Tmem108 mutant (Tmem108-LacZ) mice (MMRRC: 032633-UCD) were described previously (16, 22), and briefly, β-galactosidase protein cassette, including both stop code and a polyadenylation termination signal, was replaced the first coding exon of Tmem108 (exon 3) by homologous recombination. Mice were housed in a 12 h light/dark cycle room with ad libitum access to water and rodent chow diet. The animal protocols in this study were approved by Nanchang University Medical Sciences Committee for research on vertebrate animals, following EN Directive 2010/63/EU on the protection of animals used for scientific purposes. After terminal experiments, mice were euthanized by carbon dioxide inhalation.

According to the manufacturer’s instructions, the total RNA from mice tissues was extracted by TRIzol Reagent (Invitrogen, Thermo Fisher Scientific, USA). Real-time PCR was performed as described previously (27). The relative expression values were calculated relative to β-Actin by using the 2^-ΔΔCT methods, and the control group was set to a value of 1. In the end, the mutant group was normalized to the control. Primer specificity was tested by melting curve analysis, and the PCR product was verified by agarose gel electrophoresis. The real-time PCR primer sets are listed in Table 1 .

For GTT, after fasting the mice overnight for about 16 h, intraperitoneally (i.p.) glucose (2.5 mg glucose/g body weight) injection was performed on the mice. Then, plasma glucose concentrations were measured by a glucometer (Roche Accu-Chek Performa, USA) on each mouse before 20, 40, and 120 minutes after the i.p. glucose injection. Tail sipping was used to get blood. Before snips, the tail end was dipped into 0.25% bupivacaine for local anesthesia to reduce pain.

IRT was performed after a 6-h fast, and plasma glucose was measured before 20, 40, and 120 minutes after an i.p. insulin (0.75 mU/g body weight) injection.

Food intake, water consumption, feces outtake, and urine production were assessed by a metabolic cage (Tecniplast Metabolics Biological Monitoring, Italy). Briefly, 2-month-old male mice were individually housed in the metabolic cage 12 h light/dark cycle with ad libitum access to rodent chow diet and water. The data were collected by 12-h cycles for one month after a one-week adaption, and the data were normalized to the body weight.

Whole-body oxygen consumption (VO₂), carbon dioxide expiration (VCO₂), RER, and ambulatory activity were assessed by a Comprehensive Lab Animal Monitoring System (CLAMS, Columbus Instruments, USA). Briefly, 2-month-old male mice were acclimated to CLAMS cages individually with food and water ad libitum for 24 h. The data on oxygen consumption (VO₂) and carbon dioxide production (VCO₂) were recorded every 10 min for 3 d at 23°C following adaption. Infrared beams measured the ambulatory activity in a single horizontal plane (X-axis, Y-axis) and a vertical direction (Z-axis) (OPT-M3, Columbus Instruments). The RER was calculated as VCO₂/VO₂ automatically by the software.

Male mice were handled for habituation for 3 d before behavioral tests. The open field was a square chamber (40 × 40 × 30 cm). Mice were placed in the chamber individually and monitored for 10 minutes with an overhead video tracing system to record the animal’s locomotor activity.

Mice balance and motor coordination were measured with a rotarod apparatus (Unibiolab, Beijing, China). The first day was used for acclimation training, which consisted of two trials with a 15-min resting interval between two trials. The first trial was performed at a 13-rpm speed for 15 min, and the speeds in the second trial ramped from 13 rpm to 25 rpm in 15 min. After acclimation training, mice were used for the test trial, including speeds ramping from 13 rpm to 25 rpm in 15 min and maintaining 25 rpm for 30 min. Mice were resumed on the rotarod if they stopped running or fell off four times within 60 s of each fall. The latency falling off the rotarod of mice was recorded individually.

A muscle fatigue task test was performed on a six-lane treadmill (Columbus Instruments, USA). The speed was accelerated with a specified ramp from the starting speed of 0 m/min to the maximum speed of 20 m/min. Mice ran until exhaustion. Definition of the exhaustion was the mice’s inability to restart running by continuous contact (no less than 5 sec) even with the electrical shock at the end of the running lane. The running time of each mouse was recorded. Moreover, the distance was calculated based on the treadmill’s speed and the mouse running time.

