expression was determined
   using the delta–delta Cycle threshold. For liver, transcripts were
   normalized to RplpO and for mouse hepatocytes (human GCKR and mouse
   Gckr) to Gapdh.

2.13. RNA extraction and RNA-sequencing

   RNA was extracted from mouse livers (n = 74) and hepatocyte cultures
   (n = 36) using Trizol and purified using RNeasy kit (Qiagen #74104)
   with on-column DNase digestion (Qiagen #79254). Samples were submitted
   to the Newcastle University Genomics Core for quality assessment and
   samples with a RIN value > 6 were used for sequencing. RNA-Seq sample
   quality was assessed via FastQC and visualized with MultiQC. All
   samples had quality scores >35, no adapter contamination, and therefore
   none required trimming. Transcript-level counts were obtained using
   Salmon run against the mouse reference transcript sequences (release
   M26, GRCm39) from Gencode (gencodegenes.org/mouse/). Differential gene
   expression was determined using the DESeq2 package in R via RStudio.
   Differentially expressed genes (DEGs) were defined as an adjusted
   P-value (FDR) <0.05 combined with a fold-change increase or decrease of
   30%. One mouse, wild-type #7 (of 10), was removed from all analyses
   following a heatmap showing that its gene expression differed from all
   other mice. Pair-wise comparison of the different genotypes were
   carried out to determine gene expression changes.

2.14. Blood and tissue analysis

   Blood glucose was determined by tail vein sampling using a glucose
   meter (Roche, Accu-chek) and plasma insulin by ELISA (Mercodia,
   #10-1247-01). For oral glucose tolerance tests mice were fasted for
   2 h, gavaged with glucose (2 g/kg body wt) and tail vein blood sampled
   at time 0 (before gavage) and intervals indicated. Glucose area under
   the curve (AUC) was corrected for baseline at time zero. For insulin
   tolerance tests mice were fasted for 5 h and injected intraperitoneally
   with insulin (Actrapid, 1Unit/kg body wt) and tail vein blood sampled
   for glucose at time 0 (before insulin) and intervals indicated. For
   lipid analysis, blood was collected into heparinized tubes and
   sedimented 10,000 g for 10 min. Plasma triglycerides (GPO-POD), total
   cholesterol (CHO-POD), LDL-cholesterol (CHO-PAP), HDL cholesterol,
   inorganic phosphate (Molybdate) were determined on the AU680 Analyzer
   at Harwell, Mary Lyon Centre Pathology. Liver lipids were extracted in
   chloroform/methanol (1:2) and the lipid extract in the chloroform
   fraction was dried was reconstituted in isopropanol. Liver triglyceride
   was assayed with WAKO-triglyceride kit (Alpha Laboratories, 290–63701)
   and cholesterol with a Cholesterol Kit (Cambridge BioSciences,
   CAY10007640).

2.15. Statistical analysis

   Statistical analysis was completed using Prism 8 Software (GraphPad
   Software Inc) or SPSS (v27). Data is shown as mean ± standard error of
   the mean, except for the gene counts for RNA-seq (which are median and
   range) or associations and histopathology which are categorical scores.
   Statistical comparisons between groups of continuous variables was
   assessed by the two-tail t-test for parametric data and for
   histopathology by Chi-square test for categorical variables.
   Statistical significance is set at 0.05 unless otherwise stated.

3. Results

3.1. Expression of human or mouse GKRP (446P or 446L) in GKRP-deficient mouse
hepatocytes

3.1.1. Nuclear GKRP sequestration for mouse Gckr-transcript-1 but not
Gckr-transcript-2

   We used adenoviral vectors to express human or mouse GKRP:446 P or
   446 L in hepatocytes isolated from GKRP-deficient mice
   ([106]Figure S1A). In hepatocytes not transfected with vectors there
   was no detectable nuclear staining for GKRP or GK ([107]Figure S1B).
   Transfection with human GKRP:446 P or 446 L at adenoviral titres of
   10–30 × 10^6 pfu/ml resulted in intense nuclear staining for GKRP and
   the endogenous GK accumulated in the nucleus ([108]Figures. S1C and D).
   The GKRP nuclear/cytoplasmic (N/C) intensity ratio was greatest at the
   lowest titre of 10 × 10^6 pfu/ml ([109]Figure. S1E), indicating
   accumulation of GKRP in the cytoplasm at higher titres. Transfection
   with vectors for mouse GKRP-T2 ([110]NM_144909.2), previously
   designated canonical [[111]32] showed exclusively cytoplasmic GKRP and
   GK staining ([112]Figures. S1F and G). Transfection with GKRP-T1
   ([113]NM_001374741.1, previously designated Gckr-X1, which shares 87.5%
   identity with human GKRP, showed nuclear staining for GKRP and GK
   similar to human GKRP ([114]Figure 1H vs 1D). Mouse GKRP-T2, which has
   a 108bp deletion in exon-17, and two other mouse Gckr transcripts:
   XM-006503882; XM-006503883, now designated X1, X2 also encode
   exclusively cytoplasmic proteins ([115]Figure. S1I). Using primers
   spanning deletions of T2, X1, X2 we confirmed that Gckr-T1 is the
   predominant transcript and that Gckr-T2 is not expressed in mouse liver
   ([116]Figures. S2A–C). Amino acid alignment of human GKRP, rat GKRP and
   mouse GKRP isoforms encoded by Gckr-T1 (Isoform-1); Gckr-T2
   (Isoform-2), Gckr-X1 (X1) and Gckr-X2 (X2), shows that Isoform-2
   expressed in [[117]32] encodes a protein missing residues 489–524,
   whereas X1, skipping exon-7 is missing residues 166 to 183 and X2,
   skipping exon-2 is missing residues 21 to 72 ([118]Figure. S2D). It is
   noteworthy that only Isoform-1 localizes to the nucleus in hepatocytes
   ([119]Figure. S1).

   GKRP-T1 (Isoform-1) was used in the rest of this study. We next used
   adenoviral titres of 5 × 10^6 and 10 × 10^6 pfu/ml, which have similar
   cellular transfection efficiency ([120]Figure S1J) for comparison of
   human GKRP and mouse GKRP-T1 (446 P/L) and data from these 2 titres is
   either pooled ([121]Figure-1A) or presented separately
   ([122]Figure-1B). Gckr mRNA expression determined by RT-qPCR using
   primers for mouse or human GCKR confirmed equal expression at mRNA
   level for GKRP:446 P compared with 446 L ([123]Figure. S1K).

   In GKRP-deficient hepatocytes transfected with human GKRP:446 P, the
   GKRP immunoreactivity for the pooled data of 5 × 10^6 and
   10 × 10^6 pfu/ml titres was comparable to GKRP immunoactivity in
   wild-type mouse hepatocytes ([124]Figure 1A). However, the endogenous
   GK protein level was lower than in wild-type hepatocytes, as found
   previously for other GKRP-deficient mouse models [[125]33,[126]34].
   Consequently, the GKRP-to-GK ratio was higher than in wild-type
   hepatocytes ([127]Figure 1A). In the rest of the study GK was expressed
   by ∼2–3 fold above endogenous where indicated to attain GKRP-to-GK
   ratios spanning the physiological range.

3.1.2. Lower expressivity of GKRP:446L protein and the stabilizing effect of
GK

   Comparison of GKRP immunoreactivity in transfections at 5 × 10^6 or
   10 × 10^6 pfu/ml showed lower protein levels for 446L compared with
   446P, for both mouse and human GKRP, at the higher titre
   (10 × 10^6 pfu/ml), with smaller differences at the lower titre
   ([128]Figure 1B), despite similar mRNA expression for 446L
   ([129]Figure. S1K). To test for differences in protein stability, the
   protein synthesis inhibitor cycloheximide (CX), was added for the last
   6 h ([130]Figure 1C). There was a greater effect of cycloheximide on
   human GKRP (446P and 446L), indicating lower stability than for mouse
   GKRP and the lower immunoactivity of mouse 446 L was modestly
   accentuated (+CX 29% vs no CX 18%).

   We next tested the effect of GK overexpression by ≥ 2-fold
   ([131]Figure 1D). This showed a stabilizing effect of GK on human GKRP
   (446P and 446L) and on mouse GKRP:446L immunoactivity and the lower
   expressivity of both 446 L variants was attenuated.

