From Cell to Circuit: Investigating Functional Topological Changes in iPSC-derived Neuronal Networks in Major Depressive Disorder

doi:10.21203/rs.3.rs-4569534/v1

Download PDF

Article

From Cell to Circuit: Investigating Functional Topological Changes in iPSC-derived Neuronal Networks in Major Depressive Disorder

https://doi.org/10.21203/rs.3.rs-4569534/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The functional organization of brain networks maintains a delicate equilibrium between segregation and integration where it facilitates local neural communication together with effective global integration of information across network’s components. While numerous whole-brain imaging studies have linked alterations in functional topology to major depressive disorder (MDD), our comprehension of how these changes manifest at the cellular level remains limited. Here, we explored whether neuronal networks derived from induced pluripotent stem cells (hiPSCs) of nine depressed patients display a distinct functional topology compared to those of matched controls. Spontaneous activity of the derived neuronal networks was captured using calcium imaging, and graph theory analysis was applied to assess functional topology. We computed the graph metrics clustering coefficient and global efficiency to quantify respective network segregation and integration attributes. We also measured the average node degree to assess group differences in the overall number of connections. We observed a decrease in clustering coefficient and average node degree in MDD-derived neural networks compared to those of controls. Global efficiency also exhibited a decreasing trend in patient-derived networks across varying thresholds and network sizes. Together, our findings reveal diminished segregation properties and a reduced number of nodal connections in MDD-derived neural networks, suggesting a predisposition for a less efficient functional topology in depression already at the microscale. This work marks the first attempt to explore microscale alterations in functional topology of human-derived neural networks in MDD and highlights the power of iPSC technology in providing a human cellular model to better understand disease mechanisms.

Biological sciences/Neuroscience

Biological sciences/Stem cells

The brain is a highly complex network with a distinctive functional topology that enables efficient integration of information across its distant parts, while maintaining a level of segregation that allows for effective local neural communication and functional specialization [1, 2] (Fig. 1A). The balance between integration and segregation processes is critical for brain function, as it enables the brain to meet highly dynamic cognitive demands and generate complex, adaptive behavior [3]. Recently, altered network topology of human brain networks has been used to account for mental disorders [4–7], including major depressive disorder (MDD) [8]. MDD is a psychopathology that affects around 280 million people worldwide and is characterized by persistent low mood and disturbed cognitive function [9]. Despite the ubiquity of depression, its exact underlying pathophysiology remains elusive. Most of the evidence for altered functional network topology in depression originates from macroscale investigations, i.e., whole-brain imaging studies using functional Magnetic Resonance Imaging (fMRI) [8, 10–13]. However, the functional organization of neural networks on the microscale, at the cellular level, has remained largely unexplored in the context of depression.

Large-scale brain networks [14, 15] and in vitro neuronal circuits of cultured neurons [16–18] tend to self-organize into a so-called “small-world” architecture that is conserved across both anatomical structure [19, 20] and function [21, 22]. The term “Small-worldness”, introduced by Watts & Strogatz [23], describes a topological organization exhibiting two key network attributes: segregation and integration (Fig. 1B). The topology of a small-world network entails densely connected and functionally segregated clusters or “communities” forming among neighboring elements. Simultaneously, long-ranging connections are sparsely distributed, linking these communities across the entire network and allowing for a global integration of information. This small-world organization is favorable as it promotes efficient information transfer at a minimum cost of energy and wiring resources, allowing the system to accommodate the necessary balance between segregation and integration [23, 24].

Such a complex network topology can be investigated within the framework of graph theory, in which neuronal networks are mathematically modeled as graphs composed of nodes and edges. In this context, nodes represent the computational units of the network (such as single neurons), while edges represent anatomical connections or statistical dependencies between the nodes, known as structural or functional connectivity, respectively [21]. The two fundamental features of an efficient small-world network, segregation and integration, can be assessed by graph metrics such as clustering coefficient and global efficiency, respectively. The clustering coefficient captures the extent of segregation in the network by measuring the tendency of adjacent and functionally similar nodes to form dense clusters [25]. On the other hand, global efficiency quantifies the efficacy of parallel information transfer across the entirety of the network, reflecting its level of integration [26]. Graph theory metrics offer a simplified way to model complex systems, providing valuable insights into global network topology. While several neuroimaging investigations have reported changes in graph metrics in large-scale networks in MDD [8], graph theoretical enquiries into microscale network topology in depression are currently lacking.

The ability to derive human induced pluripotent stem cells (hiPSCs) has opened up new avenues in disease modeling and drug discovery at the cellular level [27]. This technology provides disease-relevant, patient-specific neuronal cultures accessible for investigation [28, 29]. The technique involves reprogramming human somatic cells into pluripotent stem cells, which can then be differentiated to various neuronal fates while preserving the genetic makeup of the donor cells. Initially, iPSC technology was employed for modeling diseases with a strong monogenic component. However, it has since emerged as a powerful tool for studying complex, polygenic psychiatric disorders, such as MDD, where genetic factors account for 30–40% of the disorder’s etiology [30, 31]. In the context of major depression, hiPSCs have been used to generate serotonergic neurons used to model the resistance to the antidepressants selective serotonin reuptake inhibitors (SSRIs) [32, 33] and to explore ketamine-dependent neuroplasticity in dopaminergic neurons [34–36]. iPSC-models have also been deployed to explore the potential role of bioenergetic imbalance in the pathophysiology of MDD [37, 38]. Until now, hiPSC-models have not been utilized to examine changes in the functional architecture of in vitro neural networks and their potential implication in the etiology of MDD.

The current study marks the first attempt to bridge this knowledge gap by investigating whether microscale networks derived from hiPSCs of depressed patients show differences in functional organizational properties compared to network derived from matched controls. With this approach we hope to provide a deeper insight into the etiology and pathophysiological mechanisms of MDD. We measured the spontaneous activity of the iPSC-derived neurons by means of live-cell calcium (Ca²⁺) imaging, and we constructed the microscale graphs using the functional connectivity profile of each culture. We then probed the segregation and integration properties of the graphs by computing graph theory metrics (clustering coefficient, global efficiency, and average node degree) and compared them between the study cohorts. Our findings suggest an altered functional organization of microscale neuronal networks in depression. Furthermore, these alterations are detectable even at the small-scale topology of in vitro iPSCs-derived neurons.

