Engineered Living Systems With Self-Organizing Neural Networks: From Anatomy to Behavior and Gene Expression

A great deal is known about the formation and architecture of biological neural networks in animal models, which have arrived at their current structure‐function relationship through evolution by natural selection. Little is known about the development of such structure–function relationships in a scenario where neurons are allowed to grow within evolutionarily‐novel, motile bodies. Previous work showed that ectodermal tissue excised from Xenopus embryos, develops into a three‐dimensional mucociliary epidermal organoid ex vivo and exhibits movements distinct from age‐matched tadpoles. These ‘biobots’ are autonomous, self‐powered, and able to move through aqueous environments. Here, we report a new type of biobot, the neurobot, composed of mucociliary epidermis and neural tissue. We show that neural precursor cells implanted in explanted Xenopus ectodermal tissue develop into mature neurons, extending processes both toward the surface and among each other. These self‐organized neurobots exhibit unique morphology, more complex movements, and different responses to neuroactive drugs compared to non‐neuronal counterparts. Calcium imaging confirms neuronal activity in neurobots. Transcriptomics reveals increased transcript variability, expression of genes related to nervous system development, a shift toward ancient genes, and up‐regulation of neuronal genes linked to visual perception.

Sensing cues from the environment and translating them into appropriate responses is the fundamental function of the nervous system in all animals. Critically, nervous systems endow animals with the ability to generate context‐ and experience‐dependent changes in their behavior. Nervous systems are also known to be plastic and adapt both structurally and functionally, on a much shorter timescale, to changes in sensory and/or motor effectors that might occur in the lifetime of an organism, for example, as a result of injury, amputation, or sensory deprivation [1, 2, 3, 4]. The capacity of an animal's nervous system to adapt to a new body plan is especially striking when the underlying sensory‐motor architecture is drastically altered. For example, ectopically induced eyes in the tails of Xenopus tadpoles have been shown to confer light sensitivity to otherwise eyeless hosts [5]. What are the limits of neuroplasticity in developing nervous systems? And how might wild‐type neurons establish coherent patterns and functional circuits when placed in an entirely new motile embodiment? Creating truly novel configurations of biological material allows us to probe the plasticity of evolutionary hardware to adapt on developmental (not evolutionary) timescales to truly novel circumstances and has applications for regenerative medicine, human augmentation, and biological engineering.

Two‐dimensional (2D) neuronal cell cultures grown in‐vitro provide one of the earliest demonstrations of neurons developing and functioning in a completely non‐native environment. These systems have been an extremely powerful tool for studying neural development, modeling emergent complex neural dynamics, and investigating disease mechanisms [6, 7]. Although 2D neuronal cell cultures are powerful and highly accessible systems, they lack the three‐dimensional (3D) architecture and cellular diversity of the brain, which limits their ability to model its complex circuit behaviors [8]. Recent advances in stem cell biology have enabled the creation of brain organoids, 3D self‐organizing neural tissues derived from pluripotent stem cells that can model some aspects of human neurodevelopment in vitro [8, 9]. These structures have layered neural circuits, spontaneous oscillations, early cortical‐like activity [10], have been used as a powerful tool to investigate neuropsychiatric disorders [11], and been more recently shown to have the capacity for basic forms of learning and memory [12]. Both 2D neural cultures and 3D brain organoids, however, remain non‐motile, lack the capacity to generate sensorimotor behaviors on their own, and have to be interfaced with computers or robotic systems via microelectrode arrays to perform simple tasks [13, 14, 15, 16]. Self‐contained biohybrid robots powered by muscular or neuromuscular actuation have been created by embedding muscle or combinations of neurons and muscle cells within synthetic scaffolds [17, 18, 19, 20]. These biohybrid robots, however, are not built exclusively using biological tissue, are not fully embodied, do not self‐assemble, and require external stimulation for propulsion.

Another related field is that of hybrots ‐ neural structures repurposed to drive engineered bodies, such as brains and fungi operating robotic vehicles [21, 22, 23, 24]. However, engineered robotics doesn't offer the full complexity and compatibility of biological tissues.

We sought to establish a fully biological model system in which we could investigate the morphology and function of self‐organizing neural networks in novel motile embodiments, gain deeper insight into the evolutionary developmental biology of the nervous system, and inform the design of future innervated biological robots. Here, we provide the initial characterization of an inexpensive, highly‐accessible, self‐assembling biobot model to answer fundamental questions about the persistence, morphology, and functional impact of wild‐type neural cells in a non‐standard body and identify possible sensory or behavioral endpoints for future investigation.