   A putative role for covalent modification of GKRP by P300
   acetyltransferase and Sirt2 deacetylase on GKRP protein stability and
   interaction with GK has been reported [[132]35,[133]36]. In COS-1 cells
   transfected with the GKRP-adenoviral vectors, GKRP immunoactivity was
   increased with the Sirt2 inhibitors, AGK2 and nicotinamide (NAM), and
   decreased by SIRT2 overexpression ([134]Figure. S3) consistent with a
   putative stabilizing role for acetylation [[135]36]. When AGK2 was
   tested in hepatocytes transfected with the GKRP-vectors at
   5 × 10^6 pfu/ml, it increased GKRP immunoactivity for both human and
   mouse 446L and attenuated the compromised expressivity of the 446L
   variant ([136]Figure 1E). Cumulatively, this shows lower expressivity
   of the 446L variant compared with 446P for both human and mouse GKRP
   that is greater at higher GKRP-to-GK ratios. The lower 446L
   expressivity is attenuated by GK overexpression and by Sirt2 inhibition
   consistent with a stabilizing effect of the interaction with GK.

3.1.3. Nuclear sequestration of both GKRP:446P and 446L is dependent on the
GK protein level

   In GKRP-deficient hepatocytes transfected with adenoviral vectors for
   human or mouse GKRP at 5 × 10^6 or 10 × 10^6 pfu/ml there was large
   intercellular heterogeneity in the nuclear/cytoplasmic (N/C) staining
   intensity for GKRP and GK ([137]Figure 1F), as occurs for endogenous
   GKRP and GK in hepatocytes [[138]37]. To test for differences in
   nuclear sequestration of the 446 L variant, the N/C mean pixel
   intensity ratio was determined from 60 to 100 nuclei for each
   transfection condition in each hepatocyte experiment. The data for
   experiments with only endogenous GK is summarized in [139]Figure 1G,H.
   Nuclear sequestration of GKRP:446 P was higher (18%) for human GKRP
   compared with mouse GKRP and it was lower for the 446 L variant by 17%
   for human and by 11% for mouse GKRP ([140]Figure 1G) with similar but
   smaller trends for endogenous GK ([141]Figure 1H).

   Overexpression of GK increased the sequestration of GKRP in the nucleus
   for both 446 P and 446 L by 34–37% for human GKRP and by 54–61% for
   mouse GKRP ([142]Figure 1I), indicating that increased GK protein
   promotes nuclear GKRP sequestration. When GKRP (446 P or 446 L) was
   expressed in heterologous cell lines (FTO2B hepatoma and HeLa cells)
   which lack endogenous GK, and without or with adenoviral-mediated GK
   expression, the sequestration of GKRP in the nucleus was increased ∼2
   fold by GK expression ([143]Figure. S4), confirming a role for GK in
   sequestering GKRP in the nucleus.

3.1.4. Lower nuclear GK sequestration by GKRP-446L

   In hepatocytes overexpressing GK by ∼2-fold above endogenous
   ([144]Figure 1D), the nuclear accumulation of GK ([145]Figure 1 J) was
   increased ∼2-fold, if cells expressed GKRP-446 P (human 1.7 ± 0.17,
   mouse 1.98 ± 0.23 fold,∗P < 0.02), but negligibly for GKRP-446 L (human
   1.04 ± 0.10, mouse 1.31 ± 0.16) and nuclear GK accumulation was
   significantly lower (32–35%, #P < 0.03) with GKRP-446 L indicating a
   lower affinity for GK of human and mouse GKRP:446 L.

3.1.5. Compromised nuclear GKRP-446L sequestration-independent of lower
expressivity

   To test whether the lower nuclear sequestration of GKRP-446L
   ([146]Figure 1G,I) is explained by its lower expressivity
   ([147]Figure 1B), we compared nuclear sequestration (N/C) and cellular
   expression (N + C) by immunofluorescent staining at adenoviral titres
   of 5 × 10^6 or 10 × 10^6 pfu/ml ([148]Figure 1K,L). In hepatocytes
   expressing only endogenous GK ([149]Figure 1K), total cell intensity of
   the 446L variant was comparable to 446P at the lower titre
   (5 × 10^6 pfu/ml) but not at the higher titre ([150]Figure 1K upper
   panel). However, the N/C ratio was lower for 446 L at both titres
   ([151]Figure 1K, lower panel), indicating that it is in part
   independent of compromised expressivity. With GK overexpression
   ([152]Figure 1L) total GKRP cell intensity for 446L was comparable to
   446P at both titres ([153]Figure 1L-upper panel), consistent with
   GK-mediated stabilization and the GKRP N/C ratios were elevated
   compared with endogenous GK alone (6–10 vs 3–6) and were lower for
   446 L. [154]Figure 1M summarizes the GKRP N/C ratios and the
   corresponding GK N/C ratios are shown in [155]Figure 1N. At endogenous
   GK, nuclear GK sequestration was greater for human than mouse GKRP at
   both titres ([156]Figure 1N), consistent with a higher affinity for GK
   of human GKRP [[157]25].

   Cumulatively, the results support the following conclusions: i) nuclear
   GKRP sequestration is enhanced by a higher GK-to-GKRP ratio
   ([158]Figure 1I, [159]Figure S4); ii) human GKRP has a higher affinity
   for GK than mouse GKRP as shown by the greater nuclear sequestration of
   endogenous GK ([160]Figure 1H); iii) GKRP-446 L has a lower affinity
   for GK as shown by the greater sequestration of overexpressed GK in the
   nucleus with 446P ([161]Figure 1J); iv) lower expressivity of GKRP-446L
   by immunostaining is apparent at higher GKRP-to-GK ratios
   ([162]Figure 1K) and concurs with the immunoblotting
   ([163]Figure 1B-D); v) lower sequestration of GKRP:446L in the nucleus
   is in part independent of the lower expressivity ([164]Figure 1K) and
   best explained by its lower affinity for GK; vi) differences between
   the 446L variant and GKRP-446P are qualitatively similar for human and
   mouse GKRP.

3.1.6. Glucose-dependent GK translocation and rates of glucose
phosphorylation

   Nuclear GK sequestration in rat hepatocytes is glucose-dependent and
   maximal at 5 mM glucose with half-maximal translocation to the
   cytoplasm at 10–20 mM glucose depending on the GK-to-GKRP ratio
   [[165]11,[166]17]. In hepatocytes from wild-type mice
   glucose-dependence of GK translocation ([167]Figure. S5A) concurred
   with previous findings in rat hepatocytes. Because nuclear GKRP
   sequestration is dependent on the GK-to-GKRP ratio ([168]Figure 1I) we
   tested whether glucose (10–35 mM) affects nuclear GKRP:446 P/L
   sequestration. There was no significant effect of 10–35 mM glucose on
   nuclear sequestration of human or mouse GKRP-446 P/L ([169]Figure 2A).
   However, translocation of GK from the nucleus was induced at 35 mM
   glucose in hepatocytes expressing human GKRP:446 P or 446 L (22% or
   25%, respectively) and at ≥ 20 mM glucose or ≥ 10 mM glucose in
   hepatocytes expressing mouse GKRP:446 P or 446 L, respectively
   ([170]Figure 2B–C). Translocation at lower glucose for mouse GKRP,
   concurs with the higher affinity of human GKRP for GK. Rates of glucose
   phosphorylation were inhibited by GKRP overexpression with differences
   between 446 L and 446 P manifesting at adenoviral titres
   ≥10 × 10^6 pfu/ml ([171]Figures. S5B and C), indicating that they are
   in part due to compromised GKRP-446 L expressivity.

