Unexpected partial correction of metabolic and behavioral phenotypes of Alzheimer’s APP/PSEN1 mice by gene targeting of diabetes/Alzheimer’s-related Sorcs1

Introduction Insulin resistance and type 2 diabetes mellitus (T2D) are associated with increased risk for cognitive impairment, Alzheimer’s disease (AD) and vascular dementia. SORCS1 encodes a protein-sorting molecule genetically linked to both T2D and AD. The association of SORCS1 with both AD and T2D is sexually dimorphic in humans, with both disease associations showing more robust effects in females. Based on published evidence that manipulation of the mouse genome combining multiple genes related to cerebral amyloidosis, to T2D, or both, might provide novel mouse models with exacerbated amyloid and/or diabetes phenotypes, we assessed memory, glucose homeostasis, and brain biochemistry and pathology in male and female wild-type, Sorcs1 -/-, APP/PSEN1, and Sorcs1 -/- X APP/PSEN1 mice. Results Male mice with either the APP/PSEN1 or Sorcs1 -/- genotype displayed earlier onset and persistent impairment in both learning behavior and glucose homeostasis. Unlike prior examples in the literature, the behavioral and metabolic abnormalities in male mice were not significantly exacerbated when the two disease model mice (Sorcs1 -/- models T2D; APP/PSEN1 models AD) were crossed. However, female Sorcs1 -/- X APP/PSEN1 mice exhibited worse metabolic dysfunction than Sorcs1 -/- knockout mice and worse memory than wild-type mice. The deletion of Sorcs1 from APP/PSEN1 mutant mice led to no obvious changes in brain levels of total or oligomeric amyloid-beta (Aβ) peptide. Conclusions In general, unexpectedly, there was a trend for gene targeting of Sorcs1-/- to partially mitigate, not exacerbate, the metabolic and amyloid pathologies. These results indicate that crossing AD model mice and T2D model mice may not always cause exacerbation of both the amyloidosis phenotype and the metabolic phenotype and highlight the unexpected pitfalls of creating mixed models of disease. Electronic supplementary material The online version of this article (doi:10.1186/s40478-016-0282-y) contains supplementary material, which is available to authorized users.


Introduction
Alzheimer's disease (AD) is a progressive neurodegenerative disorder and is the most common cause of dementia in the elderly. There are approximately 34 million people afflicted by AD worldwide and those numbers are expected to triple over the next 40 years [5]. Type 2 diabetes mellitus (T2D) increases the risk for dementia [15,19,42,69]. Both T2D and dementia are highly complex and share several features, including impaired cognitive function, vascular dysfunction and oxidative and inflammatory stress [15,54]. Several mechanisms have been proposed for this association, including hypercholesterolemia, vasculopathic factors, and insulin resistance [63]. Environmental factors, such as poor diet and obesity are also known to increase the predisposition to T2D and the risk for dementia [21]. The extent to which dementia is due to AD or due to vascular dementia is an area of great interest, and the pathology of dementia in T2D is usually mixed [1,43,65].
Rare, early-onset familial Alzheimer's disease (EOFAD) is believed to begin with the accumulation of oligomeric forms of Aβ in the hippocampus and cortex. EOFAD is caused by mutations in genes that directly influence Aβ metabolism, most commonly the amyloid precursor protein (APP), presenilin1 (PSEN1) or presenillin 2 (PSEN2). Mice expressing pathogenic mutant forms of human APP, PSEN, or both, are mainstays of in vivo modeling of cerebral amyloidosis in AD research. Of note, mice harboring EOFAD-pathogenic mutations in both APP and PSEN1 have been observed to develop insulin resistance (Ruiz et al. in press [56]).
Genetic studies of late-onset Alzheimer's disease (LOAD) point to a number of susceptibility genes, including several that belong to one of three classes of molecules: the apolipoprotein family (the most notable of which is apolipoprotein E [11,59]; the low density lipoprotein receptor (LDLR) family [22,31]; and the vacuolar protein sorting-10 (VPS10) domain containing receptor family. The ADlinked gene SORL1 belongs to both the LDLR family and the VPS10-domain protein family, and a deficiency in SorL1 protein has been observed in the brains of patients suffering from LOAD [12,33,55,58,60]. SorL1 has been demonstrated to modulate Aβ generation via an interaction with the core component of the retromer complex, Vps35 [2,39,41,61]. Human and animal studies in AD have implicated Vps35 and other components of the retromer complex [36,64].
Multiple studies have demonstrated a connection between insulin resistance and T2D and increased risk for cognitive impairment in humans [4,13,34,62]. SORCS1, another member of the Vps10 family, has been identified as a potential risk factor for LOAD and is decreased in AD brains [33]. SORCS1 resides at a quantitative trait locus for T2D in mice and rats [10,14] and has been associated with both T1D and T2D [13]. Previously, we initiated an investigation of the action of SorCS1 on APP metabolism. We observed that endogenous murine Aβ40 and 42 levels were increased in the brains of female but not male Sorcs1 -/-mice, and this appeared to involve an interaction mediated by the Vps35 retromer complex [29]. This was especially interesting because the genetic linkage to SORCS1 is sexually dimorphic with the linkage to women being more robust. We hypothesized that SORCS1 hypomorphism might contribute to the link between insulin resistance and increased risk for cognitive impairment in humans. A corollary of that hypothesis is that dysregulation of SorCS1 may contribute to both the Aβ disturbance underlying AD and the insulin/ glucose metabolism disturbance underlying DM, with the two genes arranged either in parallel or in series. Kebede et al. [23] demonstrated that Sorcs1-/mice were not overtly diabetic unless they were crossed with leptin deficient (ob/ob) mice. In contemplating this result in the context of the AD/T2D association with SORCS1, this observation raised the possibility that Sorcs1-/mice crossed with AD model mice (here, APP/PSEN1 mice) might manifest overt diabetes if stressed by the accumulation of cerebral amyloid. Previous studies show that the combination of a genetic proamyloidogenic phenotype and potentially diabetogenic phenotype (induced by either drug, diet, or genetic manipulations) can lead to exacerbation of either the proamyloidogenic component of the phenotype, the prodiabetogenic component of the phenotype, or exacerbation of both components of the phenotype [8,18,20,38,46,51,52,66].
Here we report characterization of the memory, glucose homeostasis and Aβ pathology in male and female wildtype, Sorcs1 -/-, APP/PSEN1 and Sorcs1 -/-mice crossed with APP/PSEN1 mice. Unexpectedly, there was a trend for gene targeting of Sorcs1-/to mitigate, not exacerbate, the metabolic, behavioral, and cerebral amyloid angiopathic pathologies.