Cultivating primary human fibroblasts and generating hiPSCs

Skin biopsies were obtained from 9 MDD patients and 9 age- and sex-matched healthy controls by the Department of Dermatology, University Hospital of Regensburg, Regensburg, Germany. The skin biopsies were acquired from MDD patients at the end of their hospitalization. The diagnosis of MDD was determined based on the ICD-10 criteria [39] and depression severity was assessed using the 21-item Hamilton Rating Scale for Depression (HAMD₂₁) [40]. All Participants provided their written informed consent, and the study was approved by the ethics committee of the University of Regensburg (ref: 13-101-0271). Human fibroblasts were cultivated, frozen until later use, and thawed as reported in [41]. Human iPSCs were generated with episomal plasmid vectors [42] as described in [37]. The quality of the generated iPSC in terms of their pluripotent nature was assessed and tested using the PluriTest® assay adjusted for next-generation sequencing (NGS) data [43].

Generating hiPSC-derived neural progenitor cells (NPCs) and differentiating neurons

We differentiated the generated hiPSCs into NPCs following the derivation protocol of Yan et al. [44]. We cultured the derived NPCs in a neural expansion medium (NEM) (49.5% Neurobasal medium (Gibco by Life Technologies), 49.5% Advanced DMEM/F12 media, and 1% neural induction supplement (Life Technologies)). NEM medium was changed every couple of days until 80–90% confluency was reached. Next, we dissociated NPCs using Accutase (StemCell Technologies) and cultured them in Matrigel-coated dishes (Corning). To start the differentiation of NPCs into neurons, we plated NPCs of passages 5 to 10 onto poly-L-ornithine/laminin-coated 35 mm µ-dishes (Ibidi, Gräfelfing, Germany) and cultured them in a neural induction medium (1% B-27 Plus supplement (Life technologies), 0.5% GlutaMax (Life Technologies), 0.5% non-essential amino acids, 0.5% Culture One supplement (Thermo Fisher Scientific), 200 nM ascorbic acid (Carl Roth), 20 ng/mL BDNF (PeproTech), 20 ng/mL GDNF (PeproTech), 1 mM dibutyryl-cAMP (StemCell), 4 µg/mL laminin (Sigma-Aldrich), 50 U/mL Penicillin (Thermo Fisher Scientific), and 50 µg/mL Streptomycin (Thermo Fisher Scientific) in Neurobasal Plus medium (Thermo Fisher Scientific)). Half of the neural induction medium was replenished every other day in the cultures of induced neurons (iNeurons) with the addition of fresh medium. We treated the neuronal cultures with Cytarabine (ara-C) at 7 days in vitro (DIV) to suppress the growth of non-neuronal cells. Cultures of mature neurons were obtained at 21 DIV and were ready for imaging (Fig. 2). For each subject’s iPS cell line, three iNeuron cultures stemming from different NPCs passages were prepared and imaged, serving as biological replicates to yield stable parameter estimates.

Immunofluorescence

For immunocytochemistry, IPS neurons grown on laminin coated µ-dishes (Ibidi) were fixed in a 4% PFA. Cells were permeabilised and blocked using a blocking solution containing 1x PBS, 0.5% Triton X-100 and 10% normal goat serum. Cells were incubated overnight at 4^oC with the following primary antibodies: anti-β-III-tubulin (mouse; 1:2000, G7121, Promega,), anti-MAP2 (chicken; 1:5000, ab5392, Abcam), anti-VGLUT1 (rabbit; 1:500, ab180188, Abcam). After three washing steps, cells were incubated for 1 hour with corresponding secondary antibodies conjugated with Alexa Fluor 488, Cy3, Cy5 and Hoechst, to stain nuclei. Cells were mounted on glass slides with DAKO fluorescent mounting medium.

Microscopy/Ca²⁺ Imaging

To prepare for imaging, we loaded neuronal cell cultures with a solution of the ratiometric calcium indicator Fura-2/AM (2 µM, Gibco by Life Technologies) and Pluronic-F127 (Thermo Fisher Scientific) in Opti-MEM solution (Gibco by Life Technologies) and incubated the cultures at 37 ℃ for 30 minutes. Cells were then washed with glucose-containing Ringer’s solution. We then examined the cultures with a Zeiss Axio Observer.Z1 inverted fluorescence microscope equipped with a Fluar 20x/0.75 objective and an AxioCam MRm CCD camera (ZEISS, Jena, Germany). To measure intracellular calcium levels, we excited Fura2-loaded cell cultures with ultraviolet light at wavelengths of 340 and 380 nm (BP 340/30 HE, BP 387/15 HE) using a fast wavelength switching and excitation device (Lambda DG-4, Sutter Instrument), and we detected fluorescence at 510 nm (BP 510/90 HE and FT 409). We used Zen imaging software (ZEISS) to control the hardware and collect a 20-minute-long recording of spontaneous neural activity in each dish. The recordings were acquired at a rate of 1.89 Hz, yielding 2265 frames or images of size 272x208 pixels (438.736x335.504 µm). It is worth noting that in our sample, Ca²⁺ imaging recordings of 2 patients and their 2 matched controls were acquired for another project at a different frame rate (0.5 Hz), yielding a different frame count (1200 frames). Although a reduced sampling rate can inflate pairwise correlations between time series, it is important to clarify that including these 4 subjects in the analysis does not bias our comparison between patients and controls. This is because the same number of two datasets from each group is equally affected by the reduced sampling rate, maintaining an unbiased group comparison. For every image within a recording, the Ca²⁺ signal was expressed as the 340/380 fluorescence intensity ratio (indicating the ratio of bound/free Ca²⁺). This ratio was computed using ImageJ plugin “Ratio_Plus” [45]. An example of the spontaneous activity of iPS-neurons captured by Ca²⁺ imaging is shown in the Supplementary Video 1.