When ectodermal tissue is excised from the animal pole of a late blastula stage Xenopus embryo, and allowed to develop ex vivo, it will develop into a self‐motile, self‐powered 3D mucociliary epithelial organoid [25, 26], which we will refer to herein as biobot. These biobots express the four cell types normally present in the embryonic skin of Xenopus tadpoles. These include multiciliated cells (MCCs), mucus‐secreting goblet cells, ionocytes that regulate ionic homeostasis of epidermis, and small secretory cells (SSCs) [27, 28]. MCCs function as motor effectors, generating flow through the spontaneous beating of their cilia. As a result, these biobots are capable of navigating aqueous environments, generating a suite of stereotyped movement trajectories and velocities [25]. Ciliary beating frequency is thought to be under serotonergic control, with the SSCs secreting serotonin and stimulating an increase in ciliary beating frequency through serotonergic receptors expressed on MCCs [29].

Building upon this knowledge, we set out to explore what would happen if we provided these biobots with the raw materials needed to build a nervous system. Previous theoretical work suggests that nervous systems evolved primarily to coordinate and modulate movement rather than to support complex cognition [30, 31]. From this perspective, introducing neurons into biobots offers a unique opportunity to test whether even nascent neural circuits can shape or enrich spontaneously generated behavior, providing an experimentally tractable platform for probing early principles of neural organization. Moreover, it may offer a way of testing hypotheses linking neurons, ciliary function, and movement trajectories in an accessible model, which otherwise have been limited to paleontological data [32, 33]. The availability of ciliary biobots could help test models of motion patterns inferred from the geometry and microstructure of trails left on/in sediment by Cambrian ctenophores and fossil larvae with ciliary swimming bands [34, 35].

We show that neural precursor cells harvested from Xenopus embryos and implanted into biobots indeed differentiate into functional neurons that extend processes both within the construct and toward its outer surface. Neurobots exhibit marked differences in morphology and behavior relative to their non‐neuronal counterparts, suggesting the emergence of neural influences on movement. Transcriptomic profiling further reveals significant upregulation of genes associated with nervous system development in neurobots compared to non‐neuronal biobots [36, 37, 38, 39], including, intriguingly, genes important for processing light stimuli. Through detailed characterization of this novel composite system, we establish a platform that reveals its key features while enabling the generation of future mechanistic hypotheses.

To characterize the structure and function of nervous systems that self‐assemble within a novel embodiment, we established an experimental procedure for implanting biobots with neural precursor cells during the first few minutes of their formation. As shown previously, biobots can be constructed by excising tissue from the animal hemisphere of a Nieuwkoop and Faber stage 9 Xenopus laevis embryo (animal cap). Over the course of 30 minutes, the excised tissue gradually heals, initially forming a “bowl” shape before closing into a spherical structure [25]. We used this brief time window before tissue closure to introduce neuronal precursor cells into the interior of the healing tissue (Figure 1a (i, ii)‐top left panel).

To obtain neural cells, we took advantage of the fact that if the animal cap is excised and dissociated at the late blastula to early gastrula stage, and the dissociated cells are allowed to remain separated for 3 hours or more, they will assume a neural fate [40, 41]. To generate aggregates of neural precursors, we dissociated animal caps from approximately 50 embryos, allowed the cells to remain separated for 3 hours, and then reaggregated them. Clumps of these reaggregated cells were subsequently placed inside a freshly excised animal cap prior to its full closure (Figure 1a (i, ii), top left panel, see Methods). Within 30 minutes the formed composite assumes a spherical shape, and by the second day, it is fully healed (Figure 1a (ii), top right panel). As with non‐neuronal biobots, by the third day, multiciliated cells start appearing on their outer surface, and the bots start moving around in the dish. Similar to biobots, neurobots have a lifespan of about 9–10 days without being fed, and survive by consuming maternal yolk platelets present in all early Xenopus embryonic tissue [25].