3.1.7. Transcriptome analysis of gene counter-regulation by GK excess and by
GKRP:P446 > L

   To test for functional differences between GKRP:446 P and 446 L we
   performed RNA-sequencing of hepatocytes expressing human or mouse GKRP
   (446 P or 446 L) at either endogenous GK or overexpressed GK (10
   conditions) after 4 h incubation with 25 mM glucose ([172]Figure 2D,E).
   GK overexpression caused significant induction of 55 genes and
   repression of 26 genes ([173]Figure 2F; [174]Figure. S5D) from
   comparison of GK overexpression alone with the 5 combined groups of
   endogenous GK (untreated −/+ GKRP-hP,hL,mP,mL). This analysis
   identified fewer differentially expressed genes (DEGs) than comparison
   with the untreated group alone but minimizes non-specific genes linked
   to the adenoviral vector. The genes up-regulated by GK overexpression
   included candidate ChREBP-target genes [[175][38], [176][39],
   [177][40]] such as Pklr, G6pc, Rgs16, Aldob, Klf10, Arrdc4, Slc2a2 and
   Rorc and other genes not previously identified as ChREBP targets such
   as Slc16a1, the lactate and pyruvate transporter and 3 mitochondrial
   transporters of the Slc25 family including Slc25a44, which transports
   branched-chain amino acids and ubiquinone precursors [[178]41] and
   Slc25a47, which is expressed predominantly in hepatocytes [[179]42].
   The genes down-regulated by GK overexpression included Hmgcr
   ([180]Figure. S5D). Eight genes were validated by RT-qPCR ([181]Figure.
   S5E).

   Expression of human or mouse GKRP-446 P/L in cells overexpressing GK,
   repressed around 60 genes of which 20 were up-regulated by GK
   overexpression ([182]Figure 2G). The gene counts are shown for 14 of
   the 20 genes and for Hmgcr ([183]Figure 2H). There was heterogeneity of
   response (446 P vs 446 L) with some genes showing similar repression by
   446 P and 446 L (Lhx6), others stronger repression by 446 P (Angptl8,
   Slc25a47) or minimal repression by mouse 446 L (Pklr, Tkfc).
   Cumulatively, this identifies ChREBP-target genes and novel genes
   conversely regulated by GK and GKRP and shows weaker counter-regulation
   by GKRP-446 L that is more pronounced for mouse GKRP consistent with
   greater glucose-induced GK translocation ([184]Figure 2C).

3.2. Lower GKRP and GK protein in GCKR rs1260326-446LL liver and in the P446L
mouse

3.2.1. Lower GKRP and GK protein in human liver homozygous for GCKR
rs1260326-446 L

   Human liver biopsies from a nonalcoholic fatty liver disease (NAFLD)
   study [[185]9] that had been genotyped for GCKR rs780094 (n = 48) were
   additionally genotyped for rs1260326^C > T, with 23 and 17 confirmed CC
   and TT, respectively. Of these 11 (CC) and 14 (TT) had a NASH
   (Nonalcoholic steatohepatitis) diagnosis according to the Clinical
   Research Network [[186]28]. Immunohistochemical staining for GKRP and
   GK showed predominant hepatocyte nuclear staining for both genotypes
   ([187]Figure 3A). Nuclear immunostaining intensity was assessed from
   the H-score (N–H), a semi-quantitative measure (scale 0 to 300) of the
   % of cells with high (x3) plus medium (x2) plus weak intensity. For
   GKRP immunostaining, the TT genotype had lower N–H scores in the entire
   data set ([188]Figure 3B, P < 0.00001), and in steatosis and NASH
   groups ([189]Figure 3C, P < 0.004; P < 0.0001) and the cytoplasmic
   H-scores (C–H) were also lower for the TT genotype ([190]Figure 3D,
   P < 0.001), showing lower overall GKRP protein by TT genotype.
   Likewise, for GK immunostaining, the TT genotype had lower N–H scores
   in the entire group ([191]Figure 3E; P < 0.000001) and in steatosis and
   NASH groups ([192]Figure 3F) and also lower C–H scores ([193]Figure 3G)
   indicating lower GK protein by TT genotype. The GK N–H scores for the
   entire set (CC and TT) correlated with the GKRP N–H scores
   ([194]Figure 3H, P < 0.001) but with different slopes for the 2
   genotypes (CC, 0.40; TT 0.08) and with a significantly higher GK
   N–H/GKRP N–H ratio for the CC genotype ([195]Figure 3I; CC, 0.47 ± 0.04
   vs TT, 0.26 ± 0.04, P < 0.0003). This concurs with the lower nuclear GK
   sequestration in hepatocytes transfected with GKRP-446 L
   ([196]Figure 1J).

Figure 3.

   [197]Figure 3
   [198]Open in a new tab

   Lower GKRP and GK protein levels in human NAFLD biopsies of the
   rs1260326-TT genotype and in the P446L mouse. A-I) Human NAFLD biopsies
   genotyped for GCKR rs1260326 variant (CC n = 23; TT n = 17) and scored
   as simple steatosis (NAFL) or NASH were stained for GKRP or GK and
   nuclear (N–H) and cytoplasmic (C–H) H-scores were determined. (A)
   representative images. (B–D) Lower GKRP N–H scores by TT genotype for
   entire cohort (B) or by −/+ NASH diagnosis (C), lower C–H scores (D)
   (E–G) Lower GK N–H scores by TT genotype for entire cohort (E) or
   by −/+ NASH diagnosis (F) and lower C–H scores (G) H) Correlation of GK
   N–H vs GKRP N–H, r^2 0.34, P < 0.0001 for entire data set. I) Lower GK
   (N–H)/GKRP (N–H) ratio by TT genotype, #P < 0.0003. J-M) GKRP P446L
   mouse aged 14-wk (regular diet, n = 4,4, PP, LL) or 28-wk (on HFHSD for
   last 20-wk, n = 10,10, PP, LL): Liver Gckr and Gck mRNA normalized to
   RplpO (J); GKRP and GK protein (K–L) and GK enzyme activity (M). J)
   Similar Gckr mRNA by genotype but lower Gck mRNA in 28-wk mice by LL
   genotype. K-L) Lower GKRP and GK immunoactive protein (n = 4–8) and
   lower GK activity (M, n = 4,4,10,10,10) by LL genotype. #P < 0.05, LL
   vs PP; ∗P < 0.05 28-wk vs 14 wk. N-R) Comparative GKRP and GK
   immunostaining of human liver biopsies (CC,TT: n = 7,6) and P446L mouse
   liver (PP, PL, LL: n = 7,4,5) showing representative images for mouse
   (N) and the N–H and C–H scores. O) Lower GKRP N–H scores for human
   compared with mouse ($) liver and by TT and LL genotypes (#) and
   negligible cytoplasmic H-scores. P) Lower GK N–H and C–H for human
   compared with mouse liver ($) and lower N–H by TT (human) or LL (mouse)
   genotype (#). Q) Correlation of GK and GKRP N–H scores. R) Relative
   distribution of GK in nucleus and cytoplasm (N–H/C–H ratio) showing
   lower nuclear GK sequestration by TT and LL genotypes. #P < 0.05 by
   genotype; $P < 0.05 human vs. mouse.

3.2.2. Lower GKRP and GK protein in the GKRP:P446L mouse

   Generation of the GKRP:P446L mouse has been reported [[199]26]. At
   14-wk of age Gckr and Gck mRNA levels in LL mice were similar to
   wild-type (PP) litter mates for male ([200]Figure 3J) and female mice
   (not shown). However GKRP immunoactivity was very low in LL mice and GK
   immunoactivity and GK enzyme activity were ∼40% lower
   ([201]Figure 3K–M). In older mice (28-wk) that were maintained on a
   high-fat high-sugar diet (HFHSD) for 20 wk, GKRP and GK immunoactivity
   and GK enzyme activity were higher (∼1.4, 1.7 and 2-fold, respectively
   28-wk-PP vs 14-wk-PP), but remained lower by LL genotype
   ([202]Figure 3K–M).

3.2.3. Comparative GKRP and GK immunostaining in human liver and in the P446L
mouse

   Concurrent immunostaining of P446L mouse liver (PP, PL, LL, genotypes
   28-wk) and human liver (rs1260326: CC,TT, n = 7,7) enabled comparison
   of GKRP protein levels by species ([203]Figure 3N–R). The mouse GKRP
   N–H scores were lower in PL and LL genotypes than wild-type
   ([204]Figure 3O) and the relative values for LL and PP concurred with
   the GKRP immunoblotting (23% vs 19%, [205]Figure 3J). Cytoplasmic GKRP
   immunostaining was very low in mouse liver with N–H/C–H ratios >300.
   For mouse GK immunostaining, the N–H scores of PL and LL genotypes (78%
   and 18% of PP) showed similar trends as GKRP N–H scores
   ([206]Figure 3P) and correlated with GKRP N–H scores ([207]Figure 3Q).
   Cytoplasmic GK immunostaining was high in mouse liver and is estimated
   at ∼60% of total GK for PP liver based on the GK activity differences
   by genotype ([208]Figure 3M). Estimates for GK nuclear/cytoplasmic
   distribution from the N–H/C–H ratio (which is an arbitrary ratio
   because of lower intensity thresholds for C–H than N–H) showed lower GK
   nuclear sequestration by LL genotype ([209]Figure 3R).