Motion correction and segmentation

The preprocessing and segmentation of Ca²⁺ recordings were carried out using MATLAB R2020b (The MathWorks Inc, Natick, MA, USA) (Fig. 3). First, we corrected for motion artifacts by registering the images in each video through the application of Non-Rigid Motion Correction (NoRMCORRE) [46] (Fig. 3A). To eliminate distortions at the recording edges, we cropped the margins of all videos by 2 pixels on each side. Subsequently, we generated a correlation image for each motion corrected video by averaging, for each pixel, the temporal correlation of that pixel’s activity with the activity of its nearest 4 neighbors [47]. This correlation image, serving as a summary image, enhanced active neurons while suppressing uncorrelated neuropil noise, thereby facilitating the subsequent segmentation step [48]. We then established a segmentation pipeline in CellProfiler [49, 50] to create a binary mask image for each correlation image. These binary masks delineated the boundaries of all neurons in the correlation image, encompassing both active and non-active neurons. The masks produced by CellProfiler underwent the same cropping as the recordings to ensure size consistency. Within these masks cell boundaries were identified and cells were labeled. Subsequently, we employed a MATLAB-based intensity-thresholding workflow to segment regions of interest (ROIs) representing active neurons in each recording (Fig. 3B). Spatial smoothing of each video with an 8-pixel Gaussian kernel helped eliminate spatially distributed temporal noise, allowing for smoother time courses. The data was then normalized, and Z-scores were calculated across the temporal dimension. We determined the maximum Z-score for each frame and selected the minimum among these values as the threshold above which a pixel was considered to represent neuronal activity. Additionally, we set a cluster-size criterion for the detected suprathreshold neighboring pixels, considering only clusters of more than 14 pixels as neurons (corresponding to 22.6 µm, which falls within the typical range of soma size seen in mammalian neurons [51]). In a subsequent step, we masked the detected ROIs with the binary mask of all cells to mitigate false-positive identification of background elements as neurons. The time-course of each segmented ROI was computed by averaging the Ca²⁺ signal across all pixels of that ROI (Fig. 3C). We validated the segmented neurons through visual inspection of their time-courses along with their spatial masks overlaid on a summary image (mean image). Cells exhibiting non-active Ca²⁺ traces were excluded from the analysis, and time-courses from all active cells in each video were saved for the graph analysis.

Graph theory analysis

We performed graph theory analyses in Python using the NetworkX 2.2 package [52] (Fig. 3). To reduce temporal noise, we first band-pass filtered the preprocessed time-courses between 0.0001 and 0.07 Hz. We determined the connectivity profile of the neuronal network in each recording by computing pairwise Pearson correlations between the filtered time-courses, resulting in one functional connectivity (FC) matrix per video (Fig. 3D). We then thresholded FC matrices to obtain binarized adjacency matrices. The resultant adjacency matrices consist of binary values of ones and zeros, where “one” indicates the presence of a suprathreshold connection between two nodes, and “zero” indicates the absence of such a connection (Fig. 3E). Subsequently, we constructed binary undirected graphs from the resultant adjacency matrices by depicting the neurons in the matrix as nodes, and assigning an edge between any pair of nodes that showed a suprathreshold connection (Fig. 3F). To evaluate network attributes of segregation and integration, we computed the graph measures clustering coefficient and global efficiency. We also computed average node degree to get an estimate of the overall number of nodal connections within the graphs (refer to the supplementary information for mathematical definitions of the computed graph measures, and to [53] for a review on their interpretations).

The number of suprathreshold edges in the constructed graphs, and thus the resultant graph measure, heavily depends on the selection of the correlation threshold r_thresh. To circumvent bias related to selecting an arbitrary threshold, we repeated the graph analysis for every r_thresh in the range of ± 0.2–0.8 (in steps of 0.1). Previous studies have reported that graphs created with too low or too high thresholds exhibit characteristics that cannot be distinguished from random or lattice networks [26]. For this reason, we excluded two thresholds from our range: the extremely liberal r_thresh (± 0.1) that typically yields spurious correlations/edges, and the extremely strict r_thresh (± 0.9) that typically results in fragmented or sparsely connected graphs.

Clustering coefficient reflects a network’s tendency to segregate and form clusters. It is computed as the proportion of a given node’s neighbors that are also neighbors of each other. Thus, clustering coefficient represents how efficient the communication is in a subgraph of one node’s neighbors [53].

Global efficiency, on the other hand, refers to the efficiency of parallel communication between all nodes in a network. It is quantified as the average inverse of the shortest path length connecting any pair of nodes in that network [21, 53]. Consequently, global efficiency serves as a metric to assess the network’s capacity for integrating information across its nodes.

Average node degree describes the number of connections (edges) that nodes in a network have on average. It is one of the most easily derived measures, obtained by dividing the sum of node degree over the total number of nodes in a network [21].

One of the challenges we faced when analyzing the Ca²⁺ imaging data was the different number of cells in each video since recordings were acquired from different cultures, each with a unique cellular organization. To account for this variability in the number of cells, we computed graph measures for 500 unique randomly sampled sets of cells for set-sizes of 10 to 47 cells. These limits were chosen such that a) 500 unique combinations were possible and b) data from at least five subjects per group could be included (Supplementary Fig. 1).

Statistical testing

We performed a three-way repeated measures ANOVA (Group x Threshold x Network size (Number of cells)) on the computed graph measures to investigate group differences between patients and healthy controls. The analysis aimed to identify a main effect of Group (controls versus patients) and potential interactions between Group and other factors (Threshold, Network size). Upon detecting a group effect or an interaction with the Group factor, post-hoc t-tests were conducted to assess group differences at each Threshold and each Network size level. To that end, we utilized Monte Carlo permutation testing to create a null-distribution of t-values for each graph measure at each level of Threshold and Network size by shuffling group labels 5000 times and computing an independent t-test. The null distributions were then used to evaluate the observed t-values of group difference. After correcting for multiple comparisons using false discovery rate (FDR), the observed t-values did not show statistically significant differences between groups. However, for exploratory purposes, we present the findings at p_uncorrected < 0.05 in the Results section.

