Motile Living Biobots Self-Construct from Adult Human Somatic Progenitor Seed Cells

Fundamental knowledge gaps exist about the plasticity of cells from adult soma and the potential diversity of body shape and behavior in living constructs derived from genetically wild‐type cells. Here anthrobots are introduced, a spheroid‐shaped multicellular biological robot (biobot) platform with diameters ranging from 30 to 500 microns and cilia‐powered locomotive abilities. Each Anthrobot begins as a single cell, derived from the adult human lung, and self‐constructs into a multicellular motile biobot after being cultured in extra cellular matrix for 2 weeks and transferred into a minimally viscous habitat. Anthrobots exhibit diverse behaviors with motility patterns ranging from tight loops to straight lines and speeds ranging from 5–50 microns s−1. The anatomical investigations reveal that this behavioral diversity is significantly correlated with their morphological diversity. Anthrobots can assume morphologies with fully polarized or wholly ciliated bodies and spherical or ellipsoidal shapes, each related to a distinct movement type. Anthrobots are found to be capable of traversing, and inducing rapid repair of scratches in, cultured human neural cell sheets in vitro. By controlling microenvironmental cues in bulk, novel structures, with new and unexpected behavior and biomedically‐relevant capabilities, can be discovered in morphogenetic processes without direct genetic editing or manual sculpting.

Keywords: biobot, bioengineering, emergence, morphogenesis, repair, self‐assembly

What is the latent space of possible functional morphologies that cells, with a wild‐type genome, can be coaxed to construct?[ 1 ] This question drives at the heart of fundamental issues in evolutionary, developmental, cell, and synthetic biology, and has been taken up by a rapidly growing field focusing on building new kinds of active living structures: biobots.[ 2 ] This emerging multidisciplinary effort to control the behavior of cellular collectives has garnered much excitement for two main reasons. First, it offers the possibility of using engineering to reach outcomes that are too complex to micromanage directly, and hence promises to revolutionize efforts to produce complex tissues for clinical applications in regenerative medicine and beyond. Second, increased control over the morphology and behavior of cellular collectives by leveraging morphogenetic tissue plasticity could enable the development of self‐constructing living structures by design with predictable and programmable functional properties and numerous practical uses, greatly extending the current abilities of traditional fabrication practices in diverse fields as robotics,[ 3 ] architecture, sustainable construction, and even space exploration.

In the last decade, interest in developing biological structures de novo has seen a rapid surge.[ 4 ] Among these efforts, a subset of functional biogenic assemblies gave rise to a special class of motile synthetic structures dubbed biobots. Early examples of biobots are hybrids between biological cells and inert chemical substances supporting them, such as gels or 3D‐printed scaffolds.[ 5 ] These assemblies incorporated living cells ranging from bacteria to diverse mammalian tissues such as nerve, muscle, and neuromuscular junctions (NMJs), as well as engineered cell lines with programmable features, all carefully crafted into diverse 3D scaffolds designed to harness and amplify the innate functionality of biological cells.[ 6 ]

A different approach resulted in Xenobots, the first fully‐biological biobots created by sculpting or molding amphibian embryonic cells into multicellular structures that can spontaneously locomote without external pacing.[ 7 ] But it was not known how general these phenomena are, whether this kind of plasticity extended to mammals, or what the throughput of this technology can be. Thus, we sought to address whether the capacity of genetically unaltered cells to generate a self‐propelled, multicellular living structure in this way is unique to amphibian embryonic cells, and whether such a living structure can be built without needing to be individually sculpted or molded, but instead coaxed to self‐construct from an initial seed cell, resulting in a high‐throughput process wherein large numbers of biobots can be grown in parallel.

Here, we introduce novel, multicellular, fully biological, self‐constructing, motile living structures created out of human lung epithelium. We refer to them as Anthrobots, in light of their human origin and potential as a biorobotics platform.[ 2 , 8 ] We quantified their emergent natural, baseline properties as an essential background characterization of their native capacities which will serve as targets for future efforts to reprogram form and function for useful purposes. Anthrobots self‐construct in vitro, via a fully scalable method that requires no external form‐giving machinery, manual sculpting, or embryonic tissues and produces swarms of biobots in parallel. They move via cilia‐driven propulsion,[ 9 ] living for 45–60 days. We quantitatively characterized the range of movement and morphological types, showing that their behaviors are strongly correlated with specific features of their anatomy. The ability of adult, somatic, human cells to form a novel functional anatomy, with unique behaviors, reveals that this plasticity is not restricted to amphibian or embryonic cell properties, and is a fundamental feature of wild‐type cells that requires no direct genetic manipulation to unlock. Furthermore, we found that Anthrobots exhibit a highly surprising behavior given their origin as static airway epithelium: they can move across scratches in (human) neuronal monolayers and induce gap closures across these scratches. Numerous in vitro and in vivo uses of such functional living structures can be envisioned, especially because they can now be made from the patient's own cells.[ 10 ]