Pre-pulse inhibition (PPI) was performed in sound-attenuated chambers (Med Associates Inc., St. Albans, VT, USA) as previously described (16) with minor revision. In brief, a 65-dB background whiter noise was maintained throughout the test sessions. Mice were acclimated to the testing chamber for 5 min as the first test session. Then, mice were individually placed in a tube fixed on a plastic frame to monitor movement. After 12 initial startle pulses (20 ms, 120 dB) were presented, 12 prepulse/startle trials (prepulse, 20-ms white noise at 70, 80, or 90 dB at 100-ms intervals; 80-ms interval; startle pulse, 20 ms,120 dB) were performed immediately. The initial startle pulses provided a baseline of startle reactivity for the following test session. The different trial types (12 times for each trial type) were presented pseudo-randomly with a 15-s inter-trial interval (10-20 s), and two consecutive trials were always different. Mice movements were analyzed for 100 ms from startle stimulus onset. Each complete test session lasted about 25 min. For each pre-pulse intensity, PPI (%) was calculated as: [1 - (startle amplitude on “pre-pulse followed by pulse” trial/startle amplitude on “pulse-alone” trial)] × 100.

The study was not pre-registered. For the assignment of each experimental group, no unique randomization method was applied. Sample sizes were determined by reference and experience. The experiments reported in this study did not require institutional approval. Exclusion criteria were not employed in this research. All Data statistics were conducted by GraphPad Prism 6.0 (GraphPad Software Inc, USA), and the values were mean ± SEM. Different people performed the analysis and experimental group assignments than the experimenter. As indicated in the figure legends, the data were analyzed by two-way ANOVA or unpaired t-test where appropriate. Significance between two groups were indicated as * p<0.05, ** p<0.01 or *** p<0.001, and p > 0.05 was considered no significant.

Tmem108 expression was verified by qPCR ( Supplementary Figure S1 ) in several indicated tissues (brain, liver, muscle, and pancreas) in adult male wild-type mice, which are mainly related to glucose metabolism. The muscle and liver are the main mobilization organ of stored glycogen, which provide the most rapid and available glucose source for themselves and other organs. The pancreas maintains the plasma glucose balance by different exocrine hormones, such as insulin and glucagon, and diabetes is a disability of pancreas function. The brain constitutes no more than 5% of the body weight. However, except for acetone body utilization, the brain’s energy consumption accounts for 20%-30% of the total glucose utilization.

Moreover, the hypothalamus in the brain could secrete insulin (central insulin) to regulate glucose levels. Tmem108 was widely expressed in mice’s indicated tissues, highly in the brain (including the hypothalamus, hippocampus, prefrontal cortex, olfactory cortex, and medulla) and liver, lowly in pancreas and muscle.

Tmem108 mutant mice reached adulthood were fertile and exhibited no apparent health issues, including average brain size and brain weight ( Supplementary Figure S2 ). We performed the growth curve analysis to characterize the growth rates of Tmem108 mutant mice ( Supplementary Figure S2 ). Tmem108 mutant mice were significantly smaller than the control (Wild-type) mice at P50 and P100. The growth curve difference was significant between Tmem108 mutant mice and the control mice, indicating the slow growth rate in the Tmem108 mutant mice during the postnatal period.

Carbohydrate metabolism was investigated in Tmem108 mutant mice using glucose tolerance (GTT) and the insulin resistance test (IRT). A GTT is used to measure the efficiency of sugar metabolism. When the sugar is metabolized abnormally, plasma glucose levels will be too high or too low to maintain homeostasis, resulting in diabetes or hypoglycemia. Insulin is essential to process food’s sugars to energy, and an IRT is used to analyze the body’s response to the insulin hormone. Plasma glucose level is the same indicator in both GTT and IRT, and a similar analysis is a calculation of the area under the curve (AUC), which depends on integrating glucose values in the GTT or IRT. Tmem108 mutant mice exhibited slower glucose clearance after bolus injection of glucose ( Figures 1A, B ) and slower clearance following insulin injection ( Figures 1C, D ), which means a lowered response to the insulin hormone in the mutant mice. This result indicated that Tmem108 mutant mice exhibited impaired glucose homeosis, with metabolized glucose more slowly and higher insulin resistance than the control mice.