Animals
Male and Female C57BL/6J wild-type (WT; n = 5-14/ group), Sorcs1 -/-(n = 10-16/group), APP Swe x PSEN1-Δexon9 (APP/PSEN1; n = 5-11/group respectively) and Sorcs1 -/-x APP Swe x PSEN1 Δexon9 (Sorcs1 -/-x APP/ PSEN1; n = 5-9/group) were bred in the Icahn School of Medicine at Mount Sinai Vivarium. We have previously described the generation of the Sorcs1 -/-line [29]. All animal studies were conducted in accordance with National Institute of Health Guidelines for the Care and Use of Experimental Animals and approved by the Institutional Animal Care and Use Committee at the Icahn School of Medicine at Mount Sinai. Mice were kept in a pathogen-free environment on a 12-h light/dark cycle, and given ad libitum access to food and water. From weaning until 6 months of age, cohorts were maintained on standard rodent chow. Cohorts were subjected to learning behavior assessment at 5 and 9 mo of age. Cohorts were placed on a 60 % high fat diet (Diet 58Y1, TestDiet, St. Louis, MO) at 6 mo of age for 4-12 weeks. Mice were subjected to metabolic profiling prior to starting the HFD at 4-5 mo of age then again after 4-12 weeks of HFD. All mice were weighed weekly. Following final metabolic testing mice were anesthetized, and were then perfused with ice-cold PBS. The brain was removed and dissected into two hemispheres. One half was snapfrozen on dry ice and stored at -80°C for Aβ biochemistry. The other half was post-fixed in 4 % PFA then cut into 30 μm coronal sections using a vibratome. Sections were stored in PBS with 0.1 % NaAz until histological analyses.

Metabolic profiling Body composition analyses
Total body fat and lean tissue were assessed using an EchoMRI quantitative magnetic resonance system (Echo Medical Systems) as previously described [57].

Glucose tolerance test (GTT)
Mice were fasted for 5 h in a quiet testing room with free access to water. Baseline fasted blood samples were then taken via tail vein bleed to measure plasma insulin, lipids, and branched-chain amino acids. Blood samples were collected into EDTA coated microvette tubes and placed on ice before being spun at 7 K for 6 min. Plasma was then removed, snap frozen on dry ice and stored at -80°C until analyses. Mice were then injected intraperitoneally (i.p) with a 0.75 or 1.5 g/kg glucose solution (Sigma Aldrich, St. Louis, MO). Blood glucose was measured via tail vein bleed immediately before the glucose bolus then at 15, 30, 60, 90 and 120 min post injection.

Insulin tolerance test (ITT)
Mice were fasted for 5 h then given an i.p. injection of 1 or 2.5U/kg insulin (Humulin R, Lilly & Co., Indianapolis, IN) solution. Blood glucose levels were measured prior to and at 15, 30, 45, 60 and 90 min following the insulin bolus.

Behavioral testing
Mice were placed in the testing room 1 h prior to testing to acclimatize to the room. All testing was completed between 8 am-4 pm. All equipment was cleaned between animals.

Novel object recognition (NOR)
Short-term non-associative memory based on the natural exploration of novelty in mice was assessed in the novel object recognition test as described [26]. Briefly, on day 1, the mouse was habituated in the NOR arena (20 cm diameter) for 10 min. On day 2, the mouse underwent the testing phase, composed of two stages. During testing phase 1, the mouse was placed in the arena and allowed to explore two identical unfamiliar objects for 10 min. The mouse was then returned to its home cage for an interval of 1 h. During this time, one of the two objects the mouse was previously allowed to explore was removed and replaced with a novel object. During testing phase 2, the mouse was placed back into the arena and allowed to explore the familiar object and the novel object for 4 min. Trials were videotaped using an overhead camera. The duration spent exploring the objects was then measured using ANY-maze (Stoelting, Wood Dale, IL). Exploration was defined as the amount of time the mouse spent pointing their nose within 2 cm of the object.