   For human CC livers (446 PP), N–H scores for both GKRP and GK were
   lower (∼50%) than for mouse PP liver ([210]Figure 3O,P) indicating
   lower protein levels in human liver. However, nuclear sequestration of
   GK (N–H/C–H) was similar or higher in human liver ([211]Figure 3R) and
   similar results were obtained using a different GK antibody (PT15629,
   not shown). Cumulatively, this shows lower nuclear GK sequestration by
   446LL genotype for both human and P446L mouse liver. Additionally,
   human liver has lower GKRP protein level compared with mouse but
   similar or higher nuclear GK sequestration consistent with the higher
   affinity for GK of human GKRP.

3.3. Liver transcriptome of the P446L mouse on regular diet

   GKRP-P446L mice on regular rodent diet were fertile with no overt
   abnormalities. Breeding of PL mice produced PP, PL and LL genotypes in
   Mendelian ratios. In male mice, despite the lower GKRP and GK protein
   abundance by LL genotype ([212]Figure 3K), blood glucose, insulin and
   triglyceride levels were not different ([213]Figure 4A). In female
   mice, which had higher GK activity than male mice (2.4 ± 0.13 vs
   1.5 ± 0.08, U/g, n = 5,4) for the PP genotype, but similar reduced GK
   activity for the LL genotype (0.9 ± 0.13 vs 0.9 ± 0.13 U/g, n = 5,4)
   blood triglycerides were higher in LL mice but liver triglycerides and
   glucose tolerance were not different ([214]Figure 4B).

Figure 4.

   [215]Figure 4
   [216]Open in a new tab

   Blood metabolites and liver transcriptome in the P446L mouse on regular
   diet. A) Male mice aged 14 wk: body weight, free-feeding blood glucose,
   plasma insulin and plasma triglyceride; n = 6–9. B) Female mice aged
   14 wk: body weight; plasma triglycerides; liver triglycerides,
   n = 15–17 and glucose tolerance, n = 7,7. C) Heat map of differentially
   expressed genes (DEGs, FDR <0.05) in livers of male mice by LL genotype
   and the direction of the log fold change (LFC) of expression values. D)
   Top pathways from the Ingenuity Pathway Analysis showing corresponding
   P value and Z scores. E) Venn diagram showing numbers of common genes
   that are significantly differentially expressed (upregulation, red;
   down-regulation green) in male mice by LL genotype and are
   differentially expressed in female PP vs male PP mice. F) Gene counts
   for P446L male and female mice for common genes from E). G) Box plot of
   gene counts from RNA-sequencing for 12 Mup genes showing lower
   expression in male LL mice and in female PP and LL mice; n = 4,4,5,5,
   (M−P, M-L, F–P, F-L)

   To assess hepatic adaptations we determined the liver transcriptome of
   male and female mice (PP and LL, 14-wk) by RNA-sequencing. Unsupervised
   clustering analysis identified distinct groups by PP and LL genotypes.
   For the male mice there were 45 differentially expressed genes (DEGs,
   FDR <0.05) by LL genotype ([217]Figure 4C) and Enrichment Gene Ontology
   (EnrichGO) and Ingenuity Pathway Analysis (IPA) identified lipid
   metabolism as the main enriched pathways or processes ([218]Figure 4D).
   Six of 24 upregulated genes by LL genotype (Cyp17a1, Ehhadh, Scd1,
   Enpep, Lepr, Tns3) corresponded to previously identified ChREBP-target
   genes in fasted-refed mice [[219]40] or high-glucose challenged
   hepatocytes or HepG2 [[220]38,[221]39]. In female mice the DEGs by LL
   genotype did not share common genes with the male DEGs nor did they
   show significant pathway enrichment by EnrichGO or IPA. Comparison of
   the male LL DEGs with differential gene expression in female mice
   showed that several of male-LL, upregulated (17 of 24) and
   downregulated (9 of 21) genes ([222]Figure 4E) were expressed at higher
   or lower levels, respectively, in the female mice ([223]Figure 4F),
   indicating that sexual dimorphic gene expression in part explains the
   lack of common DEGs.

   Comparison of the DEGs by GK or GKRP overexpression from the hepatocyte
   study ([224]Figure 2) with DEGs by LL genotype in male mice did not
   identify common genes linked to GK overexpression and identified one
   gene (Mup18) that was up-regulated by GKRP expression in hepatocytes.
   Mup18 which was expressed at lower levels in LL male mice
   ([225]Figure 4C), is one of several major urinary protein (MUP) genes
   comprising ∼21 coding and ∼30 non-coding genes [[226]43]. MUPs are
   species-specific ligand-binding 19 kDa proteins of the lipocalin family
   produced in liver and excreted in urine with roles in communication
   behaviour and glucose and lipid metabolism [[227]43]. Gene counts for
   21 Mup genes averaged 43%, 13% and 6% in male-LL, female-PP and
   female-LL respectively of male PP mice. A sub-set of these Mup genes is
   shown in ([228]Figure 4G).

   Cumulatively, the liver transcriptome of male P446L mice on rodent diet
   identified some ChREBP-target genes and down-regulation of Mups but
   there were no common genes with the GK overexpressed genes identified
   from the hepatocyte study.

3.4. Metabolic adaptations in P446L mouse hepatocytes: attenuated glucose
metabolism at elevated glucose but not at 5 mM glucose

   GK functional activity in hepatocytes is a composite function of the
   intrinsic amount of GK protein, the molar ratio of GK-to-GKRP, the
   affinity of GKRP for GK and the intrinsic concentrations of glucose and
   allosteric effectors of GKRP. To assess functional GK activity in
   hepatocytes isolated from male P446L mice we compared rates of glucose
   phosphorylation and glycolysis and metabolite and gene responses to
   substrate challenge in hepatocytes isolated from 446LL to 446 PP mice
   after 24 h-culture ([229]Figure 5). The LL hepatocytes had very low
   (<20%) GKRP protein and ∼50% lower GK protein and total enzyme activity
   ([230]Figure 5A–D) and thereby a higher GK-to-GKRP molar protein ratio.
   Rates of glucose phosphorylation or glycolysis were identical in LL and
   PP hepatocytes at 5 mM glucose but lower in LL hepatocytes at 25–35 mM
   glucose ([231]Figure 5E,F). During challenge with high glucose (15 mM
   or 25 mM) −/+ a GK activator (PF-04991532, GKA) or inhibitor of glucose
   6-phosphatase (S4048) or with xylitol or fructose ([232]Figure 5G),
   which bypass the GK reaction [[233]44], basal cell ATP was similar in
   LL and PP hepatocytes ([234]Figure 5H) but glucose 6-P and glycerol 3-P
   were higher in PP hepatocytes incubated with high glucose, but not with
   xylitol or fructose which bypass the GK reaction ([235]Figure 5I,J).
   Cell ATP normalized to the respective 5 mM glucose controls, was
   attenuated by xylitol and fructose in both genotypes and by high
   glucose + S4048 inhibitor only in PP cells, which had higher glucose
   6-P ([236]Figure 5K). Basal Gckr and Gck mRNA levels were similar in LL
   and PP hepatocytes and likewise mRNA levels for ChREBP-β and several
   target genes except for G6pc which was higher in LL-hepatocytes
   ([237]Figure 5L). The induction by high glucose challenge of
   metabolite-responsive genes including ChREBP-β, Txnip, Pklr, G6pc and
   Fgf21 paralleled the higher glucose 6-P and glycerol 3-P accumulation
   in the wild-type hepatocytes ([238]Figure 5M–S), whereas the effect of
   xylitol which bypasses the GK reaction and causes a larger elevation in
   glycerol 3-P than high glucose ([239]Figure 5J) showed opposite trends
   by genotype on gene expression. It is noteworthy that glycerol 3-P in
   incubations with xylitol is measured as a surrogate for changes in
   phosphate ester intermediates of the pentose cycle including xylulose
   5-P which show parallel changes to glycerol 3-P [[240]45].

