Chronic Stress Exacerbates Hyperglycemia-Induced Affective Symptoms in Male Mice

Diabetes mellitus (DM) is a group of metabolic diseases characterized by chronic hyperglycemia resulting from either an inability to produce insulin or a dysregulated response to insulin [1]. Approximately 537 million adults aged 20–79 years are living with diabetes, and the number of diabetic cases worldwide continues to increase [2]. While this disease is associated with long-term complications including neuropathy, retinopathy, and increased risk of arteriosclerotic cardiovascular, peripheral arterial, and cerebrovascular disease, the high degree of comorbidity between diabetes and affective disorders, particularly depression and anxiety [3, 4], is less recognized and remains undertreated [5, 6].

Depression occurs twice as frequently among individuals with diabetes than among the general population [7], with an estimated prevalence rate of 17.6% among individuals with diabetes compared to 8.3% in the general population [8, 9]. Further, patients with comorbid major depression and diabetes appear to have higher rates of melancholic depression and suicide risk [10]. These differences may relate to increased inflammation, as measured by circulating C-reactive protein (CRP), observed among patients with DM [11, 12]. Inflammation-associated depression is often characterized by somatic symptoms (e.g., fatigue, sleep disruption) [13]; these symptoms may be more difficult for both patients and healthcare providers to recognize, leading to underdiagnosis and undertreatment [7, 14‒17].

There are a number of factors that contribute to diabetic patients’ increased risk of affective dysregulation. The psychological burden of managing a chronic disease likely increases the risk of depression and anxiety in diabetic patients [17, 18]. Additionally, chronically ill patients with affective disorders are more likely to engage in unfavorable lifestyle behaviors [7, 15, 17, 19‒21], further compounding existing issues. Hyperglycemia itself may also induce neurological changes that drive depression [22, 23].

This relationship has been experimentally confirmed using murine models. Pharmacologically inducing hyperglycemia in rodents using streptozotocin (STZ) has been shown to be associated with the development of depressive-like behavior [24‒26]. STZ ablates insulin-producing pancreatic β cells inducing chronic, robust hyperglycemia [27]. The failure to produce insulin in this model is like that of type I DM. However, it deviates from the clinical condition in that STZ-treated mice remain in good health for months after induction without the need for insulin treatment. It may better be viewed as a model that allows us to isolate the impact of chronic hyperglycemia on the brain in the absence of insulin treatment and obesity. Previously, chronic hyperglycemia has been associated with impaired hippocampal neurogenesis, systemic peripheral inflammation, neuroinflammation, and both central and peripheral oxidative stress [7, 26, 28‒30]. While the affective dysregulation associated with this model is likely linked to these mechanisms, further mechanistic research is needed.

Stress is, arguably, the most common factor associated with depression, and both clinical and preclinical research have demonstrated robust causal links [31‒37]. While our understanding of stress-induced depressive symptoms is still developing, reduced synaptic plasticity and inhibited neurogenesis have been strongly implicated [33, 35, 38‒40]. Understanding the shared and unique behavioral and neurobiological features of hyperglycemia- and stress-induced depression may allow us to develop more effective therapeutic interventions. If chronic stress and chronic hyperglycemia induce depression through different mechanisms this may allow them to act additively, further enhancing the risk for individuals with diabetes. Within this study, we used a murine model to compare the affective behavioral changes induced by (1) stress and (2) hyperglycemia. We used 2 weeks of unpredictable chronic mild stress (UCMS) [41] as our stress model and STZ to induce chronic hyperglycemia. We also evaluated commonly reported inflammatory targets (e.g., Il1b, Il6, Tnf), synaptic health-related targets (e.g., Bdnf, Snap25), and a critical glucose transporter (i.e., Slc2a1) within the hippocampus and frontal cortex, as these brain areas have been linked to depression [28, 38, 42, 43]. Both UCMS and STZ have been shown to independently induce depressive- and anxiety-like behavior in murine models [24, 28, 34, 44‒50], however, there is limited research directly comparing these models.