We developed the Anthrobots by leveraging normal human bronchial epithelial (NHBE) cells’ native tissue plasticity and unlocked a novel morphology, which is not apparent from the reliable, default developmental patterning of airway epithelium, in order to fulfil target functional and structural needs of creating a self‐constructing, multi‐cellular, 3D, motile living structure of human‐origin. Airway organoids with apical out tissue organization, yielding ciliated spheroids anatomically similar to the Anthrobots, have very recently been shown using different protocols,[ 11 ] each starting from individual normal bronchial epithelial cells isolated from airway epithelium extracts. Apical out spheroids made out of intact airway epithelium extracts have also been produced[ 12 ]). Each one of these three protocols is optimized for different priorities such as ease of organoid access during its development,[ 11c ] structural uniformity in final products,[ 11a ] and ability to easily modulate resulting organoid size.[ 11b ] The common denominator between these parallel advances is that they are characterized as organotypic cultures exclusively, enabling scientists to investigate lung anatomy, function, and pathology. Beyond native tissue recapitulation, these constructs’ abilities as functional assemblies, range of behavioral and morphological patterns, as well as functional correlations between these patterns have yet to be explored.

These three methods, plus the one detailed in this paper, constitute convergent but distinct technical approaches toward producing the novel morphology of cilia‐out spheroids derived from human airway epithelium. These four protocols for creating cilia‐covered NHBE‐derived spheroids differ in their apical orientation from the earlier established approaches for creating traditional airway organoids where cilia develop as lining the lumen.[ 13 ] Boecking & Walentek[ 11c ] grow airway organoids as embedded in a collagen‐rich matrix (as opposed to the traditional Matrigel approach) and also cultures them in air‐liquid‐interface (ALI) inserts, which have traditionally been used with NHBEs for 2D differentiation into airway epithelium, enabling ease of access to the airway organoids during their developmental course.[ 11c ] After this initial ALI culture period of 14 days, the mature airway organoids are dissolved from the collagen matrix and replated into a fresh same matrix of similar composition (to remove catabolites) for another 14 days. It is in this second 14 day period that cilia localization on the surface is accomplished by administering R‐Spondin‐2 (RSPO2) and Noggin into the matrix. Accordingly, this Boecking & Walentek method consists of two consecutive 2‐week periods of matrix‐embedded growth and differentiation: first period without and the second period with RSPO2&Noggin.

The method introduced in our paper is most similar to the Boecking & Walentek method in that the initial proliferation of individual NHBEs into spheroids with cilia‐lined lumen is accomplished by culturing them as embedded in a gel‐based matrix. However, in our method, upon dissolution of spheroids from matrix at the end of this 14 day period, the cilia‐in spheroids are not plated back into the matrix, and instead, the cilia localization into the spheroid cortex is achieved by culturing these spheroids in low‐adhesive environments. Accordingly, cilia localization is observed within one week, making our method a faster (20 days between single cell to cilia‐coated spheroid as opposed to 28), less laborious (single matrix dissolution, as opposed to two), and potentially higher‐throughput (since each time matrix is dissolved, a certain percentage of spheroids are lost with it, as is also reported by Boecking & Walentek). The remaining two cilia‐out protocols (Stroulios and Wijesekara) are in contrast not leveraging the self‐construction ability of NHBEs, and instead form spheroids by means of cell aggregation in U‐bottom wells without the presence of a matrix, which in turn provide them with higher regularity of spheroid size (given each spheroid self‐assembles with a similar number of constituent cells).