Figure 5.

   [241]Figure 5
   [242]Open in a new tab

   Functional characterization of glucose metabolism and adaptive response
   to substrate challenge in hepatocytes from P446L mice. A-D) GKRP and GK
   immunoactivity and enzyme activity after 24 h-culture of hepatocytes
   isolated from male P446L mice (PP and LL genotypes, aged 8–12 wk). (C)
   Representative immunoblots: protein loading GK 30 μg; GKRP 10 μg
   (446 PP); 35 μg (446LL); D. Glucokinase activity munits/mg protein
   (n = 4–5). E-F) Glucose phosphorylation ([2–^3H]glucose, E) and
   glycolysis ([2–^3H]glucose, F) is similar at 5 mM glucose but lower by
   LL genotype at high glucose, (n = 5,5). G) Metabolic scheme showing
   metabolism of xylitol and fructose bypass the GK reaction and raises
   glycerol 3-P (G3P); whereas pharmacological activation of GK with
   PF-04991532 (PF) or inhibition of glucose 6-phosphatase (G6pc) with
   S4048 (S4) raises glucose 6-P (G6P). Note that fructose at micromolar
   concentrations also causes activation of GK via elevation in fructose
   1-P and thereby elevation in G6P. However at millimolar concentrations
   the increase in triose phosphates (G3P) is consequent to flux through
   ketohexokinase and aldolase B reactions. H–K) Hepatocyte incubations
   with glucose (15 mM or 25 mM) −/+ the GK activator PF-04991532 (PF,
   10 μM) or inhibitor of glucose 6-phosphatase (S4048, 1 μM) or xylitol
   (Xyl, 2 mM) or fructose (Fru, 10 mM) were for 1 h for determination of
   cell ATP (H), glucose 6-P (I), glycerol 3-P (J) expressed as nmol/mg
   protein or ATP normalized to respective 5 mM (K), n = 5,5. L-S)
   Hepatocyte incubations for RNA extraction and RT-qPCR analysis were for
   4 h. (L) Basal mRNA levels of genes indicated normalized to Gpd2 mRNA.
   (M–S) mRNA levels of the genes indicated normalised to the respective
   5 mM glucose control, n = 3,4; #P < 0.05 LL vs PP; ∗P < 0.05 relative
   5 mM glucose control.

   Cumulatively, hepatocytes from P446L mice on standard rodent diet had
   similar rates of glucose phosphorylation and glycolysis at basal
   glucose (5 mM) despite the higher GK-to-GKRP protein ratios and
   compromised binding affinity of GKRP for GK. But they had lower
   metabolic rates at elevated glucose and blunted metabolite accumulation
   and ChREBP-target induction. The similar rates at 5 mM glucose and
   blunted response at elevated glucose is best explained by compensatory
   adaptation of total GK activity and gene changes geared to basal
   glucose homeostasis.

3.5. Phenotype of the P446L mouse on high-energy diets

   To assess adaptations of the P446L mouse on high-energy diets, mice
   were fed either a high-fat diet (HFD) with 45% fat by energy or a
   high-fat high-sugar diet (HFHSD) with the same fat-diet and with sugar
   (glucose 10% + fructose 5%) in the drinking water in two consecutive
   studies ([243]Figure 6). Consumption of diet pellets was lower in the
   HFHSD than HFD study (2.3 ± 0.04 vs 3.0 ± 0.03 g/day).

Figure 6.

   [244]Figure 6
   [245]Open in a new tab

   Higher blood cholesterol and lower fasting blood glucose and insulin in
   P446L mice on high energy diets. A-F) Chronic study of P446L mice on
   HFD: A. Experimental design. B) Body weight, liver weight and
   epididymis fat pad weight, (n = 7,7). C) No difference in free-feeding
   blood glucose and insulin (21 wk, n = 7,7) or glucose tolerance (4-wk,
   GTT, n = 9,9). D) Higher blood cholesterol and liver triglyceride by LL
   genotype but no difference in blood triglyceride or liver cholesterol
   (n = 6–7). E) Representative hematoxylin-eosin images showing
   microvesicular steatosis in LL-genotype (red border). F) Histopathology
   scores for steatosis, microvesicular steatosis, lipogranulomas, lobular
   Inflammation, portal Inflammation and fibrosis showing higher
   microvesicular steatosis by LL genotype (n = 7,7). G-L) Chronic study
   of P446L mice on HFHSD: G. Experimental design. H. Body weight, liver
   weight and higher epididymal fat pad wt by LL genotype, n = 11–13. I)
   Free feeding blood glucose and insulin and glucose tolerance test
   (GTT), n = 11–14. J) Lower post-prandial blood glucose (2 h food
   withdrawal) and insulin and inorganic phosphate (Pi) at cull,
   n = 10–14. K) Higher blood cholesterol in LL genotype but no difference
   in blood triglycerides or liver triglycerides and cholesterol,
   n = 10–14. L) Histopathology scores showing higher lipogranuloma and
   lower fibrosis scores by LL genotype, n = 14,13. M) Representative
   hematoxylin-eosin images showing lipogranulomas and Sirius red fast
   green (SRFG) for fibrosis staining. Categorical data statistical
   differences was by Chi Square test; other analysis was by t-test
   #P < 0.05; ##P < 0.01; ###P < 0.001 for genotype effect.

   In the HFD study the LL mice had similar body weight gain
   ([246]Figure 6A,B) and no significant difference in blood glucose,
   insulin or glucose tolerance ([247]Figure 6C). There was no difference
   in blood triglyceride but blood cholesterol was higher (21%) and liver
   weight and liver triglyceride but not liver cholesterol were also
   higher in LL mice ([248]Figure 6B,D). Hepatocyte steatosis was evident
   in all mice, but microvesicular steatosis was more prominent in LL mice
   (P < 0.002). There was no evidence for hepatocellular ballooning and no
   difference between genotypes in lobular inflammation, lipogranulomas,
   Kleiner activity score, sinusoidal fibrosis, portal fibrosis or Kleiner
   fibrosis stage ([249]Figure 6E,F).

   In the HFHSD study ([250]Figure 6G) there was no difference in body
   weight gain or liver weight but adipose epididymis pad weight was
   higher in LL mice ([251]Figure 6H). There was no difference by genotype
   in free-feeding blood glucose and insulin, glucose tolerance
   ([252]Figure 6I) or insulin sensitivity (not shown). Blood glucose and
   insulin after 2 h food withdrawal were lower ([253]Figure 6J) and blood
   Pi (inorganic phosphate) was also lower in LL mice ([254]Figure 6J).
   Blood triglycerides were not different whereas total blood cholesterol
   ([255]Figure 6K), LDL-cholesterol (1.67 ± 0.09 vs 2.20 ± 0.14,
   P < 0.01) and HDL-cholesterol (3.21 ± 0.12 vs 3.74 ± 0.14, P < 0.01)
   were higher (21%, 32% and 17%, respectively) in LL mice, but there was
   no difference in liver triglyceride or cholesterol ([256]Figure 6K).
   Hepatocyte steatosis was evident in most mice, and microvesicular
   steatosis was not different by genotype ([257]Figure 6L). The LL mice
   had a higher score (P < 0.006) for lipogranulomas composed of a fat
   vacuole surrounded by mononuclear cells and macrophages
   ([258]Figure 6M) but had trends of lower sinusoidal fibrosis
   (P < 0.058) and Kleiner fibrosis stage (P < 0.051; [259]Figure 6L,M).

   Cumulatively, LL mice on HFD had higher blood cholesterol, liver
   triglycerides and microvesicular steatosis, whereas on HFHSD they had
   higher adipose pad mass but lower post-prandial glucose and insulin,
   raised blood cholesterol and raised liver lipogranulomas.