We utilized simple metabolic cages for the tests over one month to detect the mice’s consumption and production ( Figure 2A ). Compared to the control mice, food intake and feces excretion were essentially lower in Tmem108 mutant mice ( Figures 2C, D ). Furthermore, water consumption was identical in the two mice ( Figure 2E ), and urine production in Tmem108 mutant mice was slightly higher in the night than in the control mice ( Figure 2F ). However, there were no differences in the weight between Tmem108 mutant mice and the control mice ( Figure 2B ). The results suggested that Tmem108 deficiency in the mutant mice might shift the metabolism homeostasis unless the mutant mice have severe issues in the metabolic architecture.

A comprehensive laboratory animal monitoring system was also used to record the mice’s whole-body oxygen consumption and carbon dioxide production and measure the ambulatory movements in similar home-cages. Oxygen consumption (VO₂) was substantially lower in Tmem108 mutant mice than wild-type, both in the light and dark phases (as expected for nocturnal animals, Figures 3A, B ). The mice were cultured in the metabolism cages for 3 d, and the data of the last 24 h were used for analysis. The RER, also known as a respiratory quotient (RQ), is the volume ratio of carbon dioxide released to oxygen absorbed during respiration. The results showed that RER in Tmem108 mutant mice was higher than in the control mice in light ( Figures 3C, D ), indicating the mutant mice preferred consuming more carbohydrates or fewer lipid compounds than the control mice. Furthermore, Tmem108 mutant mice might have a low metabolism level because of less oxygen consumption in the mutant mice.

The locomotor activity of mice in the metabolism cages was represented by ambulatory counts ( Figures 3E, F ). Ambulatory counts of Tmem108 mutant mice were lower than the control mice in three axes, which meant horizontal locomotion (Ambulatory counts in X-axis and Y-axis) and vertical exploration (Rearing behavior, ambulatory counts in Z-axis) were reduced in Tmem108 mutant mice ( Figure 3F ). The reduction in ambulatory counts suggested a significant inhibitory/abnormal effect in Tmem108 mutant mice.

RER in Tmem108 mutant mice was higher than in control mice, indicating a high glucose oxidation ratio to lipid metabolism. Therefore, the relative expression of the rate-limiting enzyme of glycolysis (Hk) and glycogenesis (Pck) in glucose metabolism were investigated ( Figures 4A–D ), an increase of both them indicating high carbohydrate metabolism in the mutant mice, consistent with the high RER of the mutant mice. Moreover, relative expression of the rate-limiting enzyme in lipid metabolism also was determined by qPCR, including de novo lipogenesis crucial gene (Acaca), cholesterol synthesis rate-limiting enzyme (Hmgcr), and acetone synthesis critical gene (Hmgcs). Strikingly in the mutant mice, Acaca expression decreased in all the detected tissues, indicating a low level of lipid metabolism.

Furthermore, several hormone-associated genes (Gcg, Ide, Ins, Igf, and Pdges) regulating plasma glucose levels and lipid metabolism were also verified in the pancreas and hypothalamus ( Figures 4E, F ). mRNA of the detected hormone gene-regulating glucose and lipid metabolism decreased in the mutant mice.

Before the mice were used for the muscle fatigue test, the mice’s locomotor activity and balance ability were assessed by open field test and rotarod test separately. Firstly, Tmem108 mutant mice have no difference from the control mice in the total travel distance and the average speed evaluated by the open field test ( Figures 5A–C ). And then, in the rotarod wheel test ( Figures 5D–F ), Tmem108 mutant mice have similar performance with the control mice in the balance learning and keeping. Therefore, damnify of motor coordination ability was not evidenced in Tmem108 mutant mice. Intriguingly, Tmem108 mutant mice manifested weak fatigue resistance in the muscle fatigue task test ( Figures 5G–I ) because the mutant mice had quickly exhausted when running on the treadmill.