The Stroulios et al method aggregates single cells in low‐adhesive micro aggregation chambers first, then transfers them into regular wells for differentiation in matrix‐free liquid environment with bronchial epithelial differentiation medium, achieving uniformity in organoid morphology and size.[ 11a ] Similarly, the Wijesekara et al method also first aggregates individual cells into spheroids, and then either transforms them into a matrix environment for differentiation, or keeps them in the liquid environment with bronchial epithelial differentiation medium, both approaches yielding apical‐out spheroids with the ability to control the resulting ciliated spheroid size by modulating the initial aggregate size.[ 11 , 14 ] In summary, our method facilitates differentiation as embedded in extracellular matrix, enabling spheroids to self‐construct, a feature the other two methods Stroulios and Wijesekara lack; though in turn they achieve higher spheroid regularity and control over spheroid size. The major difference between our method and the Boecking & Walentek method is that upon extraction of apical‐in spheroids from Matrigel, we simply culture the spheroids in a non‐adhesive environment and achieve polarity reversal and motility within a few days. However, the Boecking & Walentek method replates the spheroids into the Matrigel for another two weeks, during which period they administer RSPO2&Noggin in order to facilitate polarity reversal, and at the end of this period they dissolve the Matrigel again to finally harvest the apical out spheroids. We believe our method provides a simple, rapid, scalable, and high‐throughput protocol that harnesses biological cells’ ability to build themselves into multicellular complex structures.

Despite their differences in methods, goals, and characterization metrics, these four novel protocols taken together help explore the space of possible morphologies of NHBEs and unravel the morphogenetic potential of human airway epithelium; they are thus significant steps for mapping the morphogenetic plasticity landscape of non‐embryonic wild‐type cells.

To explore the self‐organizing plasticity of morphogenesis without genomic change, we chose a cellular substrate in which such outcomes would be most surprising: adult, somatic, human airway tissues. To study and steer the in vitro morphogenesis of novel 3D tissues with motile appendages, we developed a novel protocol (Figure 1A) that builds upon the existing ability of human bronchial epithelial progenitor cells to form multicellular spheroids (Figure 1a.1) with cilia‐lined lumina[ 15 ] (i.e., apical‐in configuration). We modified this process by manipulating the culture environment such that it now yields cilia‐coated (i.e., apical‐out) spheroids, which exhibit spontaneous locomotive ability.

A key step in their construction, in order to obtain significant translocation, is the induction of cilia to face outward. Given that cilia naturally localize into the lumen due to the basal cells’ interaction with the surrounding high‐viscosity matrix, we hypothesized that changing the culture environment to a lower viscosity level (e.g., water‐based media instead of gel‐based matrix) may trigger the basal layer cells to migrate inward and allow the apical layer to take their place on the spheroid cortex.[ 16 ] Thus, to trigger apicobasal polarity switching, we first grew airway organoids embedded in Matrigel as reported previously[ 13 ] (Figure 1a.2), which does not yield either ciliated or motile spheroids. We have then proceeded to dissolve the surrounding matrix while keeping the spheroids intact and transferred them into a low‐adhesive environment (Figure 1a.3) and induced them with retinoic acid on a bidaily basis while performing media changes every 3 days. (Matrigel's elastic modulus is 450 Pa[ 17 ] whereas water based liquid media's elastic modulus in this low‐adhesive environment is 2*109 Pa). This new approach has enabled the originally apical‐in spheroids (that show no motility on day 0 as shown on Figure 1b.1) to became motile by day 7 (Figure 1b.2), featuring highly motile ciliary appendages on the spheroid surface. A high‐resolution high‐speed capture of ciliary movement in Anthrobots (supplemental video 1) show that they deploy a similar propelling strategy observed in multiciliate motile organisms.[ 9 ]

We next examined two aspects of the microenvironment as possible control parameters for properties of Anthrobot self‐assembly. First, we tested the role of matrix viscosity, which is known in other bioengineering contexts to impact diverse cell properties, from secretory profile[ 18 ] to mechanical attributes.[ 19 ] We observed that culture environments with higher viscosity levels than the protocol baseline result in decreased motility (Figure S2, Supporting Information) and size (Figure S3, Supporting Information), suggesting that low‐viscosity environments better facilitate the growth of functional bots as well as yield larger bots. Second, we examined cell seeding density as a factor for motility (Figure S4, Supporting Information) and size (Figure S5, Supporting Information). We set up three separate conditions: one with the default seeding density in Matrigel (x = 30 000 cells mL−1), one with double this density (2x = 60 000 cells mL−1), and one with half this default density (x/2 = 15 000 cells mL−1). After growing the bots under these three conditions for two weeks, while keeping all other protocol aspects constant, we dissolved the mature spheroids from the matrices and measured their size immediately. We then continued culturing the bots per usual maintenance protocol and measured their motility in a binary fashion (i.e., displacing mover or not) during the period where bots show peak motility (between days 9–20) on a bidaily basis. In both experiments, we observed significant differences in the resulting bot sizes and motility (on particular days) among different seeding density conditions, though the effect of introducing additional cells was not linear at any time point tested.