3.6. P446L mouse liver transcriptome: altered cholesterol homeostasis on
HFHSD

   Analysis of the P446L liver transcriptome in PP, PL and LL genotypes on
   HFHSD by unsupervised clustering identified 2 groups with either PP or
   LL genotypes and with PL distributed in the 2 groups. Subsequent
   analysis excluded the PL genotype and compared PP and LL genotypes on
   the HFHSD (P-HD, L-HD) and on the regular diet (P-RD, L-RD)
   ([260]Figure 7A). On the HFHSD there were 111 down-regulated and 55
   up-regulated genes by LL genotype (L-HD x P-HD) of which only one gene
   (Nup54, a nucleoporin) was in common with the DEGs on RD.

Figure 7.

   [261]Figure 7
   [262]Open in a new tab

   Liver transcriptome analysis in GKRP^P446>Lmice on the HFHSD (HD). A)
   Heat map for unsupervised clustering of P446L livers of PP and LL
   genotypes on regular diet (P-RD, P-HD, n = 4,4) or HFHSD (P-HD, L-HD,
   n = 9,10). B) Ingenuity pathway analysis showing selection of top
   significantly enriched canonical pathways ranked by P-adjust (FDR). C)
   Venn diagram for converse gene regulation by diet (P-RD x L-HD) and
   genotype (P-HD x L-HD). Red up-regulation, blue down-regulation. D)
   Heat map for selected DEGs from groups in C. E) RT-qPCR validation of
   DEGs (n = 4,4,10,12) and comparison with RNA-seq counts (n = 4,4,9,10)
   normalized to P-HD. F) Hmgcr Immunostaining in P446L livers (RD and HD)
   and cytoplasmic H-scores. n = 4,4 (RD); 9,9 (HD). G. Correlation of GK
   activity and RNA-seq Gck gene counts (n = 9,10)/. H) Venn diagram for
   common genes between the hepatocyte transcriptome (up-regulated by GK
   and down-regulated by GKRP, [263]Figure 2G) and genes up-regulated by
   diet (P-RD x P-HD; L-RD x L-HD) I) Correlation coefficient matrix for
   hepatocyte transcriptome ([264]Figure 2: n = 3 hepatocyte preparations,
   10 conditions) comparing expression of 17-cholesterol linked identified
   by the IPA analysis on P446L livers and 8 of 20 genes that were
   up-regulated by GK-overexpression and counter-regulated by GKRP in the
   hepatocyte study ([265]Figure 2) showing significant (P < 0.05)
   positive correlations (green) and negative correlations (purple). The
   Z-score shows up-regulation orange, down regulation grey. J)
   Correlation coefficient matrix for P446L mouse liver transcriptomes
   (P-RD, P-RD, P-HD, L-HD, n = 27) for the same genes as in (H). The
   Z-score shows relative expression by experimental group (P-RD, L-RD,
   P-HD and L-HDL-HD).

   Pathway enrichment analysis by EnrichGO or Ingenuity Pathway Analysis
   (IPA) identified cholesterol biosynthesis and LXR activation as the top
   regulated pathways by LL genotype (L-HD x P-HD). Key DEGs included
   repression of Hmgcr, Lss, Msmo1, Insig1, Insig2, Fgf1 and induction of
   Por. There were ∼2000 DEGs between the HD and RD transcriptomes for
   each genotype, and the top enriched pathways by IPA were cholesterol
   biosynthesis and lipid signalling (LXR, FXR, PPAR, PXR) which had
   negative and positive Z-scores, respectively ([266]Figure 7B). To
   determine whether the DEGs by LL genotype (L-HD x P-HD) represent a
   blunted response to diet, we compared the up-regulated and
   down-regulated genes by diet (P-HD x P-RD) with the conversely
   regulated DEGs by LL-genotype (L-HD x P-HD) ([267]Figure 7C). This
   showed that 47 of 111 down-regulated genes and 21 of 55 up-regulated
   genes by LL-genotype (HD) represent a blunted response to diet.
   Selected DEGs from the 4 groups are shown in the heat map
   ([268]Figure 7D). In addition to cholesterol biosynthesis, other
   cholesterol homeostasis genes include up-regulation of Abcg5 and Abcg8,
   involved in biliary sterol secretion [[269]46], Gpr146, linked to blood
   cholesterol homeostasis in man and mouse [[270]47] and Nr1h4, the bile
   acid receptor. Genes from the 4 groups were validated by RT-qPCR
   ([271]Figure 7E).

3.7. GK-upregulated genes correlate negatively with Hmgcr in P446L liver and
hepatocytes

   Comparison of DEGs by LL-genotype (HD) with genes linked to GK excess
   from the hepatocyte study ([272]Figure 2) identified 5 common genes:
   Slc25a47, Slc2a2, G6pc, Rgs16 (upregulated) and Hmgcr (down-regulated).
   Immunostaining for HMGCR protein showed lower protein in HD than RD
   groups but no difference by genotype ([273]Figure 7F). The latter may
   be linked to repression of Insig1 and Insig2 which regulate HMGCR
   protein stability and uncouple changes in Hmgcr mRNA and protein
   [[274]48]. Gck transcript was increased by HFHSD and repressed by
   LL-genotype and correlated with GK activity ([275]Figure 7F,G).

   To explore gene changes linked to GK excess we compared DEGs
   up-regulated by diet (P-RD x P-HD and L-RD x L-HD) with DEGs induced by
   GK expression in hepatocytes ([276]Figure 2H). There were 7 common
   genes up-regulated in both genotypes (Angptl8, Pklr, Tkfc, Rgs16,
   Slc17a4, Hcn3, Rnf144b), 2 unique to PP (Slc25a44, Aldob) and 3 to LL
   (Slc25a47, Slc2a2, Lhx6) ([277]Figure 7H).

   Correlation analysis of genes linked to GK-excess in hepatocytes (from
   [278]Figure 2) with genes linked to cholesterol biosynthesis [[279]49]
   or homeostasis from the IPA ([280]Figure 7B) is shown for both the
   hepatocyte transcriptome ([281]Figure 7I) and P446L mouse liver
   ([282]Figure 7J). In the hepatocyte study there were significant
   negative correlations between Abcg5, Abcg8, Gpr146 and cholesterol
   biosynthesis genes (n = 18) and negative correlations between Hmgcr and
   genes induced by GK excess (n = 8). Likewise, for P446L mouse liver
   ([283]Figure 7J) there were negative correlations between Abcg5, Abcg8
   and cholesterol biosynthesis genes (n = 12) or Insig1,2 (n = 3) and
   negative correlations between cholesterol biosynthesis genes and genes
   induced by GK excess (n = 51). Cumulatively, the inverse correlations
   between Hmgcr and either cholesterol excretion-linked genes or
   GK-inducible genes in both the hepatocyte model of varying GK-to-GKRP
   and P446L mouse liver support a mechanism linking GK-excess to
   cholesterol homeostasis.

3.8. Blood glucose and cholesterol in the P446L mouse correlate inversely
with GK-linked genes

   Correlation analysis of phenotype traits (blood glucose, Pi,
   cholesterol) of the P446L mouse on HFHSD with liver transcriptome
   counts for genes linked to GK excess or cholesterol homeostasis
   ([284]Figure 8) shows that blood glucose correlated negatively with
   genes linked to GK-excess (Slc2a2, Slc25a47, Angptl8, Rgs16, Rorc,
   G6pc) and cholesterol excretion (Abcg5, Abcg8) and positively with
   genes linked to cholesterol biosynthesis (Hmgcr, Msmo1, Cyp51a1,
   Insig1), supporting a link between GK-to-GKRP excess and blood glucose
   and cholesterol homeostasis. Blood inorganic phosphate (Pi) correlated
   positively with cholesterol biosynthesis genes (Hmgcr, Msmo1, Cyp51a1,
   Insig1) and negatively with G6pc, which generates Pi from glucose 6-P
   supporting a role for G6pc induction in hepatic Pi homeostasis
   [[285]50,[286]51]. Blood cholesterol (total, HDL, LDL) correlated
   positively with the Slc25a47 transporter. Common genetic variation at
   the human SLC25A47 locus associates with raised blood total and LDL
   cholesterol and lower fasting blood glucose in man (CMDKP:
   [287]hugeamp.org).

Figure 8.