Pre-pulse inhibition (PPI) is a psychophysiological measurement of sensorimotor gating. The startle reflex is suppressed when the startling stimulus is preceded by a subthreshold stimulus (28, 29). PPI is consistently standard in rodents and humans (28). PPI in Tmem108 mutant mice decreased, as described before (16).

Metformin, the most commonly prescribed medication to treat diabetes (30, 31), was executed to treat the aberrancy of glucose metabolism in Tmem108 mutant mice ( Figure 6A ). Before metformin treatment, plasma glucose levels in Tmem108 mutant mice were higher than in the control mice. Intriguingly, metformin-induced a decline of plasma glucose levels in Tmem108 mutant mice significantly ( Figure 6B ). Whether treatment with metformin or not, the baseline response to auditory-evoked startle stimulus (120 dB) was no different between Tmem108 mutant mice and the control mice ( Figure 6C ).

After metformin treatment for ten days, PPI in Tmem108 mutant mice was lower than the control mice at pre-pulse 75 dB and 80 dB ( Figure 6D ). However, after ten-day metformin treatment, PPI decreased Tmem108 mutant mice were found only at pre-pulse 80 dB ( Figure 6E ). Therefore, metformin slightly enhanced startle responses of Tmem108 mutant mice and the enhancement was also found in the two-way ANOWA analysis, genotype variance variation from 0.001 (before metformin treatment, Figure 6D ) to 0.032 (after metformin treatment, Figure 6E ).

There is a wide array of different mechanisms for illuminating the SCZ pathogenesis, including but not limited to neurodevelopmental deficits in the brain circuity, defects in one or more neurotransmitters, viral infection, and multifactorial inheritance (32). Rates of SCZ are higher in patients’ relatives than in common individuals. Studies on twins and adoption declare a tenfold increase in risk associated with the affected first-degree family. However, SCZ genetic transmission is not following Mendelian single-gene inheritance patterns, and it probably owes to multiple risk genes acting with environmental contributors. While their heritability has built up some genetic factors in SCZ, several genetic factors and environmental contributors may be sufficient to provoke the outbreak of SCZ (5–7). Tmem108 is not only a susceptible gene of SCZ (10, 11, 15–17) but also a risk gene of other psychotic disorders, such as bipolar disorder (10–12) and major depressive disorder (14), which suggest Tmem108 have a complicated role in the psychotic pathology.

For SCZ, genetic factors from the family are one of the leading causes, implying at least one risk gene in this family (33, 34). However, environmental factors also contribute to the onset of SCZ (35, 36), and the environmental factors take effect because of the risk gene existence, such as Tmem108. The environmental factors may take part in the diseases through many routes, directly or indirectly interacting with the risk genes (33, 34, 37), and whether direct or indirect interaction between Tmem108 and SCZ is unclear.

Comparing with the control mice, Tmem108 mutant mice exhibited higher glucose intolerance in GTT and had higher plasma glucose levels before metformin treatment. In some reports, dysregulated cerebral glucose metabolism is regarded as a beginning matter of SCZ pathogenesis, and it causes the secondary alteration in peripheral glucose metabolism. The immunity may be associated with the inauguration in SCZ pathogenesis, causing abnormality in glucose metabolism. Common plausible elements may be accountable for the co-occurrence of glucose abnormality and immune disturbances in SCZ (38).

The blood glucose is affected in the Tmem108 mutant mice. From the expression of glycolytic and lipogenic genes, muscle is one of the biggest organs in the body (39), and it is a major contributor to glucose homeostasis (40, 41). Though the brain constitutes no more than 5% of the body weight, the brain’s energy consumption accounts for 20%-30% of the total glucose utilization. So, brain contribution in blood glucose is not negligible. The pancreas and hypothalamus are upstream in glucose regulation (42–44), contributing more.