Y-maze spontaneous alternation (SA)
Short-term working was assessed in the Y-maze spontaneous alternation (SA) test as previously described [25] using a black opaque Perspex Y-maze with 3 arms (A, B, and C) each containing a visual cue (arm dimensions; 35x5x10 cm). Briefly, each animal was placed in turn in arm A of the Y-maze and allowed to explore for 8 min and the arm entries made by each animal were recorded. Arm entry was defined as having all 4 paws in the arm. Spontaneous alternation was defined as a successive entry into 3 different arms, on overlapping triplet sets. The percentage number of alternations was calculated as the number of actual alternations divided by the maximum number of alternations (the total number of arm entries minus 2).

Biochemical and histological analyses
Plasma insulin, lipids, and branched-chain amino acid level determination Baseline, 1 and 3 months post HFD, levels of plasma insulin were determined using a mouse insulin ELISA kit (Mercodia, Sweden) according to the manufacturer's instruction. Glycerol and triglycerides were quantified using a plasma triglyceride assay (Sigma Aldrich, St. Louis, MO), which was modified to a 96-well plate format. Non-esterified fatty acid (NEFA) levels were determined using a NEFA kit from Wako (Wako Chemicals USA, Inc. Richmond, VA) following the manufacturer's instructions. Plasma branched-chain amino acid (BCAA) levels were determined using a fluorimetric assay as described [6].

Aβ assay
Hemibrains were processed via differential detergent solubilization [40] to produce TBS-soluble, Triton-Xsoluble and formic-acid soluble Aβ fractions. Total Aβ equals the sum of Aβ concentration in TBS-soluble, Triton X-soluble and formic acid-soluble fractions. For analysis of native oligomeric Aβ protein structure, 2-4 μl native protein samples (from the TBS-soluble fraction) were spotted onto activated/pre-wetted PVDF membrane (0.22 μm; Millipore) and allowed to dry. Following protein spotting, membranes were blocked for 1 h at room temperature in 5 % w/v non-fat milk (Santa Cruz) in TBS containing 0.1 % v/v Tween-20 (Fisher Scientific; TBS-T). Membranes were incubated in the indicated primary antibody (in 5 % milk/TBS-T) overnight at 4°C, washed 4x in TBS-T, incubated in species-specific HRP-conjugated secondary antibody (in 5 % milk/TBS-T) for 1 h at room temperature, and then washed 4x in TBS-T. Membranes were subsequently developed with ECL Western blotting substrate (Pierce) using the Fujifilm LAS-3000 developer. Membranes were washed 1x in TBS-T and stripped in low pH stripping buffer [25 mM Glycine HCl, pH 2.0 and 1 % w/v SDS] to remove primary and secondary antibody, washed 3x in TBS-T, and blocked for 1 h (in 5 % milk/TBS-T) at room temperature before probing with the next primary antibody. Integrated density of immunoreactive spots was measured using MultiGauge Software (FujiFilm) and normalized to % control (vehicle). Generation, purification, and characterization of rabbit pAb A11 (anti-prefibrillar oligomers, 0.5 μg/ml), rabbit pAb OC (anti-fibrillar oligomers and fibrils; 0.25 μg/ml) and mouse mAb Nu-4 (anti-oligomers; 1 μg/ml) have been described previously [28,67]. Normalization to total APP/Aβ signal was achieved by detection of human APP transgene metabolites with the mouse mAb 6E10 (1:1000; Covance). Peroxidase-conjugated goat anti-rabbit IgG (H + L; 1:20,000; Vector Labs) or goat anti-mouse IgG (H + L; 1:20,000; Vector Labs) were used for detection. To quantify total Aβ levels, human/rat Aβ 1-40/1-42 ELISA kits (Wako) were used according to the manufacturer's instructions. Absolute concentrations of total or oligomeric Aβ were normalized to initial tissue weight prior to analysis.

Histology
Aβ was assessed via free-floating fluorescence immunohistochemistry using mAb 6E10 (1:1000, Covance, Princeton, NJ). To assess Aβ deposition within vessels sections were co-stained with collagen IV (1:300). Briefly, sections were washed 3x in PBS for 10 min, incubated in 3 % acetic acid for 10 min followed by 1 mg/ml Pepsin for 30 min. Following 1x in PBS/0.1 % NaAz for 10 min and 3x 10 min washes in PBS, sections were incubated 1 % H 2 O 2 for 20 min, 3x further 10 min washes in PBS before blocking with 5 % goat serum in TBS/Triton-X for 1 h. Sections were then incubated with mAb 6E10 (1:1000, Covance, Princeton, NJ) and collagen-IV (1:300, Abcam ab6586) overnight at 4°C. The following day, sections were given 3x 20 min washes in PBS then incubated with anti-mouse and anti-rabbit fluorescent secondary antibodies (1:200, mouse Alexa Fluor 488 and rabbit 594 respectively) at room temperature for 2 h. After 3x final 20 min washes in PBS, sections were mounted and coverslipped with Vectashield (Vector Laboratories, Inc., Burlingame, CA). Images were captured on an Olympus BX61 upright microscope with an attached Olympus DP71 camera.
Perls's Berlin blue-stained clusters of hemosiderin staining were evaluated from sections throughout the neocortex, hippocampus, and thalamus. Briefly, sections were mounted on slides and dried overnight. The next days, slides were immersed in equal parts of 20 % hydrochloric acid and 10 % potassium ferrocyanide for 20 mins. Slides were then washed 3x in dH 2 0 then counterstained with nuclear fast red for 5 min. Following another 2x rinses in dH 2 0, sections were dehydrated through 95 % and 2x changes of 100 % ethanol. Sections were then cleared in xylene, coverslipped and allowed to dry overnight before imaging.