These results show that the size and time course of maturation of motile bots could be modulated by altering the concentration of cells in the Anthrobot differentiation culture, suggesting initial cell seeding density to be a tractable control knob for Anthrobot morphology and function.

To characterize the temporal dynamics of motility initiation, we periodically (every other day) counted the number of motile spheroids for 3 weeks following dissolution and observed a sigmoidal motility profile with peak change in motility on day 10 (Figure 1C). We next confirmed that this drastic change in motility occurred as a result of a morphological reorganization event exposing cilia on the cortex (Figure 1D). We immunostained the spheroids on day 0 (pre‐motility) and day 7 (post‐motility) with DAPI and for the apical markers a‐tubulin (cilia marker) and ZO1 (tight junction marker), revealing a drastic increase in external multi‐ciliated cells on day 7 compared to day 0. Figure 1E shows the tissue organization within an ≈50‐micron depth of a typical Anthrobot. Upon observing increased multi‐ciliated cell presence in motile subjects, we sought to definitively attribute the emergence of motility to the presence of surface cilia. We administered the efficient blocker of cilia motion, ciliobrevin,[ 20 ] and observed the expected drastic decrease in motility (See Figure S6, Supporting Information), confirming that the motility of Anthrobots is cilia‐driven.

Despite their wild‐type human genome and somatic origin, these self‐motile constructs exhibited a wide range of behaviors and an anatomy that differs from the species‐specific body morphology. To characterize this diverse landscape and uncover the developmental features of Anthrobots, next we sought to characterize these behavioral and morphological capabilities, and investigated a potential correlation between their form and function. One initial key task for Anthrobots, as with any new behavioral subject,[ 21 ] is to determine whether its major morphological properties and activities are discrete, continuous, or uniform characters.[ 22 ] Thus, we quantitatively analyzed their range of behavior modes in time‐lapse videos of ≈200 randomly‐selected motile spheroids (Figure 2A, and Videos S2–S5, Supporting Information) for 5 h in groups of 4 or 5 Anthrobots, and extracted their movement trajectory coordinates. We then split up these 5 h‐long trajectories into 30 s periods to classify behavior with a higher degree of granularity and in an aggregate manner. To identify patterns within a potentially unlimited set of possible movements, we characterized these periods by how straight and/or circular they are as all possible trajectories can be explained together by these two properties. To this end, we used two main trajectory characterization metrics: straightness and gyration indices (see Experimental Section for detailed description of how these indices are calculated) and plotted all viable trajectorial periods along these two indices (Figure 2B). We then ran the unsupervised clustering algorithm Ward.D2, a common hierarchical clustering method (see methods for more details), on this 2D plot and observed four statistically distinct clusters to emerge (Figure 2C).

Further investigation of these clusters reveals that each represents a distinct movement type: circular, linear, curvilinear and “eclectic” (Figure 2D). Further analysis of each cluster in terms of its homogeneity (measured by “average dissimilarity” index), which is a measure of the intra‐cluster variation, and its size (measured by “% of observations”) was performed (Figure 2E), as well as a quantitative comparison between different clusters along the two major movement indices (Figure 2F).

As a result of these behavioral characterization analyses, we observed that the circular bots (type 1, Figure 2D) score the highest on gyration and lowest on straightness indices (Figure 2F). They also have highly similar trajectories and are very common among the behaviors of Anthrobots given this cluster has the smallest homogeneity and a representation of over 30% of all the recorded periods (Figure 2E). We also observed that the linear bots (type2, Figure 2D) score the highest on straightness and lowest on gyration indices (Figure 2F). They have less homogeneity than circular bots but also have the greatest representation out of all clusters (Figure 2E). Accordingly, circular and linear bots together make up more than half of the population, and each have highly homogeneous populations. Finally, the third most common (Figure 2E) type of bot is the curvilinear bot (type 3, Figure 2D), which scores high on both the gyration and straightness indices (Figure 2F) and has the second most heterogeneous trajectories (Figure 2E). Bots with most disorganized trajectories and smallest representation in the overall population (Figure 2E) are the eclectic bots (type 4, Figure 2D), which score the lowest on both the gyration and straightness indices (Figure 2F) due to exhibiting eccentric trajectories that are often a combination of the other three types.