Comparison of the gross morphology of neurobots and biobots revealed that, by Day 6, neurobots tend to exhibit a more elongated shape than biobots and become significantly larger (Figure 1b,c). To investigate whether the difference in size and elongation is simply due to implanting the animal caps with additional cells, we generated a third type of bot (sham neurobots) in a manner similar to neurobots, except that the implanted cells were not allowed to remain separated for 3 hours. Instead, they were reaggregated shortly after dissociation (within 30 minutes) to prevent the induction of neural fate. We found that the sham neurobots were not elongated and did not show a significant size difference compared to biobots (Figure S1a,b). These results suggest that the elongation and increase in size may be due to neuronal growth within neurobots.

To determine whether the implanted cells had indeed differentiated into neurons, we fixed and stained the neurobots with an antibody that specifically binds to acetylated α‐ tubulin, which is abundantly present in neurons and multiciliated cells (see Methods; neurons and multiciliated cells are readily distinguishable from each other due to their distinctive morphology). We found that the implanted neural precursor cells indeed differentiate into neurons (Figure 2). The neurons extend their processes not only within the neurobot, but also toward the outer surface (see arrows in Figure 2c,d). Such projections toward the cells lining the surface of the bot suggest the possibility of neurons modulating the activity of surface effectors including multiciliated cells and/or the activity of those that modulate the ciliary beating frequency, for example, small secretory cells.

In a subset of neurobots, we performed co‐labeling with acetylated α‐tubulin and microtubule‐associated protein 2 (MAP2). MAP2 preferentially labels neuronal cell bodies and dendrites and is absent from axons [42], whereas acetylated α‐tubulin is most abundant in stable microtubules within proximal axons and present at lower levels in dendrites [43]. We observed clear MAP2 expression in neurobots, with distinct and partially overlapping labeling patterns between MAP2 and acetylated α‐tubulin. This differential distribution suggests that neurobots develop both axonal and dendritic compartments (Figure 3a–c).

To evaluate the presence of synaptic structures, we co‐labeled neurobots with an anti‐synapsin‐1 antibody, which serves as a presynaptic marker by labeling synaptic vesicles. Numerous synapsin‐1‐positive puncta were detected throughout the neurobots, frequently colocalizing with regions positive for acetylated α‐tubulin (Figure 3d–f, Secondary‐antibody‐only control: Figure S2). This overlap suggests that stable microtubule‐rich processes in the neurobots contain presynaptic specializations or vesicle clusters, consistent with the formation of putative synaptic contacts.

There was considerable variability in the pattern of neural growth among different neurobots. No two neurobots showed identical neural architecture (Figure S3). This is not surprising given the variability of initial conditions resulting from their manual construction and the inevitable variability of the amount of implanted tissue. This variability, however, allowed us to investigate correlations between various physical and behavioral characteristics of neurobots, which we will discuss in more detail in the following sections. Despite these variabilities, we found that neural processes in most neurobots tended to emanate from one or more nuclear regions (labeled using a nuclear stain), presumably corresponding to the implanted clumps of cells (Figure 1a and Figure 2c,d; Figure S3). Interestingly, these regions were often surrounded by regions with seemingly no nuclear staining (Figure 2b–d; Figure S3). The presence of these seemingly empty spaces is intriguing. We hypothesize that these regions may be acellular, may contain support structures such as the extracellular matrix (ECM), and/or may be occupied by neural processes that lack acetylated α‐tubulin, a marker of stable microtubules found in mature neurons but absent from immature ones. Support for the ECM hypothesis includes cases in which we observed neurites traversing long distances in this empty space along a straight line (Figure 2b, yellow arrow).

In order to test these hypotheses, we stained neurobots with phalloidin dye, which stains F‐actin filaments, which are present in all eukaryotic cells [44]. Additionally, we performed Second Harmonic Generation imaging to assess the presence of collagen fibers [45], which are the most common component of ECM. Although we found some evidence for collagen fibers and puncta, their expression was rather sparse (Figure S4b, green arrows; Figure S4e). This space may therefore contain other ECM proteins that do not form fibrils such as proteoglycans, laminins, or fibronectin [46]. Similarly, phalloidin staining was largely absent from this space and mostly overlapped with acetylated α‐tubulin, with only a few instances of differential labeling (Figure S4c,d). These findings suggest that the space is predominantly acellular but may still contain a small number of neural processes, consistent with the observed sparse labeling of acetylated α‐tubulin and synapsin‐1 within this region (Figure 3d–f). Future experiments are needed to further investigate the molecular composition of this space.