Statistical analyses
All data are presented as the mean ± s.e.m. Statistical significance (P < 0.05) was determined using Student's t tests or one/two-way ANOVA with Bonferroni posthoc analyses (GraphPad Prism, San Diego, CA).

Results
Glucose homeostasis and learning behavior were tested in mice at 4-5 mo of age. Young male and female WT, Sorcs1 -/-, APP/PSEN1 and Sorcs1 -/-x APP/PSEN1 mice had similar levels of fasting glucose and insulin (Additional file 1: Figure S1a-d) and normal ability to clear a glucose load when assessed by a GTT (Additional file 1: Figure S1a,c). Fasting plasma glycerol (Additional file 1: Figure S1i,k) and triglycerides (TG) (Additional file 1: Figure S1j,l) levels were unaltered indicating normal lipid homeostasis in mice of both sexes. Female Sorcs1 -/-mice displayed lower body weight, shorter body length, and reduced lean mass (Additional file 1: Figure S1e,f,m) but comparable fat mass (Additional file 1: Figure S1n) when compared to WT mice. Female Sorcs1 -/-mice were also reduced in weight and lean mass (Additional file 1: Figure S1e,m) as compared with APP/PSEN1 mice. No differences in body weight or fat mass were seen among male cohorts (Additional file 1: Figure S1g,p). However, male Sorcs1 -/-mice were shorter in length than APP/PSEN1 mice and Sorcs1 -/-x APP/PSEN1 mice had reduced lean mass as compared to WT mice (Additional file 1: Figure S1h,o). Crossing Sorcs1 -/-with APP/PSEN1 mice did not promote the onset of any metabolic phenotype in either female nor male mice.
When behavior was assessed in Y-maze, both male and female young Sorcs1 -/-, APP/PSEN1 and Sorcs1 -/x APP/PSEN1 mice all had worse SA memory performance than WT mice, with only Sorcs1 -/x APP/PSEN1 reaching significance in the females and Sorcs1 -/-in the males ( Fig. 1a-b). Memory was also assessed in the novel object recognition (NOR) test, both male and female WT mice had intact memory whereas male and female Sorcs1 -/-, APP/PSEN1, and Sorcs1 -/-x APP/PSEN1 mice were all impaired and to a comparable degree ( Fig. 1c-d).
In order to determine whether cohorts in this study were susceptible to the induction of metabolic dysregulation by high fat feeding, a commonly used metabolic stressor to unveil a metabolic phenotype, all mice were placed on a 60 % high fat diet (HFD) for 4 weeks. All female mice were resistant to diet-induced perturbations in glucose homeostasis (Fig. 2a) with no genotypedependent differences in fasting plasma metabolic parameters detected at 7 mo of age, consistent with the notion that female sex is protective against a metabolic phenotype (Fig. 2d-e, h-I,l). After 4 weeks of HFD feeding, differences in body weight observed on regular chow were no longer statistically significant (Fig. 2m). As noted above for regular chow feeding, female Sorcs1 -/-mice on HFD continued to have less lean mass than WT or APP/PSEN1 mice (Fig. 2p). While all mice gained a significant amount of fat mass on HFD, body composition analysis revealed no differences among cohorts (Fig. 2q).
After 4 weeks on HFD and in comparison to WT mice at baseline (4-5 mo of age), male APP/PSEN1, Sorcs1 -/-, and Sorcs1 -/-x APP/PSEN1 mice displayed impaired glucose tolerance (Fig. 2b) and APP/PSEN1 and Sorcs1 -/-x APP/PSEN1 mice impaired insulin sensitivity (Fig. 2c). Unexpectedly, APP/PSEN1 male had worse glucose tolerance and insulin sensitivity than Sorcs1 -/-x APP/ PSEN1 mice (Fig. 2b-c). Insulin sensitivity was also worse in APP/PSEN1 mice than in Sorcs1 -/-mice (Fig. 2c). This was also reflected in higher levels of fasting plasma insulin in male APP/PSEN1 mice compared to all other groups (Fig. 2f ). All other metabolic parameters were unaltered (Fig. 2g, j-k, n) as was body weight (Fig. 2o) and lean mass (Fig. 2r). As with females, all males gained a significant amount of fat mass on the HFD as seen by body composition analyses (Fig. 2s). These data are consistent with enhanced susceptibility to HFD induced metabolic dysregulation of APP/ PSEN1 mice (Ruiz et al.,in press [56]).  1 Female but not male Sorcs1 -/-x APP/PSEN1 mice display impaired memory. Spontaneous alternation (SA) was assessed at 5 mo (a, b) and retested again at 9 mo (e, f), while novel object recognition (NOR) was assessed at 5 mo (c, d) in cohorts of male and female WT (n = 11-14/group), Sorcs1 -/-(n = 15-16/group), APP/PSEN1 (n = 6-11/group) and Sorcs1 -/-x APP/PSEN1 (n = 9/group). One-way ANOVA with Bonferroni posthoc analyses; *P < 0.05, **P < 0.01 and ***P < 0.001. Data expressed as mean ± s.e.m Aging is a risk factor for both AD and T2D. To determine whether cohorts were more susceptible to aginginduced insulin resistance and/or glucose intolerance, mice were maintained on the HFD for an additional 8 weeks i.e., to a total of 12 weeks on HFD. During this extension of HDF feeding, abnormalities in glucose homeostasis emerged in females cohorts. Both female APP/PSEN1 and Sorcs1 -/-x APP/PSEN1 mice displayed impaired glucose tolerance (Fig. 3a), and APP/PSEN1 mice displayed impaired insulin sensitivity (Fig. 3b). The metabolic phenotype was more severe in female APP/ PSEN1 and Sorcs1 -/-x APP/PSEN1 mice than in female Sorcs1 -/-mice ( Fig. 3a-b). After 12 weeks of HFD, fasting plasma insulin levels were elevated in all genotype groups but this elevation only reached statistical significance in female APP/PSEN1 mice (Fig. 3e). All other metabolic parameters were unaltered (Fig. 3f, i, j, m) across other cohorts. Sorcs1 -/mice had lower body weight than WT and APP/PSEN1 mice (Fig. 3n).
When learning behavior was re-tested after 12 weeks on HFD, all groups of female mice (regardless of WT vs genetically manipulated) displayed similar performance (Fig. 1e). As observed at baseline and as mentioned above, male Sorcs1 -/-mice had poorer performance than WT mice in SA memory (Fig. 1f ). No other groups of male mice displayed learning behavior deficits that reached statistical significance. After 12 weeks of HFD feeding, mice were sacrificed to determine whether Sorcs1 deficiency exacerbated brain Aβ accumulation. Sorcs1 -/-x APP/PSEN1 mice displayed similar levels of total brain Aβ40 (Fig. 4a-f ), Aβ42 (Fig. 4 g-l), Aβ42/40 ratio (Fig. 4m-r) and Aβ oligomers (Fig. 4s-x) compared to APP/PSEN1 mice in both male and female cohorts. Similarly, no differences in Aβ plaque load were detected between the two genotypes (Fig. 4y). Notably, females had much higher levels of insoluble Aβ in the formic acid fraction than males.
As expected due to the absence of human Aβ, no plaques were observed in either male or female WT and Sorcs1 -/-mice (data not shown). Interestingly, Aβ was deposited as cerebral amyloid angiopathy (CAA) in large and small blood vessels within the brains of male and female APP/PSEN1 mice as well as Sorcs1 -/-x APP/ PSEN1 mice (Fig. 5). Quantification of CAA is challenging at best, but, at the qualitative level, male and female Sorcs1 -/-x APP/PSEN1 mice appeared to have less vascular Aβ than APP/PSEN1 mice and when present was confined to large vessels (Fig. 5). Microhemorrhages were also detectable within the brains of all cohorts (Fig. 6). As observed with the CAA, the brains of both male and female APP/ PSEN1 mice appeared to display more cerebral microhemorrhages than WT mice, and the brain microhemorrhage count was less in Sorcs1 -/-X APP/PSEN1 mice than in APP/PSEN1 mice. Unexpectedly, the microhemorrhage count in Sorcs1 -/mice was similar to that of Sorcs1 -/-X APP/PSEN1 mice despite the absence of human Aβ or CAA in Sorcs1 -/mice.