   [288]Figure 8
   [289]Open in a new tab

   Correlation of liver transcripts with phenotypic traits in the P446L
   mouse on the HFHSD. A) Correlation of phenotypic traits: liver weight,
   triglyceride, cholesterol (LWt, LTG, LCE) and blood metabolites with
   corresponding gene counts from the liver transcriptome for the P446L
   mouse (PP,LL, n = 9/10) on the HSFSD showing: (i) Blood glucose (BG)
   and blood insulin (BI) correlated negatively with genes induced by GK
   excess (Scl2a2, Slc25a47, Rgs16, G6pc, Rorc) in the hepatocyte
   transcriptome ([290]Figure 2). (ii) Blood glucose and insulin
   correlated positively with Hmgcr and other genes linked to cholesterol
   biosynthesis (Msmo1, Insig2, Cyp51a1) and negatively with genes linked
   to cholesterol excretion (Abcg5, Abcg8). (iii) Inorganic phosphate (Pi)
   correlated negatively with G6pc; (iv) Blood cholesterol (TC, HDL, LDL)
   correlated positively with Slc25a47 and Slc25a10 and negatively with
   Tm7sf2).

4. Discussion

4.1. Mouse GKRP protein phenocopies the human GKRP P446 > L substitution

   The P446 > L substitution has qualitatively similar effects for mouse
   and human GKRP when GKRP is expressed at physiological levels relative
   to GK. Two key properties of the 446 L variant are compromised protein
   expressivity and a lower affinity for sequestering GK in the nucleus.
   The latter is not contingent on the lower expressivity. However, the
   lower expressivity is best explained by the lower affinity for GK based
   on the following considerations. First, overexpression of GK causes
   increased nuclear sequestration of GKRP. This manifests in both
   hepatocytes and heterologous cells (FTO2B and HeLa) and for both
   GKRP:446 P and 446 L and concurs with the small difference in affinity
   for GK between GKRP-446 P and 446 L. Second, GK overexpression
   stabilizes GKRP protein as determined by immunoblotting. Thirdly,
   cytoplasmic GKRP staining is barely detectable in mouse liver in either
   wild-type or the P446L mouse which has very low levels of GKRP protein
   in the LL-genotype. This can be explained by the very slow turnover
   rate of GKRP in mouse liver in vivo [[291]52] and a requirement for
   nuclear sequestration of GKRP for stabilization.

   Small differences between mouse and human GKRP in the magnitude of the
   effect of the P446 > L substitution on glucose-induced translocation of
   GK concur with the higher affinity of human GKRP compared with rodent
   GKRP for GK [[292]25] and with the higher nuclear sequestration of both
   GKRP and GK in hepatocytes transfected with human compared with mouse
   GKRP. It may also explain the similar sequestration of GK in the
   nucleus in human (446 PP genotype) compared with mouse liver despite
   the lower GK and GKRP protein levels. The question whether rodent GKRP
   adequately models the human-GKRP:P446 > L substitution had remained
   unsettled based on two considerations. First, significant differences
   between GKRP:446 P and 446 L in GK kinetic studies were found for human
   [[293]18] but not rat GKRP [[294]24]. Second, a study expressing
   chimaeric fluorescent-GKRP constructs (446 P/L) in wild-type mouse
   hepatocytes found similar nuclear GK sequestration for rat 446 L and
   446 P despite lower sequestration for human 446 L [[295]19]. Two points
   are noteworthy: first, differences in nuclear GK sequestration between
   446 L and 446 P are small and critically dependent on the GK-to-GKRP
   ratio; second, because of the higher affinity for GK of human GKRP,
   differences between 446 L and 446 P manifest at different GKRP-to-GK
   ratios necessitating transfections over a range of GK-to-GKRP ratios.
   We infer that mouse GKRP:P446 > L substitution phenocopies the human
   GKRP:P446 > L despite the higher affinity of human GKRP for GK. For
   both species there is lower expressivity of GKRP:446 L and lower
   nuclear GK sequestration.

4.2. Lower liver GK and GKRP protein in human rs1260326-446LL genotype and in
the P446L mouse

   Human NAFLD tissue homozygous for the 446 L allele (rs1260326,TT) had
   lower GKRP and GK protein levels and lower ratios of nuclear GK/nuclear
   GKRP, which concurs with the lower expressivity of the GKRP-446 L
   variant and the lower nuclear GK sequestration. It also concurs with
   the P446L mouse which has 70–90% reduction in GKRP protein and 40–50%
   reduction in GK protein. The very low GKRP protein levels in young mice
   on regular diet was surprising given the modestly lower expressivity of
   the 446 L variant when transfected in hepatocytes at high mRNA
   transcript levels to attain physiological GKRP protein levels within
   ∼24 h culture. It is best explained by the very slow turnover rate of
   GKRP protein in mouse liver in vivo (t[1/2] 6.5 days) compared with
   either GK (t[1/2] 4.6 days) or the average value (t[1/2] < 1 d) for
   mouse liver protein [[296]52]. The P446L mouse had a reduction in total
   GK protein and activity but a relative excess of GK-to-GKRP protein
   ∼2-4-fold compared with wild-type mice and shares some similarities
   with Gckr-knock-out mice [[297]32,[298]33] which have ∼50% reduction in
   GK but maintain glucose homeostasis on a standard diet. It is widely
   assumed that the low GK protein in Gckr-deficient states is consequent
   to a stabilizing role of GKRP. However direct evidence for such a
   mechanism is lacking. In the present study in hepatocytes
   overexpressing GK and GKRP there was evidence for GK stabilizing GKRP
   but weaker non-significant trends for GKRP stabilizing GK. The lower GK
   protein levels in the P446L mouse may be in part due to transcriptional
   regulation analogous to induction of G6pc [[299]51,[300]52]. The
   metabolic studies on hepatocytes from the P446L mouse showed that
   glucose disposal rates were identical at basal glucose (5 mM) but lower
   for the 446LL genotype at elevated glucose. Likewise, the induction of
   ChREBP-β and its target genes during high-glucose challenge was also
   blunted. The lower metabolic capacity of 446LL hepatocytes at high
   glucose challenge indicates that despite the relative GK-to-GKRP excess
   there is effective adaptation to maintain basal glucose homeostasis on
   the regular diet. The mice on the HFHSD (both wild-type and 446LL) had
   a ∼2-fold increase in GK and an increase in GKRP. An increase in GK
   activity on a high-fat diet has been reported and concurs with the
   raised insulin [[301]53]. Given the established role of insulin in
   promoting hepatic GK expression [[302]11] the concurrent increase in
   GKRP is at least in part consequent to the stabilizing effect of GK.

4.3. Lower blood glucose and insulin in the P446L mouse on the HSHSD

   The GCKR locus is one of ∼240 loci that associate with blood glucose or
   type 2 diabetes risk [[303]1,[304]54]. The majority of the associated
   variants are intronic and of the ∼20 associated missense variants, few
   have been validated in mouse models [[305]55,[306]56]. The human
   GCKR-446 L allele is associated with higher blood triglycerides and
   lower blood glucose, with a smaller effect size on blood glucose (<4%)
   than on triglycerides (8–25%) [[307][3], [308][4], [309][5]] and with
   the widely held assumption that it represents increased hepatic
   conversion of glucose to triglyceride [[310]5]. The P446L mouse on the
   HFHSD had lower post-prandial blood glucose and insulin without an
   increase in blood triglyceride. The negative correlation between blood
   glucose and transcript levels of genes that are induced by GK
   overexpression supports a role for functional GK excess in the lower
   blood glucose. Two mechanisms can be considered consequent to the
   higher GK-to-GKRP protein ratio increased hepatic glucose uptake and
   lower hepatic gluconeogenesis. The latter is supported by the higher
   rates of endogenous glucose production in people with GCK inactivating
   mutations [[311]57]. Lower endogenous glucose production in
   hyperglycaemic clamps was also found in association with GCKR
   rs780094^C > T [[312]5]. The negative correlation with blood glucose
   and inorganic phosphate (Pi) of the G6pc transcript which encodes the
   final enzyme in hepatic glucose production, is best explained by the
   role of this enzyme in hepatic phosphate homeostasis [[313]51,[314]52].
   The lower blood insulin may be consequent to the lower blood glucose or
   the lower hepatic GK protein because lower insulin was found in
   liver-selective GK-deficient models [[315]58]. Cumulatively, the P446L
   mouse on the HFHSD replicates the lower blood glucose and insulin but
   not the raised blood triglyceride found in genome-wide associations
   with the GCKR locus.