In this study, we examined differential functional network topology in cell cultures of iPSC-derived neurons of 9 MDD patients in comparison to their matched healthy controls. Using graph theory metrics, we explored whether the functional architecture, depicted by Ca²⁺ imaging of spontaneously active neuronal cultures, is disrupted in MDD, affecting the two most fundamental principles or network organization: segregation (measured by clustering coefficient) and integration (measured by global efficiency). Additionally, we computed average node degree to assess the overall number of nodal connections in the networks and how it differs between the experimental groups.

Demographic details and sample characteristic

We examined group differences in age, gender, body mass index (BMI), and HAMD scores. As the control group of healthy, non-depressed subjects was intentionally chosen to match MDD patients in terms of gender and age, there was no differences between the two groups neither in gender (4 males and 5 females in each group), nor in age (age_control= 33.22 (± 3.44), age_MDD= 32.33 (± 3.82), t(16) = 0.19, p = 0.85). Similarly, BMI did not differ between the groups (BMI_control= 22.67 (± 0.72), BMI_MDD= 24.58 (± 1.29), t(16) = 1.29, p = 0.22). However, the two groups showed significant differences in HAMD scores, as would be expected (HAMD_control= 0.4 (± 0.25) (no depression), HAMD_MDD= 26 (± 1.63) (moderate to severe depression), t(16) = 12.19, p < 0.001).

Neuronal marker expression in iPS-derived neurons

iPS-derived neurons expressed the typical neuronal markers Beta-III tubulin (B3TUB) and microtubule associated protein 2 (MAP2). Furthermore, iPS-neurons displayed immunopositivity for the specific synaptic proteins vesicular glutamate transporter 1 (VGLUT1) indicating the successful differentiation of NPCs into functional neurons with glutamate-mediated signaling (Supplementary Fig. 2). iPS-derived neurons from a parallel study, generated using the identical protocol, demonstrated functional electrophysiological properties [37].

Segregation properties

We assessed segregation attributes of cultured neuronal networks by computing the clustering coefficient providing an estimate of the proportion of a node’s neighbors that are also connected to each other. The metric indicates the network’s ability to form clusters, potentially facilitating the execution of specialized functions within these clusters (see the supplementary information for an in-depth description of employed graph measures). We observed a significant decrease in the clustering coefficient in patient-derived neuronal cultures compared to those of controls (main effect of Group, F(1,8) = 5.97, p = 0.04, Fig. 4). This reduction indicates a diminished segregation capacity in in-vitro networks of MDD patients compared to healthy controls. Moreover, the clustering coefficient exhibited an increase as the network size grew larger (main effect of Network size, F(37, 296) = 73.59, p < 0.001) with a decreased and thus more liberal r_thresh (main effect of Threshold, F(6, 48) = 1003.25, p < 0.001).

We performed post-hoc t-tests to identify the specific network sizes and r_thresh at which a significant decline in clustering coefficient occurred (Fig. 4). Results revealed that most group differences in Clustering coefficient were observed at larger network sizes (refer to the Supplementary Table 1 for post-hoc statistics). To demonstrate that group differences have not been driven by outliers, we refer the reader to the Supplementary Fig. 3.

Integration properties

We assessed global efficiency to measure the integration attributes of the networks and their ability to route information across their distributed elements. Our data revealed an overall trend of reduced global efficiency in neuronal networks derived from MDD patients compared to those of controls across different network sizes and r_thresh. However, this reduction trend in global efficiency did not reach statistical significance (no main effect of Group, F(222,1776) = 1.63, p = 0.194, Fig. 5). In addition, we did not detect any significant influence of different network sizes or r_thresh on group differences in global efficiency (no three-way interaction effect Group x Network size x Threshold, F(222,1776) = 1.86, p = 0.152).

Nevertheless, the data shows that global efficiency significantly increased as a function of network size (main effect of Network size, F(37, 296) = 23.56, p < 0.001). Additionally, global efficiency significantly decreased as the r_thresh used for constructing binarized graph increased (main effect of Threshold, F(6, 48) = 1157.36, p < 0.001).

Average node degree

Average Node degree is one of the most fundamental measures in graph theory analysis. It represents the number of functional connections the nodes in a network have on average. We computed the average node degree in our neuronal microcircuits and found a significant decrease in MDD patient-derived neuronal cultures as compared to those of the control cohort (main effect of Group; F(1,8) = 5.44, p = 0.048, Fig. 6). Similar to the previously reported graph metrics, average node degree exhibited a significant overall increase with the expansion of the network size (main effect of Network size, F(37, 296) = 167.82, p < 0.001) and under more lenient thresholds (main effect of Threshold, F(6, 48) = 622.24, p < 0.001).

We also performed post-hoc t-tests to determine the levels of Network size and Thresholds at which significant group differences in average node degree emerged. As seen in clustering coefficient, differences between the groups became more evident at larger network sizes (more than 20 cells) (refer to Supplementary Table 2 for post-hoc statistics). To demonstrate that group differences have not been driven by outliers, we refer the reader to the Supplementary Fig. 4.

In this work, we examined whether iPSC-derived neuronal networks exhibit differential functional organization in MDD. We investigated neuronal functional networks by means of Ca²⁺imaging and employed graph theory analysis to characterize global functional network topology, focusing on the properties of segregation and integration. We show an altered functional architecture reflected mainly by decreased clustering coefficient and average node degree of iPSCs-derived networks of depressed patients.

The significant decrease in clustering coefficient detected in patient-derived neuronal cultures signifies an altered local organization and diminished network segregation and functional specialization at the cellular level in depression. The impact of clustered topology on neural function [54, 55] and its role in shaping network dynamics [18, 56, 57] suggest that the breakdown of such clustered topology is likely to have implications for the network’s dynamics. This, in turn, can impact neural activity and potentially contribute to the restricted dynamics repertoire previously reported in major depression at the macroscale [58–60]. Nevertheless, further multiscale investigations are warranted to examine the relationship between topological changes and disease manifestation and to establish a causal link between these two phenomena. Additionally, segregated topology is important for the network’s robustness and resilience to insult [18, 61]. This influence suggests that a network lacking strong local interconnectivity can no longer confine and impede the propagation of pathological processes within the system. As a consequence, the network's capacity to withstand damage diminishes, rendering it more vulnerable to the spread of psychopathologies across its parts [5].