This study was conducted in male C57BL/6J mice from Jackson Laboratories (Bar Harbor, ME, USA). Mice arrived at 6 weeks of age and were allowed to acclimate to the Baylor University vivarium for 2 weeks prior to the start of experiments. They were maintained at 22°C on a 12-h light/dark cycle with food and water ad libitum, were housed with 2–3 mice per cage, and were handled for 5 days prior to injections. All testing took place during the light cycle phase (0700–1900 h). All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee of Baylor University (IACUC protocol number: 1550457) and were in compliance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals.

We used a 2 (STZ vs. buffer) × 2 (UCMS vs. control) factorial design with 12–13 mice per group (N = 49). The week of STZ or buffer administration was designated week 0. The UCMS protocol was conducted during weeks 4 and 5. Behavioral testing was conducted during week 6, and the mice were terminated at the start of week 7 (Fig. 1a). This study was run in two cohorts, with the same schedule except for the addition of the splash test (ST) in the second cohort. Tissue analyses were conducted on mice from the first cohort. Behavioral testing began the day after the final UCMS session with marble burying task on day 1, light/dark aversion (LDA) on day 2, open field task (OFT) and forced swim test (FST) on day 3, and ST (cohort 2 only) on day 4.

Hyperglycemia was induced through administration of STZ [27]. Starting at 8 weeks of age, mice were injected intraperitoneally with 50 mg/kg/day of STZ in a volume of 10 μL/g body weight citrate buffer for 5 consecutive days. Control mice received an equal volume of buffer. Body weight and blood glucose (post 4 h of fasting) were monitored at least biweekly to ensure the health of the mice as well as confirm hyperglycemic status. Hyperglycemia was defined as a blood glucose level over 250 mg/dL. Blood glucose was measured using an AUVON handheld glucometer and test strips following a nick to the tail vein.

Mice in the UCMS treatment groups were exposed to a quasi-randomized schedule of twice-daily stressors for 14 consecutive days [41]. The stressors used are described in Table 1. The predator odor stressor protocol was run as previously described [51].

The marble burying task has been used to assess anxiety-like and/or compulsive behavior, although the task responds to various pharmacological interventions including antidepressants [52‒55]. Marble burying has previously been shown to be reduced in STZ-treated mice [56, 57], and while the precise dimension it is evaluating is unclear, it provides a robust behavioral endpoint for the STZ model. Each mouse was placed in an arena (30 cm × 19 cm × 13 cm) with fresh bedding and 20 marbles (arranged in 5 rows of 4 marbles) for 30 min. The number of marbles buried at least two-thirds of the way under the bedding was recorded by an experimenter blind to experimental conditions [58].

The LDA task has been used to evaluate anxiety-like behavior [59‒61]. This task uses a two-chamber arena (each chamber 28 cm × 14 cm × 25 cm) with one open chamber illuminated by white lights and a dark, enclosed chamber. Mice were placed into the dark chamber, and the door separating the two compartments was removed. The mice were allowed to move freely between the chambers for 10 min. Each trial was recorded, and the latency to first exit the dark chamber was recorded by an experimenter blind to experimental conditions, while the number of transitions was determined by Noldus EthoVision XT software (Leesburg, VA, USA).

The OFT can be used to assess activity in a novel environment, while center time in the field is used to assess anxiety-like behavior [62]. Mice were placed in a novel plexiglass arena (30 cm × 30 cm × 40 cm) for 5 min and activity was recorded. Distance traveled and center duration were determined using Noldus EthoVision XT software (Leesburg, VA, USA).

The ST has been used to assess motivation toward grooming and has been demonstrated to be sensitive to stress and responsive to antidepressants [63, 64]. Mice were placed in an arena (30 cm × 30 cm × 40 cm), and the dorsal surface, from tail to nose of the mouse, was sprayed with a 10% sucrose solution. Grooming behavior was recorded for 5 min. An experimenter blind to experimental conditions analyzed the time to first groom and the total amount of time the mouse spent grooming.

The FST was developed as a screen for antidepressant activity of drugs in rodents [65]. It was run as previously described [66]. Each mouse was placed in a clear cylindrical swim chamber (diameter 19 cm) filled with ∼3.5 L of 24 ± 1°C water for 6 min. Each trial was recorded, and duration of immobility during the last 5 min of the trial was analyzed by an experimenter blind to experimental conditions.