Discussion
Multiple epidemiological studies have demonstrated an association between insulin resistance and T2D and increased risk for cognitive impairment in humans [4,13,34,62]. SORCS1 has been associated with both T2D and AD. Furthermore, the human linkage to SORCS1 is stronger for women in both T2D and AD. Here, we tested the hypothesis that a mouse model with a compound proamyloidogenic/prodiabetogentic genotype might be useful in studying how the two conditions coexist with possible bidirectional exacerbation (i.e., that cerebral amyloidosis exacerbated glucose intolerance and vice versa). In order to test this hypothesis, we assessed longitudinal aging-related memory behavior, glucose homeostasis, and Aβ pathology in both male and female WT, Sorcs1 -/-, APP/PSEN1 and Sorcs1 -/mice crossed with APP/PSEN1 mice.
In the present study we demonstrate that when young, both male and female Sorcs1 -/-, APP/PSEN1 and Sorcs1 -/-x APP/PSEN1 mice display normal glucose homeostasis. This fits with previous studies that have shown that young APP/PSEN1 mice have similar glucose tolerance to WT mice when fed a standard diet [17]. At this age, crossing Sorcs1 -/-mice with APP/PSEN1 did not precipitate earlier onset of metabolic dysfunction in either males or females. Interestingly, young female Sorcs1 -/-displayed lower body weight, shorter body length and reduced lean mass when compared to WT mice. Female Sorcs1 -/-mice were also shorter in length than APP/PSEN1 and Sorcs1 -/-x APP/PSEN1 had reduced lean mass compared to WT mice. This is the first study to report such difference and the cause is unknown. Lean mass is important for insulin action and metabolic control in that it has salutary effects. If two mouse models have the same metabolic flexibility but one has a higher lean mass, one can argue that the beneficial effects of higher lean mass could offset disturbances in metabolic flexibility.
When behavior was assessed in the Y-maze, the Sorcs1 -/mice X APP/PSEN1 cross was associated with significantly impaired learning behavior in the females and trend for worse performance in Sorcs1 -/and APP/PSEN1 mice, although this did not reach statistical significance. In males, all cohorts had a worse performance in the Y-maze than WT mice but only Sorcs1 -/-were significantly impaired. When behavior was re-assessed in the females, cohorts were no longer significantly different, the reason for this appears to be the worsening of memory in the WT group, the reason for this is unknown but could possibly due to the high fat diet, or age related decline, rather any change in the other cohorts which were already poor. In the male cohort, the Sorcs1 -/mice continue to be impaired when compared to WT mice. When NOR was assessed, both male and female WT mice displayed intact learning behavior whereas male and female mice of Sorcs1 -/-, APP/ PSEN1 and Sorcs1 -/-x APP/PSEN1 genotypes were all impaired. This is the first study to demonstrate learning deficits in Sorcs1 -/-or Sorcs1 -/-x APP/PSEN1 mice. Unexpectedly in the present study both male and female APP/PSEN1 mice and Sorcs1 -/-mice displayed very early onset learning behavior deficits in NOR. For this reason, the NOR test was not repeated at later time points. Overall, the behavioral data for the Y-maze and NOR for both genders trend in the same direction for the most part. We cannot discount that lack of significance between some groups may be due to lack of power or sensitivity of these tests at the time points tested. Additional studies are required to dissect the temporal onset of NOR deficits in the Sorcs1 -/-x APP/PSEN1 mice and any potential sex differences in Y-maze.
In order to deliver a metabolic stress to unveil a metabolic phenotype, APP/PSEN1 mice have previously been placed on either a high fat diet or crossed with transgenic models of obesity and/or diabetes. When we placed mice on a 60 % HFD, we found that male APP/PSEN1, display evidence of transient elevated fasted plasma insulin as well as both glucose intolerance and insulin insensitivity. During the development of T2D, β-cells initially compensate for insulin resistance by increasing the amount of insulin in circulation. This is consistent with previous studies where APP/PSEN1 mice have increased fasted insulin, develop glucose intolerance and become less insulin sensitive when placed on a HFD [9,35,52,70]. These results demonstrate that APP/PSEN1 mice are susceptible to metabolic impairment on a HFD, consistent with a predisposition for the development of diabetes. Male Sorcs1 -/-and Sorcs1 -/-x APP/PSEN1 displayed impaired glucose tolerance when compared to WT mice. Unexpectedly, metabolic abnormalities in male Sorcs1 -/-mice were not exacerbated when crossing this line with the APP/ PSEN1 mouse. Indeed, there was a clear trend for Sorcs1 deficiency to mitigate APP/PSEN1 phenotypes. In general, the metabolic phenotype in the Sorcs1 -/-line was much milder than expected, especially when compared to the APP/PSEN1 line.
The female mice in the present study were resistant to metabolic abnormalities when placed on the diet for only 4 weeks. It is well known that female mice are resistant to metabolic effects of a high fat diet, which may account for their delayed onset of metabolic dysfunction when compared to males here. Petterson et al. [44] showed that in response to a high fat diet, plain C57Bl/6 male but not female mice developed hyperinsulinemia, hypotrophic islets, low grade systemic inflammation and an increased macrophage population in their intraabdominal adipose tissue. In contrast, female mice had an elevated anti-inflammatory regulatory T cell population in their adipose tissue [44]. In the present study, after 12 weeks of HFD, female APP/PSEN1 mice displayed metabolic abnormalities and increased fasted plasma insulin levels. Both female APP/PSEN1 and Sorcs1 -/-x APP/ PSEN1 mice had impaired glucose tolerance and APP/ PSEN1 mice had impaired insulin sensitivity. Unlike males, the female Sorcs1 -/-x APP/PSEN1 had significantly worse glucose tolerance and insulin sensitivity than female Sorcs1 -/-mice. The lack of deficit in glucose tolerance in males between Sorcs1 -/-x APP/PSEN1 and Sorcs1 -/is most likely due to the latter showing a significant deficit less visible in the females.