To assess whether neurons within neurobots are functional, we built neurobots using neural cells extracted from embryos with genetically encoded calcium indicators (GCaMP6s, see Methods); this tool is commonly used to study neural activity [47]. We used a custom‐built widefield fluorescence microscope with a large field of view that enabled measurement of calcium activity in freely moving neurobots. Because the focal plane of the microscope was fixed in our setup, neural activity could only be recorded within a single plane of focus. It was therefore critical to maintain regions of interest within that focal plane, which proved challenging, particularly in neurobots exhibiting extensive rotational movements (See e.g., neurobot shown in Video S3). Neurobots with a flatter, disc‐like shape would be less likely to generate rotational movements and would therefore help maintain regions of interest in focus [48] (see Methods). Figure 4 shows an example of calcium signals recorded from a freely moving flattened neurobot, which exhibited circular movements around its center (Video S1). Motion‐corrected videos were then analyzed to extract the fluorescent activity (see Methods, Video S2). We found that the implanted cells indeed show calcium activity in all recorded neurobots. We occasionally observed synchronized activity in nearby or distant regions of interest (see e.g., arrowheads in Figure 4c), which may result from connectivity between these regions, although this could also be attributable to chance.

We found that keeping regions of interest in focus over longer periods of time was still quite challenging even in flattened neurobots. Further, although we did not formally study the impact of flattening on the movement patterns of neurobots, we found that they were in general less likely to move compared to regular ones. Consequently, we were unable to obtain repeated measurements of specific calcium activity sequences or movement patterns for statistical correlation analyses. Moreover, since every neurobot showed a different pattern of neural expression (Figure S3), findings in one neurobot could not be reproduced and confirmed in others. An important future direction will be to develop methods that enable consistent neural expression and experimental paradigms permitting repeated observation of directional movement and controlled rotation of regular, non‐flattened neurobots. Achieving these capabilities will be essential for rigorously linking neural activity to behavior.

To investigate whether neural activity could influence bot movement, we recorded the spontaneously generated movements of both biobots and neurobots. We reasoned that if neurons exhibit spontaneous and variable activity patterns, as observed in our calcium imaging experiments, and if those neurons can modulate the bot's behavior, then neurobots should exhibit movement dynamics that differ from those of their non‐neural counterparts. We Recorded the spontaneous movements of the bots in small 8‐well plates over 30 minute periods (n = 46 neurobots and 48 biobots, Figure 5a, Video S3: single neurobot moving, Video S4: neurobots moving in an 8‐well plate). We then used an automatic tracking software [49] to measure the position of each bot at each video frame. Figure 5b shows the details of the 2D trajectories of the 8 neurobots depicted in panel (a). There was a large degree of variability in these trajectories, with some bots moving in circular/oval trajectories with relatively constant diameter (Figure 5b, bots #3, #8); bots that followed circular trajectories varying in diameter over time (Figure 5b, bot#4); ones that made more complex, sometimes spirograph‐like patterns (Figure 5b, bots #1, #5, #6); those that were seemingly following the dish's boundaries (Figure 5b, bot #7); bots that circled over very small areas (Figure 5b, bot #2); and those that did not move at all (data not shown). Interestingly, all moving bots tended to exhibit repeating behavioral motifs.

With the positional information obtained from tracking, we used custom Python code to extract 4 kinematic parameters from the bots to compare neurobots with biobots. These were the total distance travelled, average speed, average acceleration, and the percentage of the well that was traversed in 30 minutes (see Methods). We found no significant difference in the total distance traveled, the percentage of the well covered, and average speed and acceleration (Figure S5). However, we found that the minimum movement speed of neurobots was significantly higher than that of biobots, indicating that neurobots tended to move more than biobots, remaining idle less often.

We had observed that some bots show simpler circular movement trajectories with relatively constant radii (e.g., Figure 5b‐bot #3), whereas others showed changes in their circling radius (e.g., Figure 5b‐bot #4) or a generally more complex movement pattern (e.g., Figure 5‐ bot #5). More complex movement patterns in neurobots would support the hypothesis that neural activity may play a modulatory role in their spontaneously generated movements. To further investigate potential differences in the movement patterns of neurobots and their non‐neural counterparts, we used a spectral analysis, calculating the Welch power spectral density (PSD) of the trajectory time series along the x and y coordinates (Figure 6a,b, two exemplar trajectories and their corresponding time series on the x‐coordinate, see Methods). We then detected significant peaks in the x and y PSDs, summed the number of unique peaks in x and y PSDs, and defined this number as the Complexity Index (CI, Figure 6c). In this analysis, a CI of one indicates a circular trajectory with constant diameter (where both x and y PSD have one peak at the same location), and the number increases as the trajectory becomes more complex. A complexity index of zero signifies non‐moving bots. We further used a time‐frequency analysis (wavelet‐transform) and surrogate data to confirm the significance of the power at the peaks identified in the Welch analysis (see Methods).