Following completion of behavioral testing, the mice were euthanized by CO₂. Blood was collected via cardiac puncture, and mice were transcardially perfused. Tissue was collected, immediately snap frozen in liquid nitrogen, and stored at −80°C until processing.

RNA was extracted from the hippocampus and frontal cortex of the brain using the E.Z.N.A. RNA Isolation Kit II (Omega BioTek, Norcross, GA, USA). cDNA was transcribed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems by Life Technologies, Grand Island, NY, USA), and quantitative real-time polymerase chain reaction (qPCR) was carried out in a QuantStudio 6 PCR machine using TaqMan gene expression assays (Applied Biosystems by Life Technologies, Grand Island, NY, USA). Reactions were performed in duplicate with the fold difference for each gene calculated using the 2^−ΔΔCT method. Targets analyzed are as follows (Table 2).

Two-way ANOVAs were conducted for all behavioral tasks and qPCR analyses. Full ANOVA results are presented in supplemental Tables 1 and 2 (for all online suppl. material, see https://doi.org/10.1159/000534669) (online suppl. Table S1, S2). Effect size was calculated using partial eta squared for significant results. A mixed-effects model with two group factors (STZ and UCMS), time as fixed factor, and individual subjects as a random effect was conducted for body weight due to incomplete data in cohort 2 for the 6-week time point. Two-way repeated measures ANOVA was run for blood glucose. Tukey post hoc tests were performed to compare significant group differences as needed. Correlational analyses were conducted across the entire population to explore the relationships between outcome variables. These results are presented in online supplementary Tables S3 and S4. To gain perspective on overall affective dysregulation, we conducted an exploratory analysis on the computed composite symptom score. Such methodology has been utilized in a variety of similar contexts and has been shown to yield robust measures across complementary tasks [67‒69]. Behavioral assessments were reverse-scored as needed, such that higher scores represent higher symptoms. The composite Z score was calculated and planned comparisons were conducted to compare the four groups. Statistical analyses were conducted using Prism 9 GraphPad and IBM SPSS Statistics 26. Effects were considered significant at p < 0.05. Data are displayed as mean ± standard error of the mean (SEM).

Fasting blood glucose was measured throughout the experiment (Fig. 1b). There was a significant time × STZ interaction (F(4, 180) = 117.4, p < 0.0001, η²_p = 0.7229) such that STZ-treated mice rapidly developed and maintained elevated blood glucose levels, as well as a significant time × UCMS interaction (F(4, 180) = 4.244, p = 0.0026, η²_p = 0.0862). This appeared to be driven by stress-induced elevations in blood glucose at week 5. While the STZ × UCMS × time interaction failed to reach significance (p = 0.0608), this stress-induced elevation appears more pronounced in STZ mice, indicating that the STZ mice may be more vulnerable to stress-induced perturbations in blood glucose levels.

Body weight of mice was also monitored throughout to ensure the health of all animals (Fig. 1c). As anticipated, there was a significant time × STZ interaction (F(5, 201) = 57.38, p < 0.0001), such that STZ mice failed to gain weight during the study. A significant negative correlation was observed between body weight and blood glucose level (r² = 0.49, p < 0.0001, see Table S3), such that mice with higher blood glucose had lower body weight. Further, there was a significant time × UCMS interaction (F(5, 201) = 25.46, p < 0.0001), such that we observed a dip in body weight corresponding to the stress exposure weeks.

Latency to first exit the dark arena was analyzed for the LDA task (Fig. 2a). There was a main effect of UCMS on latency to exit, such that stress mice exited the dark arena more rapidly (F(1, 45) = 5.737, p = 0.0208, η²_p = 0.1131). There was no main effect or interaction with STZ. We also looked at the total number of transitions between the light and dark arenas (Fig. 2b). As with the time to exit analysis, we observed a main effect of UCMS, such that UCMS significantly increased the number of transitions (F(1, 45) = 7.714, p = 0.0080, η²_p = 0.1463). No main effect or interactions were observed for STZ.