4.4. Raised blood cholesterol and changes in cholesterol-linked genes in the
P446L mouse

   The P446L mouse on the high-energy diets had raised blood cholesterol
   without raised blood triglycerides and with a liver transcriptome
   characterized by altered cholesterol homeostasis. This included
   repression of cholesterol biosynthesis enzymes (Hmgcr, Msmo1, Lss) and
   the cholesterol regulators Insig1 and Insig2 which function as sterol
   sensors and have a dual role in anchoring the transcription factor
   SREBP-SCAP to the ER membrane and promoting ubiquitination and
   degradation of the Hmgcr protein, the limiting enzyme of cholesterol
   biosynthesis. It also included induction of Abcg5 and Abcg8 encoding
   cotransporters of biliary sterol excretion [[316]46] and Gpr146 which
   is linked to blood cholesterol in man and mouse [[317]47]. Hmgcr gene
   repression was identified by RNA-sequencing in hepatocytes
   overexpressing GK, indicating a candidate link from the GK-to-GKRP
   ratio to altered cholesterol regulation. Moreover, in both the
   hepatocytes expressing varying GK-to-GKRP ratios and in the P446L
   mouse, Hmgcr transcript levels correlated positively with Msmo1 and Lss
   and negatively with several genes induced by GK excess including
   putative ChREBP target genes (G6pc, Rgs16, Angptl8, Slc2a2) and various
   mitochondrial transporters including Slc25a47 which was reported to
   correlate negatively with cholesterol biosynthesis [[318]42]. The lack
   of difference in Hmgcr protein in the P446L mouse despite lower
   transcript levels may be linked to repression of Insig1 and Insig2
   because down-regulation of these genes leads to increased Hmgcr protein
   compared with mRNA levels [[319]48]. Common human variants for HMGCR,
   INSIG1, INSIG2, ABCG5, ABCG8 and GPR146 are associated with blood
   cholesterol levels ([320]https://hugeamp.org), although with modest
   effects for HMGCR variants [[321]59]. This concurs with the complexity
   of HMGCR regulation whereby sterols inhibit transcription whereas
   nonsterol isoprenoids inhibit translation and both metabolites promote
   proteasomal degradation [[322]60].

   The mechanisms linking molar GK-to-GKRP excess with altered cholesterol
   homeostasis remain to be elucidated. An increase in both blood
   cholesterol and triglyceride occurs in G6pc deficiency in man and mouse
   [[323][61], [324][62], [325][63]]. This is a more severe phenotype of
   hepatic glucose 6-P accumulation and ChREBP activation because of the
   major role of glucose 6-phosphatase in hepatic phosphate homeostasis
   [326][50], [327][51], [328][52] as shown by the greater accumulation of
   glucose 6-P by inhibition of glucose 6-P hydrolysis relative to GK
   activation with pharmacological activators ([329]Figure 5). In the
   hepatocyte model of GK-to-GKRP excess challenged with high glucose, the
   repression of Hmgcr correlated negatively with multiple genes induced
   by metabolite accumulation including the hepatokine Angptl8 and the
   transcriptional regulator Rorc, linked to cholesterol metabolism
   [[330]64]. Because significant repression of Hmgcr in the P446L mouse
   occurred after the HFHSD and together with induction of genes
   characterized by GK excess (G6pc, Rgs16, slc2a2, Slc25a47), altered
   cholesterol homeostasis occurs in conjunction with ChREBP activation.
   Elevated expression of cholesterol biosynthesis genes by ChREBP
   knock-down in a metabolically challenged state has been reported
   [[331]65].

4.5. Blood triglycerides in the P446L mouse compared with human GCKR
associations

   The lack of raised blood triglycerides despite raised blood cholesterol
   and lower blood glucose in the P446L mouse was confounding, given that
   the human GCKR variant associates with a greater effect on blood
   triglycerides than on blood cholesterol [[332]7,[333]8] or blood
   glucose [[334][3], [335][4], [336][5]]. Mouse models of Gckr-deficiency
   which are essentially models of GK-to-GKRP excess likewise did not have
   raised blood triglycerides [[337]33]. Amongst Gck-transgenic models,
   which represent GK-excess at endogenous GKRP, raised blood and liver
   triglycerides was found in 12-month old mice which had raised blood
   glucose and insulin [[338]66] but not in young insulin-deficient mice
   with a greater GK-excess [[339]67]. Additionally, there was no increase
   in liver triglyceride in a Gck-transgenic model on a HFD, which had
   lower blood insulin in conjunction with 2-fold GK excess [[340]53].
   Cumulatively, in mouse genetic models of relative GK-to-GKRP excess,
   the raised blood and liver triglycerides in conjunction with raised
   blood insulin and glucose [[341]66] implicates insulin resistance. In
   human GCKR association studies it was noted at the outset that the GCKR
   effect on blood triglycerides was greater in populations with higher
   body mass index [[342]4]. This was later corroborated by a larger
   effect of the GCKR variant on blood [[343]68] and liver [[344]69]
   triglycerides in obesity and type 2 diabetes [[345]70,[346]71]. An
   alternative hypothesis to gene–phenotype interactions is
   “quantile-dependent expressivity” where the genetic difference on
   triglycerides depends on the magnitude of the triglyceride
   concentration in the population [[347]72]. In the context of relative
   associations of the GCKR variant with blood triglycerides and
   cholesterol, studies on children corroborate the above hypotheses
   [[348][73], [349][74], [350][75]] and in healthy paediatric cohorts
   with substantially lower triglyceride concentration than in the adult
   population the GCKR variant associates with raised blood cholesterol
   but not triglyceride [[351]73]. Raised blood triglycerides occurs
   through raised hepatic secretion and/or impaired extrahepatic
   clearance, and clinically high blood triglycerides generally reflect
   compromised triglyceride clearance as occurs in obesity [[352]76]. The
   lack of increased blood triglycerides in the P446L mouse could be
   linked to efficient extrahepatic triglyceride clearance, because of
   deficiency in cholesteryl ester transfer protein [[353]77] or other
   mechanisms. The raised liver triglyceride and microvesicular steatosis
   on the HFD and the increased fat pad mass on the HFHSD could be
   consistent with this interpretation. Raised blood triglycerides occurs
   in human G6PC deficiency and in mouse models of G6pc deficiency
   [[354][61], [355][62], [356][63]]. This represents an analogous but
   more severe hepatic dysregulation of glucose 6-P than relative
   GK-to-GKRP excess. The hypertriglyceridemia in mouse models of G6pc
   deficiency is associated with impaired extrahepatic clearance
   [[357]62].

   The liver produces the hepatokine ANGPTL8 which functions as an
   inhibitor of triglyceride clearance by oxidative tissues thereby
   promoting raised blood triglycerides and/or diversion of triglyceride
   to adipose sites [[358]78,[359]79]. In man circulating ANGPTL8 levels
   correlate with obesity and interestingly also with the human GCKR
   rs1260326 variant [[360]80]. The current finding that hepatocyte
   expression of Angptl8 correlates with GK excess over GKRP provides an
   explanation for a causal mechanism between the GCKR variant and the
   raised blood ANGPTL8 levels in man [[361]80]. ANGPTL8 thereby
   represents one candidate, consequent to GK excess that can amplify
   blood triglyceride levels in association with the GCKR variant.
   However, the possibility that other gene variants at the GCKR locus on
   neighbouring genes that are in linkage disequilibrium can have
   additional synergistic effect on blood triglycerides needs to be
   considered.

Author contributions

   L.A. directed the project, conceived and designed the experiments,
   contributed to the analysis and co-wrote the paper. B.E.F. and S·S.C.
   conducted the experiments, analyzed the data and co-wrote the paper.,
   D.T. conducted the histopathology analysis and interpretation; K.R.,
   T.M., Z.A.O., H.L.R., Q.M.A., J.M.S., R.J.F., D.M.S., contributed to
   data collection or analysis. All authors approved the final version
   prior to submission for publication.

Acknowledgements