After having characterized each major movement type observed in Anthrobots, we next investigated the transition probabilities between each pair of behavior types. In order to estimate the stability of each trajectory and state transitions between different movement types, we used a Markov chain model shown on Figure 2G, which revealed the degree of commitment to a given behavior (persistence) and provided an ethogram of Anthrobot behavior. We observed that the most stable movement pattern for an Anthrobot is circular motion, followed by linear/curvilinear motion. The eclectics act more like an intermediate and over time, at least probabilistically, resolve into one of the 3 other categories. Therefore, we conclude that the vast majority of Anthrobot movements can be broken down into simpler, highly consistent patterns like linear, circular, curvilinear, with eclectics acting as a transient intermediary. The fact that Anthrobots exhibit movement types with high “consistency” and low rates of inter‐type conversion (e.g., between circulars and linear) suggests that Anthrobots self‐organize into discrete and stable movement types, each bot having a distinct motility fingerprint.

Having observed several distinct movement types, we next asked whether the range of Anthrobot morphologies was continuous or again composed of discrete categories.[ 23 ] This question is important for both understanding the macro‐scale rules of self‐assembly, and for future efforts to control their functional properties. We hypothesized the primary parameters of this possible underlying morphological framework to be a function of the Anthrobots’ 3D shape and overall cilia distribution pattern, since Anthrobot motility is generated by cilia. Accordingly, we collected 3D structural data (Figure 3a.1) from ≈350 Anthrobots through immunocytochemistry / immunofluorescence (ICC/IF) and confocal microscopy, focusing on shape and cilia distribution pattern properties, and binarized these morphological features (Figure 3a.2) to extract quantitative information on cilia and body boundaries. We then plotted this information for ≈350 Anthrobots along eight different morphological characterization indices we developed, each quantifying a different aspect of the Anthrobot shape and cilia pattern. (Figure 3B; Figure S7A, Supporting Information).

The shape‐related indices among these eight formal morphological characterization indices included the ratio between the longest and shortest distance within a spheroid (i.e., “aspect”), longest distance within a spheroid (i.e., “max radius”), how invaginating or protruding the spheroid surface is (i.e., “shape smoothness”); the cilia‐related indices included the total area covered by cilia signal on a spheroid surface (“cilia points”), cilia signal per unit area on a spheroid surface (“cilia points/area”), proximity of current cilia point distribution to a complete random uniform distribution (“cilia distribution homogeneity”), how clustered the cilia are on a spheroid surface (“polarity”), and how many free‐floating cilia points there are that are not a part of a cluster (“noise points”). See Experimental Section for more details on how these morphological indices were calculated.

Next, we performed a Principle Component Analysis (PCA) followed by an unsupervised clustering algorithm on this 8D data cloud and observed the emergence of three statistically distinct clusters (Figure 3C), each representing a distinct morphological type (morphotype). Figure 3D shows a quantitative characterization of each cluster along 8 different morphological indices. This analysis revealed the following two characteristics to be the most important distinguishing factors (to an equal degree, both ranking the top place in Principle Component (PC)1 contributions) between different morphotypes: the size of the Anthrobot (measured by “max radius”), and the uniformity of its shape (measured by “shape smoothness”). These two most distinguishing characteristics formally describe type 1 bots to be significantly smaller in size and smoother (spherical) in its volume, while type 2 bots to be the largest and least uniformly shaped, and type 3 to be somewhere in between the two both. (See Materials and Methods section for PC1 and PC2 contribution rankings of different indices used to uncover this hierarchy.)