We observed a reduced average node degree in our iPSC-derived neural networks of MDD patients compared to those of controls. This finding is in line with previous postmortem investigations that have indicated a link between MDD symptoms and reduced synaptic density in prefrontal cortex and hippocampus [62, 63]. This link was further supported by findings in iPSCs of patients carrying a common mutation associated with major psychiatric disorders and implicated in synaptic regulation [64]. While we did not directly evaluate synaptic density and dendritic complexity in our neuronal cultures, we postulate that the observed decrease in overall node degree in our patients’ data could be a result of such reduced synaptic density, hindering the formation of connections between neurons. This assumption aligns with previous reports originating from both in vitro and in vivo investigations that highlight the impact of synaptic loss on network connectivity and dynamics in MDD. For instance, an in vitro study linked diminished dendritic complexity and synaptic density to altered network dynamics in iPSC-derived neurons of patients exhibiting mitochondrial pathology with depressive mood manifestations [65]. In vivo, Holmes et al. [66] observed changes in network connectivity related to reduced synaptic density in MDD patients. This established influence of molecular alteration on global network organization suggests that the observed reduction in average node degree in patient-derived networks might be a consequence of genetically determined variations in synaptic formation that affects micro-level network architecture and communication.

We did not observe any statistically significant differences in integration properties among our experimental groups, as measured by global efficiency. However, our patient data revealed a consistent trend toward reduced global efficiency, suggesting a potential significance that may become more apparent with a larger sample size. Effective integration of information relies on a relatively small number of efficient, yet energy-expensive [26, 67], long-ranging connections linking remote parts of the network. The reduction trend in global efficiency in our microscale networks of MDD patients might be a consequence of inadequate energy metabolism, primarily hindering the formation of these resource-intensive “shortcuts” that are crucial for optimal integration function. Interestingly, our group has previously documented bioenergetic imbalance and mitochondrial dysfunction in both fibroblasts and NPCs of MDD patients, with a significant overlap between the subjects in these studies and the participants in our current sample [37, 41]. The findings of these reports, coupled with evidence from large-scale studies, support the hypothesis linking MDD etiology to mitochondrial dysfunction [68, 69]. However, further inquiries are necessary to establish a direct link between impaired bioenergetic status and altered functional topology in depression.

The skin biopsies in this study were obtained at the end of the patients’ hospital stay after having received antidepressants and having nearly achieved remission. Variations between patients in use of medication and other epigenetic, environmental factors, are expected to be eliminated following multiple cell divisions of the fibroblasts and upon reprogramming [29, 70]. Consequently, the alterations in functional topology observed in the neural cultures of depressive patients suggest a genetic predisposition for neural networks to functionally organize differently in MDD. Given the pivotal role that genetics plays in shaping and regulating network organization and connectivity [71–73], the utilization of iPSC technology becomes invaluable for identifying measures of microcircuit topology as potential endophenotypes of neuropsychiatric disorders. Nevertheless, it is crucial to acknowledge that iPSCs could still retain residual epigenetic memory from their parent somatic cells, due, for example, to incomplete programming [74, 75]. Therefore, additional research is needed before making strong claims about the exclusive genetic or epigenetic origin of the microscale topological alterations observed in neuronal networks of MDD patients.

While our post-hoc analysis did not survive correction for multiple comparisons, the omnibus test of variance (ANOVA) demonstrated statistically significant results. This underscores that the observed group differences in clustering coefficient and average node degree are nonrandom and show substantial variations that could become more statistically pronounced with a larger sample size. Moreover, the graph topology in the cultured neuronal networks was markedly influenced by both the size of the network and the threshold used to binarize FC matrices. These two factors determine the number of nodes and edges that will be included in the graph. Larger network sizes (more nodes) and lower thresholds (more edges) expand the range of connectivity patterns available to the network. This variability in FC configurations increases the potential for an improved functional topology, explaining the noted increase in graph metrics as the size of the network increases and as the correlation threshold becomes more liberal.

Our study is subject to several limitations. Firstly, classifying members of the control group as non-depressed relied solely on self-reported absence of any history of depression. Consequently, we were unable to control for potential genetic or environmental risk factors that might have been present in the control cohort and predisposed individuals to depression. Secondly, the cultures used in this experiment were pure neuronal cultures. While human-derived monocultures of neurons offer a powerful approach for studying psychiatric disorders that is superior to using animal models or postmortem tissue, they inherently represent an oversimplified model of human brain tissue. Other alternatives include hiPSC-derived co-cultures of neurons and glial cells, as well as 3D cultures, encompassing both biology-based models, such as organoids, and engineering-based models, such as scaffolds and microfluidic platforms. These modelling strategies all carry a better resemblance to brain tissues, providing a microenvironment supportive of the interaction among different cell types, which is an essential element for neuronal health, functionality, and dynamics [76–79]. Another limitation of this study is the small sample size, which has been dictated by constraints in feasibility and resources. This particularly limited analyzing larger networks as subjects began to drop out due to a limited field of view in imaging (Supplementary Figure 1). While our analysis would certainly benefit from a larger sample, we observed robust group differences in clustering coefficient and average node degree that were not driven by outliers (Supplementary Figure 3 and Supplementary Figure 4, respectively).

To our knowledge, this is the first study that investigates the functional network topology in depression using the innovative technology of iPSCs. Our results support the view that considers depression as a network disorder involving aberrant information transfer and disturbed functional topology. Whether these alterations seen at a cellular resolution persist in fully developed brain networks is still unclear. Brain imaging research using resting state fMRI has remained inconclusive, showing disturbed segregation and integration attributes in depression but with inconsistent results [8, 10–13]. An exciting direction for future research would be to examine the micro-macro association of neural network topology by bringing together data from different spatial scales. There is accumulating evidence that underlines the interdependencies (and interaction) between different scales of spatial brain organization and the influence of one scale on shaping the other [80, 81]. Therefore, a scale-bridging or a “multiscale” approach is important to describe how network alterations in one level of brain organization can affect structure and function on other levels, and how this knowledge can aid in explaining cognitive dysfunction in the context of mental disorders.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding authors on request.