The distance traveled and total duration a mouse remained in the center of a novel empty open field were analyzed (Fig. 2c, d). Mice treated with STZ showed a reduction in total distance traveled (F(1, 45) = 9.584, p = 0.0034, η²_p = 0.1756). UCMS did not significantly impact distance traveled. However, we did observe a main effect of UCMS on center duration, such that stressed mice spent significantly less time in the center (F(1, 45) = 10.92, p = 0.0019, η²_p = 0.1952). There were no main effects or interactions for STZ on center duration.

Within the ST, we quantified the latency to first groom and total grooming time (Fig. 3a, b). STZ increased the latency to start grooming (F(1, 20) = 5.562, p = 0.0286, η²_p = 0.2176) and decreased the total time grooming (F(1, 20) = 4.928, p = 0.0382, η²_p = 0.1977). Further, we observed an STZ × UCMS interaction for total groom time (F(1, 20) = 8.738, p = 0.0078, η²_p = 0.3041). Post hoc analyses indicate that this effect was driven by a reduction in the non-stressed mice, without an effect in the UCMS mice (p = 0.0078).

Immobility duration was recorded during the last 5 min of the FST (Fig. 3c). STZ significantly increased immobility time (F(1, 45) = 31.39, p < 0.0001, η²_p = 0.4109), and the effect of UCMS on immobility time approached significance (F(1, 45) = 3.930, p = 0.0535, η²_p = 0.0803).

The number of marbles buried in the marble-burying task was recorded (Fig. 3d). It was demonstrated that both STZ (F(1, 45) = 24.51, p < 0.0001, η²_p = 0.3526) and UCMS (F(1, 45) = 12.06, p = 0.0012, η²_p = 0.2113) significantly decreased the number of marbles buried.

The composite behavioral score was calculated by averaging Z scores across the affective assessments (Table 3). Planned comparisons demonstrated that STZ-control mice showed a worse symptom profile than buffer-control mice and that STZ-UCMS mice had a higher composite symptom score than all other groups (ps < 0.001).

To determine if STZ and UCMS differentially impact neuroinflammation, we ran qPCR on tissue collected from the hippocampus and frontal cortex (Fig. 4). Within the hippocampus, we observed a significant STZ-induced increase in Tnf (F(1, 19) = 6.009, p = 0.0241, η²_p = 0.2403) and Bdnf (F(1, 19) = 5.784, p = 0.0265, η²_p = 0.2334). Further, these two markers were correlated (r² = 0.286, p = 0.008, see Table S2). No effects were observed for any of the other markers, and UCMS did not significantly impact the markers we evaluated in the hippocampus.

Within the frontal cortex, we observed an effect of both STZ and UCMS on Il1b expression. STZ increased Il1b expression (F(1, 21) = 5.647, p = 0.0271, η²_p = 0.2119), while UCMS significantly decreased its expression (F(1, 21) = 4.971, p = 0.0368, η²_p = 0.1914). The increase in Il1b expression in the cortex was not significantly correlated to changes in hippocampal Tnf or Bdnf (Table S4).

Within this study, we aimed to compare the affective and neuroinflammatory profiles of UCMS and chronic hyperglycemia. STZ-treated mice displayed the anticipated increase in fasting glucose levels and reduced body weight. We also observed increased neurovegetative depressive-like behavioral symptoms and elevations in markers of neuroinflammation. In response to UCMS, mice displayed transient reductions in body weight as well as alterations in anxiety-like behavior, a trend toward increased immobility in the FST, and decreased Il1b expression in the frontal cortex. Evidence of increased overall affective dysregulation from the addition of UCMS to STZ was also shown through exploratory analyses. These data supports our hypothesis that stress and hyperglycemia have distinct behavioral profiles and indicates that these profiles are likely mediated by distinct neurobiological profiles.