While APP/PSEN1 overexpression facilitated glucose dysregulation, despite differences in glucose metabolism seen between Sorcs1 -/and Sorcs1 -/-x APP/PSEN1 after 12 weeks of HFD, unexpectedly, Sorcs1 -/crossed onto APP/PSEN1 mice did not modulate brain Aβ levels or oligomer levels beyond the changes already induced by the APP and PSEN1 mutations. The unexpected lack of worsening of pathology in the Sorcs1 -/-x APP/ PSEN1 mice may indicate that Aβ deposition may be the cause rather than the effect of alterations in glucose metabolic observed. In support of this formulation, we have previously described the appearance of metabolic inflexibility in APP/PSEN1 mice following high fat feeding (Ruiz et al.,in press [56]). In the present study, Aβ levels were measured from whole forebrain extracts; therefore, we cannot exclude possible subtle changes in regional Aβ levels relevant to memory behavior (such as the hippocampus and entorhinal cortex) and/or regulation of energy balance (the hypothalamus). Future studies would need to pick apart regional Aβ response and hypothalamic insulin signaling in Sorcs1 -/and Sorcs1 -/x APP/PSEN1 mice. We cannot also exclude potential subtle changes in endogenous murine Aβ levels as have previously shown that female Sorcs1 -/mice have elevated endogenous murine levels compared to WT mice [29]. Previously, Reitz et al. [53] reported that genetic variants in SORCS1 are associated with increased risk of AD and that over-expression of SorCS1 reduces Aβ levels and γ-secretase activity, whereas suppression of SorCS1 increases Aβ levels and APP processing by γ-secretase activity. Unexpectedly, however, ablation of Sorcs1 in the present study impacted memory without affecting production/accumulation of whole brain levels of Aβ. Although we can't discount subtle altered levels of endogenous murine Aβ in Sorcs1 -/-mice as we have previously reported [29] nor changes in regional Aβ as discussed above, there may be other more generalized effects that SORCS1 may have on memory and neurodegeneration. Prominent neuronal expression of the Vps10p-domain sorting receptor family, of which SORCS1 is a member, has resulted in recent data supporting their role in neuronal activity, plasticity related processes and neurogenesis [53]. Furthermore, several studies have demonstrated links between SNPs in SORCS1 and cognition. Printy et al. [49] assessed atrophy profiles against genetic markers in the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, finding associations of cerebral atrophy with SNPs on APP but also ventricular enlargement with SNPs on SORCS1. Reitz et al. [53] also assessed the impact of genetic variation in SORCS1 and memory retention and found three SNPS were associated with memory retention.
Studies in other systems examining the relationship between AD pathology and diabetes have yielded variable results. One study revealed no association of the two pathologies [16]. In another study, cerebral amyloidosis was only exacerbated in diabetics carrying the APOE ε4 allele [43], while in another, the result was that AD pathology was reduced among diabetics [7]. Interestingly, one of these studies also showed that the combined effects of carrying the ε4 allele and being diabetic increased the risk for a dementia that was clinically classified as vascular dementia [43]. Previous studies have also shown an association between both diabetes and vascular dementia [27,30,42]. More recently, Arvanitakis et al. [3] found that diabetes was associated with increased risk of brain infarction but not AD pathology in older individuals. At a recent workshop on AD and T2D at the US National Institutes of Health, the consensus was that the clinical cognitive decline in diabetics was probably more attributable to vascular dementia (Stoeckel, Gandy, and Arvanitakis, manuscript in preparation).
In the present study we assessed vascular Aβ, finding Aβ in blood vessels within the brains of both male and female APP/PSEN1 mice as well as Sorcs1 -/-x APP/ PSEN1 mice. Qualitative assessment revealed that both male and female APP/PSEN1 mice displayed abundant deposits of Aβ in both large and small caliber vessels. Female Sorcs1 -/-x APP/PSEN1 mice appeared to have less vascular Aβ than APP/PSEN1 mice and, when present, CAA was restricted to large caliber vessels. Male Sorcs1 -/-x APP/PSEN1 appeared to display relatively less vascular Aβ, and, again when present was restricted to large vessels. It is possible that reduced CAA could be a marker of altered Aβ trafficking and/or clearance, which should be an area of interest in future study. We also assessed microhemorrhages, and bleeds were detected in all mice assessed after HFD. Microhemorrhage in the WT mice is consistent with increased risk of hemorrhagic strokes associated with poor diet and obesity [45]. Both male and female APP/PSEN1 mice tended toward greater numbers of bleeds than WT mice. This is consistent with previous studies that have shown that APP/PSEN1 mice are prone to vascular alterations, which are exacerbated on a HFD [52]. Both male and female Sorcs1 -/-mice revealed more bleeds than WT mice.