Author contributions

JVS, CHW, and RI conceived and designed the experiments. RI, VMM, and CHW performed the experiments and/or contributed to data acquisition. RI, SW, and JVS analyzed and interpreted the data. RI, JVS, CHW, VMM, and SW wrote the manuscript. RR provided resources, and JVS and CHW were responsible for general supervision. All authors approved the final version of the article.

Funding

The work has been supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) project number 422182557 to CHW, project number 422179811 to JVS, and project number GRK2174 to CHW, JVS, and RR. This work was also supported by the Deutscher Akademischer Austauschdienst (DAAD) program number 57381412 to RI.

Acknowledgments

The authors would like to acknowledge Tatjana Jahner for the excellent technical assistance.

Competing interests

Authors declare that they have no competing interests.

Supplementary information is available at MP’s website.

Sporns O, Tononi G, Edelman GM. Connectivity and complexity: the relationship between neuroanatomy and brain dynamics. Neural Netw 2000; 13: 909–922.
Sporns O. The human connectome: a complex network. Ann N Y Acad Sci 2011; 1224: 109–125.
Cocchi L, Zalesky A, Fornito A, Mattingley JB. Dynamic cooperation and competition between brain systems during cognitive control. Trends Cogn Sci 2013; 17: 493–501.
Bassett DS, Bullmore ET. Human brain networks in health and disease. Curr Opin Neurol 2009; 22: 340–347.
Fornito A, Zalesky A, Breakspear M. The connectomics of brain disorders. Nat Rev Neurosci 2015; 16: 159–172.
Fornito A, Bullmore ET. Connectomics: A new paradigm for understanding brain disease. Eur Neuropsychopharmacol 2015; 25: 733–748.
van den Heuvel MP, Sporns O. A cross-disorder connectome landscape of brain dysconnectivity. Nat Rev Neurosci 2019; 20: 435–446.
Gong Q, He Y. Depression, Neuroimaging and Connectomics: A Selective Overview. Biol Psychiatry 2015; 77: 223–235.
World Health Organization [WHO]. Depressive disorder (depression) [Fact sheet]. Geneva Switz World Health Organ 2023.
Meng C, Brandl F, Tahmasian M, Shao J, Manoliu A, Scherr M et al. Aberrant topology of striatum’s connectivity is associated with the number of episodes in depression. Brain 2014; 137: 598–609.
Zhang J, Wang J, Wu Q, Kuang W, Huang X, He Y et al. Disrupted Brain Connectivity Networks in Drug-Naive, First-Episode Major Depressive Disorder. Biol Psychiatry 2011; 70: 334–342.
Ye M, Qing P, Zhang K, Liu G. Altered network efficiency in major depressive disorder. BMC Psychiatry 2016; 16: 450.
Yu Z, Qin J, Xiong X, Xu F, Wang J, Hou F et al. Abnormal topology of brain functional networks in unipolar depression and bipolar disorder using optimal graph thresholding. Prog Neuropsychopharmacol Biol Psychiatry 2020; 96: 109758.
Achard S. A Resilient, Low-Frequency, Small-World Human Brain Functional Network with Highly Connected Association Cortical Hubs. J Neurosci 2006; 26: 63–72.
Bassett DS, Meyer-Lindenberg A, Achard S, Duke T, Bullmore E. Adaptive reconfiguration of fractal small-world human brain functional networks. Proc Natl Acad Sci 2006; 103: 19518–19523.
Perin R, Berger TK, Markram H. A synaptic organizing principle for cortical neuronal groups. Proc Natl Acad Sci 2011; 108: 5419–5424.
Shimono M, Beggs JM. Functional Clusters, Hubs, and Communities in the Cortical Microconnectome. Cereb Cortex 2015; 25: 3743–3757.
Yamamoto H, Moriya S, Ide K, Hayakawa T, Akima H, Sato S et al. Impact of modular organization on dynamical richness in cortical networks. Sci Adv 2018; 4: eaau4914.
Santos-Sierra D de, Sendiña-Nadal I, Leyva I, Almendral JA, Anava S, Ayali A et al. Emergence of Small-World Anatomical Networks in Self-Organizing Clustered Neuronal Cultures. PLOS ONE 2014; 9: e85828.
Yu S, Huang D, Singer W, Nikolić D. A Small World of Neuronal Synchrony. Cereb Cortex 2008; 18: 2891–2901.
Bullmore E, Sporns O. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci 2009; 10: 186–198.
Downes JH, Hammond MW, Xydas D, Spencer MC, Becerra VM, Warwick K et al. Emergence of a Small-World Functional Network in Cultured Neurons. PLOS Comput Biol 2012; 8: e1002522.
Watts DJ, Strogatz SH. Collective dynamics of ‘small-world’ networks. Nature 1998; 393: 440–442.
Bassett DS, Bullmore E. Small-World Brain Networks. The Neuroscientist 2006; 12: 512–523.
Fornito A, Zalesky A, Bullmore E. Fundamentals of Brain Network Analysis. Academic Press, 2016.
Achard S, Bullmore E. Efficiency and Cost of Economical Brain Functional Networks. PLOS Comput Biol 2007; 3: e17.
Takahashi K, Tanabe K, Ohnuki M, Narita M, Ichisaka T, Tomoda K et al. Induction of Pluripotent Stem Cells from Adult Human Fibroblasts by Defined Factors. Cell 2007; 131: 861–872.
Bardy C, Greenberg Z, Perry SW, Licinio J. Personalized psychiatry with human iPSCs and neuronal reprogramming. In: Personalized Psychiatry. Elsevier, 2020, pp 127–146.
Soliman MA, Aboharb F, Zeltner N, Studer L. Pluripotent stem cells in neuropsychiatric disorders. Mol Psychiatry 2017; 22: 1241–1249.
Sullivan PF, Neale MC, Kendler KS. Genetic Epidemiology of Major Depression: Review and Meta-Analysis. Am J Psychiatry 2000; 157: 1552–1562.
Shadrina M, Bondarenko EA, Slominsky PA. Genetics Factors in Major Depression Disease. Front Psychiatry 2018; 9.https://www.frontiersin.org/articles/10.3389/fpsyt.2018.00334 (accessed 28 Dec2022).
Vadodaria KC, Ji Y, Skime M, Paquola AC, Nelson T, Hall-Flavin D et al. Altered serotonergic circuitry in SSRI-resistant major depressive disorder patient-derived neurons. Mol Psychiatry 2019; 24: 808–818.
Vadodaria KC, Ji Y, Skime M, Paquola A, Nelson T, Hall-Flavin D et al. Serotonin-induced hyperactivity in SSRI-resistant major depressive disorder patient-derived neurons. Mol Psychiatry 2019; 24: 795–807.
Cavalleri L, Merlo Pich E, Millan MJ, Chiamulera C, Kunath T, Spano PF et al. Ketamine enhances structural plasticity in mouse mesencephalic and human iPSC-derived dopaminergic neurons via AMPAR-driven BDNF and mTOR signaling. Mol Psychiatry 2018; 23: 812–823.
Collo G, Cavalleri L, Chiamulera C, Merlo Pich E. (2R,6R)-Hydroxynorketamine promotes dendrite outgrowth in human inducible pluripotent stem cell-derived neurons through AMPA receptor with timing and exposure compatible with ketamine infusion pharmacokinetics in humans. NeuroReport 2018; 29: 1425.
Collo G, Cavalleri L, Chiamulera C, Merlo Pich E. Ketamine increases the expression of GluR1 and GluR2 α-amino-3-hydroxy-5-methy-4-isoxazole propionate receptor subunits in human dopaminergic neurons differentiated from induced pluripotent stem cells. NeuroReport 2019; 30: 207–212.
Triebelhorn J, Cardon I, Kuffner K, Bader S, Jahner T, Meindl K et al. Induced neural progenitor cells and iPS-neurons from major depressive disorder patients show altered bioenergetics and electrophysiological properties. Mol Psychiatry 2022. doi:10.1038/s41380-022-01660-1.