STZ induced robust changes in immobility time in the FST and groom time in the ST. Both are classic tests of antidepressant sensitivity. They have also been associated with lack of motivation or increased apathy [63]. These particular depressive-like behaviors have been reported and replicated following STZ administration by a variety of research groups [24, 25, 28, 43, 46, 48, 70, 71]. STZ also significantly reduced distance traveled in the OFT. This effect has also been reported previously [24, 50, 72]. There are a variety of reasons that a mouse may have less activity in a novel environment; it could be indicative of a depressive-like lack of exploratory interest [73, 74] or it could also point to increased fatigue [75]. Both depression and fatigue are reported at higher rates in diabetic individuals than in the general population [76‒79]. Further research is needed to parse out this effect, as fatigue can represent a neurovegetative symptom of depression as well as exist independently of mood symptoms. In the context of the STZ model, it is possible that reduced energy availability due to poor glucose uptake may contribute to behavioral changes. While there are a number of studies that report reduced open field center time [46, 50, 70, 72], we were not able to detect any significant anxiety-like behaviors in our STZ mice. We also observed decreases in marble burying in response to STZ.

The UCMS mice showed an anxiety-like phenotype with significantly decreased center duration in the OFT. This result has been observed in other studies of UCMS [80‒83]. In the LDA task, the UCMS groups displayed a significantly shorter latency to exit the dark arena as well as a significantly higher number of transitions between arenas. A longer latency to exit the dark area is typically associated with anxiety-like behavior [61], and therefore this brings into question why the UCMS groups showed robust anxiety-like behavior in OFT but not in LDA. The protocol we used initially places the mouse in the dark arena; another study using this protocol showed an STZ-induced increased latency to exit the dark arena [50]. However, a number of groups place mice in the illuminated arena first and measure the latency to enter the dark arena [25, 46, 84]. One such study found that chronic stress significantly decreased latency to enter the dark area [47]. A potential explanation for our unexpected result may be a consequence of the stress procedure. Mice may have developed a learned aversion to novel environments due to repeated exposure to novel stressors. If the UCMS mice were expecting a stressor when placed in the dark box, they may have exited quickly as an avoidance response. The lack of correlation between the LDA task and other affective measures also suggests that we were not measuring the affective domain we were targeting (see Table S3). Further research is needed to elucidate the impact of UCMS on the LDA task as well as our failure to detect anxiety in the STZ mice. We anticipate that the behavioral profiles of the STZ and UCMS groups would be similar if they shared neurobiological mechanisms; however, their distinct profiles suggest that chronic hyperglycemia and stress modulate affective dysregulation through distinct mechanisms.

While most assessments did not show additive effects between STZ and UCMS, additivity was observed for marble burying. In this task, the STZ-UCMS group buried significantly less marbles compared to all other groups. While there is lack of clarity in the literature concerning the domain this task evaluates, greater number of marbles buried is generally interpreted as a compulsive or anxiety-like phenotype. Our laboratory has repeatedly found that STZ induces a robust decrease in marbles buried. This reduced marble burying behavior in response to STZ has been reported previously [56, 57]. In our model, this task may be indicative of reduced motivation or fatigue-like behavior, as evidenced by its correlation with FST performance. While a stepwise pattern was observed in FST, such that STZ-UCMS mice showed the highest immobility times, the effect of UCMS on immobility was only trending toward significance. The already high immobility time in this task may produce a bit of a ceiling effect, making it difficult to observe a mild stress-induced exacerbation in the depressive phenotype. When considering all tasks analyzed, the STZ-UCMS group had the greatest number of significant dysregulated behavioral responses, and our exploratory analysis indicated more severe affective dysregulation.

With the hippocampus and frontal cortex, we evaluated a variety of neurobiological markers, examining neuroinflammation, neuroplasticity, and glucose transport. The only markers we saw dysregulated in our hippocampal samples were Tnf and Bdnf. The hippocampal mRNA expression for both markers was significantly elevated in STZ-treated mice. Current literature confirms that STZ induces a significant upregulation of hippocampal expression of Tnf at times of affective dysregulation [28, 48, 85]. While STZ has previously been reported to decrease hippocampal expression of Bdnf [28, 48, 71], indicative of a decrease in synaptic plasticity, our study found elevated Bdnf expression. It is possibly a consequence of timing, potentially as a late compensatory response caused by early-on downregulation of Bdnf. Further research is necessary to elucidate the mechanism of this upregulation. In the frontal cortex, STZ induced an elevated expression of Il1b, which is confirmed by other STZ studies [28, 86].