The extent to which dementia in humans is due to AD rather than to vascular dementia is an area of great interest, and the pathology of dementia in T2D is usually mixed [1,43]. Human studies have yielded a range of results, with some, but not all, linking cerebral amyloidosis to insulin resistance [68]. Mouse models have been similarly variable. When APP mice are crossed with the leptin-deficient ob/ob model they showed no change in Aβ burden, however both amyloid angiopathy and angiitis were increased together with exacerbated cognitive dysfunction [66], while in contrast when Tg2576 mice are crossed with Irs2-/-insulin resistant mice show reduced Aβ pathology and cognitive improvement [24]. Recently, Ramos-Rodriguez et al. [51] crossed APP/ PSEN1 mice with the with the morbidly obese and diabetic db/db model, finding fewer plaques, a shift in soluble/insoluble pool of Aβ, increased microglia activation and increased hemorrhage burden in the crossed mice. Niedowicz et al. [38] also crossed APP/PSEN1 mice with db/db mice, finding that the crossed mice displayed extreme obesity, diabetes and parenchymal Aβ deposition with strikingly severe cerebrovascular pathology of aneurysms and small strokes. As with the present study, the crossed mice in the Niedowicz study did not develop elevated levels of Aβ deposition. The crossed mice had more impairment in learning behavior in the Morris water maze than db/db or APP/PSEN1 alone. Neither the crosses of Ramos-Rodriguez [51] nor those of Niedowics [38] predicted the results that we observed wherein the diabetogenic mutant mitigated the phenotype of the AD model.
Previous studies in diabetic models have documented increased angiogenesis and arteriogenesis, consisting of immature, unstable blood vessels leading to increased permeability of the blood-brain barrier [32,38,47,48]. In the present study, we found evidence of both vascular Aβ and microhemorrhages, however, Sorcs1 -/-x APP/ PSEN1 mice did not display an exacerbated phenotype; if anything, there was a trend toward reduced vascular Aβ and hemorrhage load in the Sorcs1 -/-x APP/PSEN1 mice when compared with the APP/PSEN1 mice. Similarly, a recent study in which diabetes was induced in APP/PSEN1 mice using streptozotocin treatment documented reduced amyloid angiopathy when compared to sham controls. This was associated with a slight reduction in plaque load and increased microglial activation [50]. In the case of the genes and models that we have studied, we tentatively conclude that metabolic abnormalities in female Sorcs1 -/-x APP/PSEN1 mice are not linked to changes in Aβ metabolism nor in brain bleeds; however, further detailed quantitative investigation of the vasculature in this model would be required to establish this proposed conclusion. The role of inflammation in the Sorcs1 -/-x APP/PSEN1 mice also requires further investigation.
In accordance with the present study, Kebede et al. [23] assessed the metabolic profile of female Sorcs1 -/mice over time, finding that without intervention they do not develop diabetes. Glucose intolerance was apparent in Sorcs1 -/mice tested at 20 weeks of age. As with the present study, no mutation-related differences in fasted plasma insulin were observed up to 18 weeks. No abnormality in insulin sensitivity was detected at 12 weeks. The authors went on to place the Sorcs1 -/mice on a HFD, and, while they saw trends toward elevated glucose levels in response to a GTT, the trends were not significant. When Kebede et al. [23] crossed the female Sorcs1 -/-mice with the leptin-deficient model of obesity (ob/ob), the crossed mice went on to develop diabetes, with elevated levels of fasting glucose and insulin as well as impaired glucose tolerance and sensitivity. On top of a diabetic phenotype, Sorcs1 deficiency in the ob/ob model lead to severe depletion of insulin granules in pancreatic β-cells, leading the authors to suggest that Sorcs1 is involved in vesicular trafficking and possibly biogenesis of insulin granules. As with the present study, the metabolic phenotype of Sorcs1 -/mice as characterized by Kebede et al. [23] required genetic or diet-induced obesity to be triggered. Therefore, as with the unexpectedly mild metabolic and amyloidosis phenotypes of Sorcs1 -/-x APP/PSEN1 mice as opposed to the exacerbated phenotypes of other combined diabetic/cerebral amyloidotic mouse models, the APP/PSEN1 transgenes were apparently less stressful to beta cells and/or insulin action when compared to the metabolic stress induced by db/db. Previous studies show that the combination of a genetic proamyloidogenic phenotype and potentially diabetogenic phenotype (induced by either drug, diet, or genetic manipulations) can lead to exacerbation of either the proamyloidogenic component of the phenotype, the prodiabetogenic component of the phenotype, or exacerbation of both components of the phenotype [8,18,20,38,46,51,52,66]. Unexpectedly, not only did the two genetic lesions not exacerbate each other in the present study, the Sorcs1 deficiency tended to partially correct the diabetogenic phenotype of the APP/PSEN1 mice. It is worth noting that "bidirectional exacerbation" is not universally observed, since Murakami et al. [37] reported that insulin receptor mutation-induced insulin resistance failed to exacerbate an Alzheimer's-like phenotype in mice. Thus, while our experience is in the minority, our observations of no bidirectional exacerbation are not singular.
Kebede et al. [23] have demonstrated that stressing Sorcs1-/mice by crossing with db/db is required to reveal the islet failure phenotype. The formulation for the basis of that model is that missorting and misprocessing of insulin constitutes the underlying pathogenesis. Our data would suggest that the APP/PSEN1 proamyloidogenic phenotype is not sufficiently metabolically stressful to cause islet failure. This is our qualitative explanation for the non-exacerbation of the metabolic phenotype. Our formulation for the mechanism by which Sorcs1-/modulates Aβ generation involves retromer dysfunction and promotion of generation of the pool of Aβ that arises from wildtype APP. For wildtype APP, 75 % of secreted Aβ is generated by endosomes and 25 % is generated in the trans Golgi network (TGN). However, when the substrate is Swedish mutant APP (i.e., APPKM670/671NL), the compartmental stoichiometry is altered, such that 75 % of secreted Aβ derived from Swedish APP is generated in the TGN and only 25 % is generated in the endosome. We would propose that Sorcs1-/-, by disrupting protein sorting in the TGNendosomal system, reduces access of TGN BACE (β-APP sorting enzyme) to Swedish APP, thereby partially reversing the proamyloidogenic phenotype.