Cardon I, Grobecker S, Jenne F, Jahner T, Rupprecht R, Milenkovic VM et al. Serotonin effects on human iPSC-derived neural cell functions: from mitochondria to depression. Mol Psychiatry 2024; : 1–12.
World Health Organization [WHO]. The ICD-10 classification of mental and behavioural disorders. World Health Organ 1993.
Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry 1960; 23: 56.
Kuffner K, Triebelhorn J, Meindl K, Benner C, Manook A, Sudria-Lopez D et al. Major Depressive Disorder is Associated with Impaired Mitochondrial Function in Skin Fibroblasts. Cells 2020; 9: 884.
Okita K, Matsumura Y, Sato Y, Okada A, Morizane A, Okamoto S et al. A more efficient method to generate integration-free human iPS cells. Nat Methods 2011; 8: 409–412.
Schulze M, Hoja S, Winner B, Winkler J, Edenhofer F, Riemenschneider MJ. Model Testing of PluriTest with Next-Generation Sequencing Data. Stem Cells Dev 2016; 25: 569–571.
Yan Y, Shin S, Jha BS, Liu Q, Sheng J, Li F et al. Efficient and Rapid Derivation of Primitive Neural Stem Cells and Generation of Brain Subtype Neurons From Human Pluripotent Stem Cells. Stem Cells Transl Med 2013; 2: 862–870.
Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nat Methods 2012; 9: 671–675.
Pnevmatikakis EA, Giovannucci A. NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data. J Neurosci Methods 2017; 291: 83–94.
Smith SL, Häusser M. Parallel processing of visual space by neighboring neurons in mouse visual cortex. Nat Neurosci 2010; 13: 1144–1149.
Giovannucci A, Friedrich J, Gunn P, Kalfon J, Brown BL, Koay SA et al. CaImAn an open source tool for scalable calcium imaging data analysis. eLife 2019; 8: e38173.
McQuin C, Goodman A, Chernyshev V, Kamentsky L, Cimini BA, Karhohs KW et al. CellProfiler 3.0: Next-generation image processing for biology. PLOS Biol 2018; 16: e2005970.
Stirling DR, Swain-Bowden MJ, Lucas AM, Carpenter AE, Cimini BA, Goodman A. CellProfiler 4: improvements in speed, utility and usability. BMC Bioinformatics 2021; 22: 433.
Yan C, Li A, Zhang B, Ding W, Luo Q, Gong H. Automated and Accurate Detection of Soma Location and Surface Morphology in Large-Scale 3D Neuron Images. PLoS ONE 2013; 8: e62579.
Hagberg A, Swart P, S Chult D. Exploring network structure, dynamics, and function using networkx. Los Alamos National Lab. (LANL), Los Alamos, NM (United States), 2008https://www.osti.gov/biblio/960616 (accessed 25 Dec2022).
Rubinov M, Sporns O. Complex network measures of brain connectivity: Uses and interpretations. NeuroImage 2010; 52: 1059–1069.
Bisio M, Bosca A, Pasquale V, Berdondini L, Chiappalone M. Emergence of Bursting Activity in Connected Neuronal Sub-Populations. PLOS ONE 2014; 9: e107400.
Shein-Idelson M, Ben-Jacob E, Hanein Y. Engineered Neuronal Circuits: A New Platform for Studying the Role of Modular Topology. Front Neuroengineering 2011; 4.https://www.frontiersin.org/articles/10.3389/fneng.2011.00010 (accessed 28 Dec2022).
Wang. Sustained activity in hierarchical modular neural networks: self-organized criticality and oscillations. Front Comput Neurosci 2011. doi:10.3389/fncom.2011.00030.
Litwin-Kumar A, Doiron B. Slow dynamics and high variability in balanced cortical networks with clustered connections. Nat Neurosci 2012; 15: 1498–1505.
Holtzheimer PE, Mayberg HS. Stuck in a rut: rethinking depression and its treatment. Trends Neurosci 2011; 34: 1–9.
Kaiser RH, Andrews-Hanna JR, Wager TD, Pizzagalli DA. Large-Scale Network Dysfunction in Major Depressive Disorder: A Meta-analysis of Resting-State Functional Connectivity. JAMA Psychiatry 2015; 72: 603–611.
Tognoli E, Kelso JAS. The Metastable Brain. Neuron 2014; 81: 35–48.
Teller S, Granell C, Domenico MD, Soriano J, Gómez S, Arenas A. Emergence of Assortative Mixing between Clusters of Cultured Neurons. PLOS Comput Biol 2014; 10: e1003796.
Kang HJ, Voleti B, Hajszan T, Rajkowska G, Stockmeier CA, Licznerski P et al. Decreased expression of synapse-related genes and loss of synapses in major depressive disorder. Nat Med 2012; 18: 1413–1417.
Duman RS, Aghajanian GK. Synaptic Dysfunction in Depression: Potential Therapeutic Targets. Science 2012; 338: 68–72.
Wen Z, Nguyen HN, Guo Z, Lalli MA, Wang X, Su Y et al. Synaptic dysregulation in a human iPS cell model of mental disorders. Nature 2014; 515: 414–418.
Klein Gunnewiek TM, Van Hugte EJH, Frega M, Guardia GS, Foreman K, Panneman D et al. m.3243A > G-Induced Mitochondrial Dysfunction Impairs Human Neuronal Development and Reduces Neuronal Network Activity and Synchronicity. Cell Rep 2020; 31: 107538.
Holmes SE, Scheinost D, Finnema SJ, Naganawa M, Davis MT, DellaGioia N et al. Lower synaptic density is associated with depression severity and network alterations. Nat Commun 2019; 10: 1529.
Bullmore E, Sporns O. The economy of brain network organization. Nat Rev Neurosci 2012; 13: 336–349.
Tobe E. Mitochondrial dysfunction, oxidative stress, and major depressive disorder. Neuropsychiatr Dis Treat 2013; : 567.
Rezin GT, Amboni G, Zugno AI, Quevedo J, Streck EL. Mitochondrial Dysfunction and Psychiatric Disorders. Neurochem Res 2008; 34: 1021.
Garbett KA, Vereczkei A, Kálmán S, Wang L, Korade Ž, Shelton RC et al. Fibroblasts from patients with major depressive disorder show distinct transcriptional response to metabolic stressors. Transl Psychiatry 2015; 5: e523–e523.
van den Heuvel MP, van Soelen ILC, Stam CJ, Kahn RS, Boomsma DI, Hulshoff Pol HE. Genetic control of functional brain network efficiency in children. Eur Neuropsychopharmacol 2013; 23: 19–23.
Fornito A, Zalesky A, Bassett DS, Meunier D, Ellison-Wright I, Yücel M et al. Genetic Influences on Cost-Efficient Organization of Human Cortical Functional Networks. J Neurosci 2011; 31: 3261–3270.
Smit DJA, Boersma M, van Beijsterveldt CEM, Posthuma D, Boomsma DI, Stam CJ et al. Endophenotypes in a Dynamically Connected Brain. Behav Genet 2010; 40: 167–177.
Kim K, Doi A, Wen B, Ng K, Zhao R, Cahan P et al. Epigenetic memory in induced pluripotent stem cells. Nature 2010; 467: 285–290.
Ohi Y, Qin H, Hong C, Blouin L, Polo JM, Guo T et al. Incomplete DNA methylation underlies a transcriptional memory of somatic cells in human iPS cells. Nat Cell Biol 2011; 13: 541–549.
Feldt S, Wang JX, Shtrahman E, Dzakpasu R, Olariu E, Żochowski M. Functional clustering in hippocampal cultures: relating network structure and dynamics. Phys Biol 2010; 7: 046004.
Fields RD, Stevens-Graham B. New Insights into Neuron-Glia Communication. Science 2002; 298: 556–562.
Lemke G. Glial Control of Neuronal Development. Annu Rev Neurosci 2001; 24: 87–105.
Ma DK, Ming G, Song H. Glial influences on neural stem cell development: cellular niches for adult neurogenesis. Curr Opin Neurobiol 2005; 15: 514–520.
van den Heuvel MP, Scholtens LH, Kahn RS. Multiscale Neuroscience of Psychiatric Disorders. Biol Psychiatry 2019; 86: 512–522.
Bassett DS, Siebenhühner F. Multiscale Network Organization in the Human Brain. In: Multiscale Analysis and Nonlinear Dynamics. John Wiley & Sons, Ltd, 2013, pp 179–204.