Abnormal glucose homeostasis emerges from the illness onset in SCZ, indicating that SCZ patients are at high risk of diabetes mellitus (DM) before medications (45). The co-occurrence of DM and SCZ was already documented in the 19th century. It is long before the occurrence of antipsychotic medications. Moreover, at that time, the propensity of diets could not induce metabolic derangements (45).

There are two main types of DM. Type I DM is caused by the damage of the pancreatic β cells. 90% of DM belongs to type II DM, accompanying insulin resistance and disruptive secretion (46). Rates of type II DM in SCZ are estimated to be 10% to 15%, much higher than that in ordinary individuals, and antipsychotic medication adds complexity to this correlation by adding DM’s risk (45, 47). Intriguingly, studies in SCZ patients’ relatives have disclosed elevated prevalence of DM similar to the result from a meta-analysis of the first episode psychosis (9, 45, 48, 49).

Treatments for SCZ mainly include typical antipsychotics emerging in the 1950s (also as first-generation antipsychotics) and atypical medication (known as second-generation antipsychotics, SGAs) 1980s. However, in recent years, the metabolic side-effects of SGAs have been substantially reported, such as impaired glucose regulation and manifest DM (50). The asymptomatic preclinical phase of DM is also distinguished by impaired glucose regulation and metabolism dysregulation. Some recent studies (51) have also advocated a potent high risk of DM in first episode SCZ populations before atypical antipsychotic administration, which may be associated with diabetogenic lifestyle factors and a genetic predisposition for DM (51). Atypical treatments replace typical treatments of SCZ because of extrapyramidal side effects and inadequate therapeutic potential (52). Although second-generation treatment of SCZ is secure and effective for movement disorders, high-risk metabolic side effects are related to them, which underly fasting hyperglycemia, impaired glucose tolerance, prediabetes, and diabetes (53). Antipsychotic medications are associated with severe metabolic abnormalities and fail to remedy premorbid cognitive deficits, though they are the cornerstone treatments for SCZ (9).

Tmem108 mutant mice exhibited high insulin resistance in IRT. Central insulin resistance is the intersection of cognitive and metabolic diseases (54). Central insulin can mediate striatal dopamine levels (55), and striatal dopamine and central insulin can manage peripheral glucose metabolism (56–58). From a comprehensive analysis of 14 studies composing 1345 individuals, Toby Pillinger et al. found that participants with the first-episode SCZ exhibited high-level fasting plasma glucose and high insulin resistance (45). It is still difficult to establish the direct influence of atypical antipsychotic medication on glucose metabolism. However, short-term dosing research has suggested that antipsychotic medication involves weakened insulin sensitivity and impaired glucose tolerance (51). An insulin signaling pathway is impaired in first-episode psychosis patients and siblings (59–61), and the intriguing correlation between SCZ and DM is always a fruitful and active inquiry (9).

The highest active organ in whole-body metabolism is the brain (38). The brain consumes 20% of the whole oxygen supply and half of the whole glucose supply, although the brain-to-body mass ratio is no more than 5%. Due to the high energy consumption, cerebral glucose supply must be continuous, while energy storage substance is limited in the brain. Instead, ketone bodies and lactate can also be consumed as an energy supply in the brain. Ketone bodies originate from fatty metabolism, and lactate is the product of anaerobic glycolysis. Insulin and Insulin receptors are distributed through the CNS, not limited to be expressed in the hypothalamus (62).

Tmem108 was widely expressed in the brain, including the hypothalamus, which was considered one area secreting insulin to regulate glucose levels. Therefore, we explored several hormone gene expressions adjusting glucose metabolism in the hypothalamus. The results showed that hormone genes associated with glucose and lipid metabolism were abated in the mutant mice.

The CNS insulin origin continues to be an argument between central and peripheral origins. In 1967, evidence showed that peripheral insulin infusion led to an insulin elevation in cerebrospinal fluid, suggesting peripheral insulin can pass through the blood-brain barrier (BBB). Insulin is transported to cross the BBB by the receptor-mediated system, such as N-methyl-D-aspartate receptors (NMDARs) and cannabinoid receptor type 1(CB1) (57, 63, 64). Evidence for the central origin of insulin is that some neurons can synthesize and secrete insulin due to the appearance of insulin mRNA and C-peptide in the brain and the release of insulin by in vitro primary neuronal cell cultures (9). Insulin expression decreased in Tmem108 mutant mice.