Conclusions
In the present study we assessed memory, glucose homeostasis, and brain biochemistry and pathology in male and female WT, Sorcs1 -/-, APP/PSEN1, and Sorcs1 -/-X APP/PSEN1 mice. Male mice with either the APP/PSEN1 or Sorcs1 -/genotype displayed earlier onset and persistent impairment in both glucose homeostasis and in learning behavior. Unlike prior examples in the literature, the behavioral and metabolic abnormalities in male mice were not exacerbated when the two disease model mice (Sorcs1 -/-models T2D; APP/PSEN1 models AD) were crossed. In contrast to the male Sorcs1 -/-X APP/PSEN1 mice, female Sorcs1 -/-X APP/PSEN1 mice showed later onset but worse metabolic dysfunction than Sorcs1 -/-knockout mice and worse memory than WT mice. The deletion of Sorcs1 from APP/PSEN1 mutant mice led to no obvious changes in brain levels of total or oligomeric Aβ peptide. In general, unexpectedly, there was a trend for gene targeting of Sorcs1-/to mitigate, not exacerbate, the metabolic, behavioral, and amyloid pathologies. Mixed pathology is a feature of someperhaps mostlate life dementia, raising the likelihood that compound genotypes may be required in order to approximate the pathology in mice, these results indicate that compound genotypes of highly complex diseases (such as AD plus T2D) may yield unexpected phenotypes that limit their utility.

Competing interests
Within the past 5 years, SG has held research grants from Amicus Therapeutics. SG is also a member of the Data and Safety Monitoring Board for the Pfizer-Janssen Alzheimer's Immunotherapy Alliance.
Authors' contributions EMK designed the experiments with CB, MEE and SG. EMK, HHR, SHK, JCK and WH performed the experiments. EMK analyzed the data. AA created the Sorcs1-/mice and contributed to the design of the experiments and the editing of the manuscript. EMK, MEE and SG wrote the manuscript. CB, MEE and SG supervised the project. CB and WLK provided material support. All authors read and approved the final manuscript.