The authors have declared there is NO conflict of interest to disclose

SupplementaryVideo1.mp4
Supplementary Video
MPsupplements.pdf

Download PDF

Unknown event
15 Jul, 2024
Editorial decision: Reject before peer review
12 Jul, 2024
Editor assigned by journal
13 Jun, 2024
Submission checks completed at journal
13 Jun, 2024
First submitted to journal
12 Jun, 2024

You are reading this latest preprint version

From Cell to Circuit: Investigating Functional Topological Changes in iPSC-derived Neuronal Networks in Major Depressive Disorder

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Cultivating primary human fibroblasts and generating hiPSCs

Generating hiPSC-derived neural progenitor cells (NPCs) and differentiating neurons

Immunofluorescence

Microscopy/Ca²⁺ Imaging

Motion correction and segmentation

Graph theory analysis

Statistical testing

Results

Demographic details and sample characteristic

Neuronal marker expression in iPS-derived neurons

Segregation properties

Integration properties

Average node degree

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1

From Cell to Circuit: Investigating Functional Topological Changes in iPSC-derived Neuronal Networks in Major Depressive Disorder

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Cultivating primary human fibroblasts and generating hiPSCs

Generating hiPSC-derived neural progenitor cells (NPCs) and differentiating neurons

Immunofluorescence

Microscopy/Ca2+ Imaging

Motion correction and segmentation

Graph theory analysis

Statistical testing

Results

Demographic details and sample characteristic

Neuronal marker expression in iPS-derived neurons

Segregation properties

Integration properties

Average node degree

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1

Microscopy/Ca²⁺ Imaging