Furthermore, central insulin involves about 50% of glucose metabolism in the brain. It is reasonable that insulin signaling dysfunction should be responsible for the cognitive deficits because the dysfunction occurs before antipsychotic administration. It is distinct that glucose uptake partly depends on central insulin in some brain regions (57). Insulin resistance is related to impaired glucose utilization (65). CNS insulin resistance may emerge with alteration in metabolic health. Hence, increasing insulin availability can plausibly improve cognitivism and metabolism (66).

Food and oxygen consumption in Tmem108 mutant mice were lower than the control mice, but urine production was higher, consistent with the elevated RER in Tmem108 mutant mice. Insulin performance in the CNS is the primary regulator of cognitive processes and energy balance (9). Insulin resistance is related to change in homeostatic signals impacting glucose metabolism and food intake. Evidence has underlined a role for antipsychotic medication in regulating central insulin signaling pathways (9). Cerebral insulin performs well in the hypothalamus and modulates glucose homeostasis, food intake, and cognitive processes (67–69). SCZ is conceptualized as an illness of dopaminergic dysfunction, links to abnormal metabolism (particularly type II DM), and may share plausible overlaps with CNS insulin dysregulation because of premorbid cognitive deficits (9). The mechanisms for interpreting these effects are largely unknown (9). Central insulin signaling represents a fundamental relation between cognitive and metabolic diseases (9).

Tmem108 mutant mice manifested weak fatigue resistance in the muscle fatigue task test. Before the test, the locomotor activity and the balance ability were assessed by the open field test and rotarod test separately. The results suggested that Tmem108 mutant mice have no difference from the control mice in the two types of tests. However, ambulatory counts representing locomotor activity were lower in Tmem108 mutant mice than in the control mice in the metabolism cages. It looks weird that locomotor activity was not inconsistent in different behavior tests, resulting in the difference of accustomed time before recording the data. The mice were accustomed for a few minutes in the open field test, so movements involved novelty exploration. However, the mice were cultured in the metabolism cages for 3 days, including 2 days accustomed, and the collecting data reflected living status like the home-cages.

It is known that defective PPI is seen in multiple neuropsychiatric disorders, including SCZ, where reduced gating is believed to be a possible neurobiological mechanism undertaking some essential abnormal cognitive functions associated with this disorder (29). PPI is a classic and plausible measurement in SCZ research (26). In this research, although metformin induced a decline of plasma glucose levels in Tmem108 mutant mice significantly, metformin addition slightly improved sensorimotor gating of Tmem108 mutant mice.

Besides PPI in SCZ, cognitive deficits/impairment are also a hallmark of SCZ (70). Cognitive dysfunction risk factors may contribute to cognitive impairment in SCZ (70). SCZ with comorbid diabetes was more cognitively impaired than SCZ without diabetes or diabetes only, especially in working memory and attention (70–74). Diabetes-related factors including insulin resistance, hyperglycemia, and lipid metabolic disorders may affect cognitive function (54, 75, 76). The levels of insulin resistance are elevated and correlated with the severity of cognitive impairment in first-episode SCZ patients (77–79).

Metformin is an effective treatment of prediabetes, and herein, it can attenuate the metabolic abnormality induced by antipsychotic medication without abridging the therapeutic effects (52, 53, 80). Metformin can reverse MK-801 induced schizophrenia-like symptoms, such as deficits, anxiety-like symptoms, hyperactivity, and spatial memory impairment (81).In SCZ treatment, metformin addition can improve physiological parameters and behaviors. SCZ frequently accompanies depression, with a rate of 20%-60% depending on the pathology phase (52, 82–84). Metformin can enhance antipsychotic, antidepressant, and anxiolytic activities (52, 62, 85).