While UCMS did not significantly dysregulate neuroinflammatory response in the hippocampus, a downregulation in Il1b expression was observed in the frontal cortex. This was not the direction we had anticipated, as various studies have demonstrated elevations in Il1b mRNA expression following stress protocols [33, 49, 83, 87, 88]. However, other studies that have shown that UCMS can produce a depressive-like phenotype without a neuroinflammatory profile [32, 89], implicating HPA-axis dysregulation and neurotrophic changes as potential neurobiological mediators instead [32, 90]. The immunomodulatory effects of stress are complex and often variable; for a review of this topic, see [91]. Although we used a 2-week stressor protocol, there is significant variability in protocol length – ranging from 2 to 10 weeks [32, 40, 81, 92, 93]. Studies with longer stressor protocol lengths have more consistently reported increases in neuroinflammatory markers (e.g., Il1b, Il6, Tnf). The lack of neuroinflammatory consequences observed in our protocol may relate to our shorter stress protocol duration. Alternatively, the length of time between the end of the stress protocol and tissue collection may have impacted our results. While we expected the stress protocol to induce persistent neuroinflammatory effects, it is possible that having just over a week between the cessation of stress and tissue collection resulted in us missing an inflammatory signal and instead picking up a compensatory, recovery response. Further research is necessary to parse out the timeline of the pro-inflammatory profile in response to chronic stress and how it pairs to the recovery from behavioral effects.

Overall, the behavioral and neuroinflammatory results suggest that UCMS and STZ act via distinct mechanisms and, thereby, promoting their ability to have additive effects. While our data did not show increased sensitivity to stress-induced depressive symptoms in hyperglycemic mice, the known struggles associated with disease management in diabetic individuals may contribute to their high incidence of depression.

Our study did have two important limitations that should be noted. First, we only utilized male mice. Male mice have historically been used to study effects of STZ, in large part due to the difficulty of inducing hyperglycemia in female mice [94‒99]. Given the higher prevalence of depressive-like behavior in female clinical populations, it is critically important to study female models of hyperglycemia-associated depression. Current research using a higher STZ dose to induce hyperglycemia in female mice is underway within our laboratory.

Another limitation, mentioned earlier, concerns the UCMS protocol. While we used a 2-week stressor, we recognize that longer protocols are often used. We utilized a shorter stress protocol as we wanted to determine if there would be potential synergy in symptom severity between STZ and UCMS. However, it is possible that a longer protocol known to induce more robust depressive-like and neuroinflammatory responses [37, 49, 82, 83] may have had a different effect. Additionally, there is significant variability in the set of mild stressors utilized between research groups. We also chose to end the stress protocol prior to starting behavioral data collection and approximately 1 week prior to tissue collection. This selection was to avoid the highly variable acute effects of the stressor on behavioral and inflammatory outcomes. As such, we do not know what the behavioral effects and neuroinflammatory signature of the brain were during and immediately following the stress protocol.

In summary, we used a murine model to investigate the effects of chronic stress and hyperglycemia on neuroinflammatory and affective behavioral profiles. While chronic stress and hyperglycemia individually produced neurovegetative and affective symptoms, their combination exacerbated the overall symptom burden. Further, they were associated with distinct neuroinflammatory profiles. This is indicative of distinct neurobiological mechanisms of depression. Since chronic illness is often accompanied by chronic stress, it is vital to investigate how chronic stress interacts with mechanisms such as hyperglycemia. Elucidating these mechanisms will provide insight into more efficient and targeted treatment for diabetic patients suffering from depression/anxiety.

All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee of Baylor University (IACUC protocol number: 1550457) and were in compliance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals.

The authors have no conflicts of interest to declare.

This work was funded by Baylor University.

Riley G. McCready and Kayla R. Gilley contributed to designing and performing the experiment, analyzing the data, and writing the manuscript. Grace M. Hall and Laura E. Kusumo contributed to performing the experiment and analyzing the data. Elisabeth G. Vichaya guided the planning of this project and the writing of this article. All authors have read and approved the final manuscript.

Riley G. McCready and Kayla R. Gilley contributed equally to this work.

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author. Raw data will be made available via the Baylor Electronically Accessible Research Data (BEARdata) repository.