Interestingly, we found that neurobots showed a significantly higher degree of trajectory complexity compared to biobots (Figure 7a). This increased complexity could not be explained by the roundness or size of the biobots and neurobots, as these variables were not significantly correlated (Figure 7b,c). The increased complexity of neurobots could be a result of increased variability in the beating frequency of cilia in MCCs, changes in their spatial distribution, changes in the 3D structure of the bot, or changes in the activity or distribution of other cell types (including neurons) that may modulate the ciliary beating frequency, among other factors. When we measured this index in sham neurobots, we also found an increase, albeit not significant, relative to biobots (Figure S1c), indicating that the increased complexity we observe in neurobots is at least partially due to factors other than neural signaling. Like biobots, sham neurobots were more likely to show longer periods of immobility, and their minimum speeds were not significantly different from those of biobots (Figure S1d). Finally, consistent with the finding that minimum speed was significantly higher in neurobots (Figure 7d), we found that most inactive bots (Npeaks = 0) were in the biobot category, with 6 out of 48 (12.5%) biobots inactive, whereas only one out of 47 (2.1%) neurobots was inactive.

To further investigate whether neural activity could play a role in modulating trajectory complexity, we performed pharmacological experiments, treating groups of neurobots and biobots with entylenetetrazole (PTZ), a GABAA receptor antagonist used for its seizure‐inducing effects in animal studies [50]. Although we do not know the identity of neuronal constituents of neurobots, we reasoned that a positive result, for example, an increase in the complexity index after PTZ treatment, could be an indirect indication of the presence of GABAergic control of movement. To test this, we performed experiments in which we video recorded the movement of neurobots and biobots in regular media and compared the complexity indices after they were transferred to dishes containing 15 mM PTZ. To get an estimate on the baseline complexity index and account for potential variability due the transferring, we followed the experimental protocol depicted in Figure 8a.

The behavior of bots (16 neurobots and 16 biobots) was video recorded for 30 minutes in 8‐well plates filled with regular media (control 1), after which the bots were transferred to the second and third sets of dishes containing regular media (control 2, control 3). The bots were then transferred to dishes containing PTZ (Figure 8a). For each bot, we calculated the CIs for the three control conditions and used the average to calculate relative complexity measures for PTZ and the wash. To control for non‐specific effects that may occur upon the fourth transfer, we performed the same sequence of transfers for a second set of bots (8 neurobots and 8 biobots), where the fourth transfer was into wells that contained regular media and not PTZ (Figure 8c).

We found that all except two biobots showed a relative decline in their movement complexity while in PTZ (Figure 8b), thereby significantly reducing the CI relative to control (one‐sample t‐test, t(df) = 15, p = 0.009). The effect of PTZ on biobots suggests that PTZ may act on GABA receptors expressed by non‐neuronal cells or exert other off‐target effects. Indeed, GABAergic receptors are present on goblet cells lining the surface of both biobots and neurobots, which could indirectly modulate movement through altered mucus secretion (see Discussion).

The majority of neurobots, on the other hand, showed increased complexity relative to control, although a few did show a decline (Figure 8b, left panel). As a result of this dichotomy in the impact of PTZ, the average CI in neurobots was not significantly different relative to control (one‐sample t‐test, t(df) = 15 p = 0.39). When we compared the relative complexity after PTZ treatment, however, we found a significant difference between neurobots and biobots, with neurobots showing significantly higher values for relative complexity (Figure 8b, right panel). This effect was not observed when we transferred bots to the control media a fourth time (Figure 8c,d), confirming that the observed effect was specific to PTZ. The variability of the impact of PTZ on neurobots is in fact not surprising given the potential variability in the identity and the degree of expression of neurons (Figure S3). What is very interesting is the significant differential impact on the biobots and neurobots. Because neurobots are biobots + neurons, the fact that most of them showed an increase in their CI suggests a role for neural activity acting against the default inhibitory effect of PTZ on the movement of biobots. Alternatively, neural expression could indirectly contribute to these effects through its impact on the expression of MCCs or other cell types present on the outer surface of the bots.

To quantify the overall amount of neural expression and its relationship with neurobot morphology and behavior, we used the confocal images of neurobots immunostained with acetylated α‐tubulin antibody, which labels neurons and cilia in the MCCs. Using this data, we traced the neural processes and quantified the position of the MCCs using Imaris software (Figure 9a,b, see Methods). In this analysis, we did not distinguish between axons or dendrites and could not determine whether these processes belonged to individual neurons. We could, however, obtain rough estimates of the total amount of neural tissue and the degree of branching by calculating the total length of neurites and the number of terminals, which were defined as the number of nerve endings, agnostic of their identity (i.e., axonal or dendritic). Figure 9 (top panels) shows two examples of such traces in cases of neurobots with very few and many neurites. We calculated the correlation between the degree of neural growth, expression of MCCs, bot size, and shape, as well as the CI of the trajectories across all neurobots for which we had both behavioral and structural data (Figure 9c). We also calculated the correlation of all these parameters with the relative amount of neural tissue that was implanted on Day 1 as the neurobots were constructed. This ratio was calculated by dividing the area of the implanted clumps by that of the external shell (Figure 9c: Neu/Ect, Figure 1a (ii), see Methods).

Based on this analysis, we found that the total number of neuron terminals was highly correlated with both absolute (Pearson correlation coefficient = 0.9, t‐test, t(df) = 20, p = 2.63e‐09) and area‐normalized neural length (neurite density, Pearson correlation coefficient = 0.9, t‐test, t(df) = 17, p = 2.77e‐06). Interestingly, we found a significant negative correlation between neurite density and MCC expression density (Pearson correlation coefficient = −0.6, t‐test, t(df) = 12, p = 3.33e‐02); neurobots with higher neurite density tended to have lower overall density of MCCs. Consistent with this finding, we found that biobots (which do not have neurons) have a significantly higher density of MCCs compared to neurobots (Figure S6, Kruskal–Wallis test, p = 0.027). Additionally, we found a significant negative correlation between the Roundness Index (RI) and neurites’ absolute length (Pearson correlation coefficient = −0.6, t‐test, t(df) = 20, p = 5.09e‐03), neurites’ normalized length (Pearson correlation coefficient = −0.5, t‐test, t(df) = 17, p = 1.72e‐02), and total number of terminals (Pearson correlation coefficient = −0.5, t‐test, t(df) = 20, p = 1.72e‐02). That is, the more elongated the bot, the more neural expression, suggesting that the elongation could be a result of neural processes growing within the neurobot. This hypothesis is consistent with the finding that sham neurobots were not different from biobots in their RI (Figure S1a, Kruskal–Wallis test, p = 0.89). We did not find a significant correlation between CI and neural expression metrics although the neurobot with the highest complexity index also had the largest number of terminals and neurite length (see the outlier data point in the panels).

Similarly, we found only a small correlation (non‐significant) between the relative amount of implanted tissue and the degree of neural expression metrics.

Previous studies showed that treatment with zolmitriptan, which is a selective 5‐hydroxytryptamine (5‐HT) 1B/1D receptor agonist, increased the degree of ectopic (but not native) neural growth in Xenopus embryos [51]. We investigated whether this treatment would have an impact on the degree of neural growth in neurobots, where the neurons are all ectopic. Interestingly, we found that this treatment increased the degree of neural expression in neurobots as well, although it did not have a significant effect on any of the behavioral measurements including trajectory complexity (Figure S7). In this group, we found a tight correlation between the ratio of neural implant to ectoderm shell and total number of terminals, as well as total neural length (Figure S7). These results indicate that neurons in a neurobot behave as ectopic, not native, cells. Notably, 4 of the 16 neurobots in the PTZ study presented in the previous section were cultured in zolmitriptan (filled circles Figure 8b; Figure S7), three of which showed an increase in relative complexity. This result points to the possibility that this treatment may bias neural expression toward those that respond to PTZ that is, GABAergic neurons. Further experiments are required to characterize the impact of zolmitriptan treatment on the neural expression patterns within neurobots.