At the second level of importance in distinguishing between these three morphotypes is a set of four indices (all ranking equally top place in PC2 contributions), pertaining to cilia characterization. While the first two of these indices characterize cilia count, i.e., the density of cilia per Anthrobot (measured by “cilia points”), and the density of cilia per unit area of Anthrobot (measured by “cilia points/area”); the remaining two indices characterize the pattern in which these cilia are distributed: how tightly grouped the cilia are (measured by “polarity,”), and the number of “free‐floating” ciliary patches that are not within a group (measured by “noise points”). These four indices together describe type 2 bots as being significantly more ciliated than type 1 and type 3 bots, and type 3 bots as having a significantly more polarized cilia distribution pattern (with the least amount of extra‐cluster noise) in comparison to type 1 and type 2 bots (Figure 3D).

The third most important (scoring a second level rank in both PC1 and PC2) characteristic in distinguishing between the different morphotypes is a function of both the size/shape of the Anthrobot and the localization pattern of its cilia: the homogeneity of cilia distribution on the surface of the Anthrobot (measured by “cilia distribution homogeneity”). This local index is related to, but not directly anti‐correlated with, the polarity index, because while cilia distribution homogeneity characterizes local neighborhood patterns, polarity (along with its supporting index “noise points”) characterizes the global (entire Anthrobot‐level) cilia distribution. (See Experimental Section for more information). In this way, we obtain both a local and a global view of the cilia distribution patterns at once and identify type 1 bots as both globally and locally homogeneous, type 2 bots as globally homogeneous but locally heterogeneous, and type 3 bots as both globally and locally heterogeneous with high degree of global polarization.

In summary, our morphological characterization pipeline suggest that Anthrobots self‐organize into 3 major morphotypes (Figure 3E) and this relationship can be represented by a developmental decision tree shown on Figure 3F wherein the first “decision point” determines the Anthrobot size and shape. Accordingly, bots that are small and regularly shaped (morphotype 1) form one branch, whereas bots that are larger and more irregularly shaped (morphotypes 2 and 3) form the alternating branch. On this alternating branch a second decision point forms further downstream and determines Anthrobot cilia pattern. Anthrobots with a non‐polarized cilia pattern form one branch (morphotype 2), and Anthrobots with a polarized cilia pattern form the other (morphotype 3).

Finally, one characteristic that does not seem to be changing in any significant way between these three morphotypes is the ratio between the longest and shortest distance within a spheroid (measured by “aspect”). Although the 3 morphotypes differ significantly in terms of the volumetric regularities (measured by the shape smoothness index) as explained above, their aspect ratios are statistically very similar.

Having observed the emergence of several discrete types of movement (Figure 2) and morphology (Figure 3), we next decided to investigate whether there is a mapping between Anthrobots’ different movement types and morphotypes. To do this, we incorporated an additional level of movement‐type information into the PCA analysis used for identifying the morphotypes as introduced in Figure 3. During the initial sample collection process for this analysis, we had been able to definitively distinguish between non‐motile Anthrobots (non‐movers) and motile Anthrobots (movers) as described in Figure 1C. To further represent the movement types observed within the mover population, we randomly sampled from the set of motile subjects, targeting 30 Anthrobots that translocated (i.e., displacing movers) to assign a movement type. Selected displacing movers were randomly collected from the two most orthogonal movement types, circulars and linear, in approximately equal proportions.

Next, we identified these non‐mover and displacing mover Anthrobots within the PCA cloud presented in Figure 3C and assigned them this additional layer of information, i.e., movement type (Figure 4A) without changing anything else in our sample pool or analysis workflow. In result, 62% of non‐movers were identified within morphotype 1 cluster (with the remaining 38% falling into the morphotype 2 cluster). A 100% of displacing bots were identified in morphotype clusters 2 and 3, with ≈85% of linear bots being in cluster 2, and 88% of circular bots being in cluster 3. We have further computed the statistical significance of these overlaps (using the Fisher test, see Materials and Methods) and conclude that the non‐movers, linear, and circulars significantly correspond with the morphotypes 1, 2 and 3, respectively.

The fact that none of the displacing bots were identified within the morphotype 1 cluster suggests that the movers identified within this cluster (i.e., those morphotype cluster 1 data points which are not labeled as “non‐movers”) are statistically likely to be non‐displacing movers, displaying a stationary wiggling motion. Accordingly, we conclude that morphotype 1 bots are likely to assume either non‐mover behavior or wiggler behavior. This may be attributable to their spherical shape with homogeneously distributed cilia where the propulsion forces generated by the ciliary motion are more prone to canceling each other out due to the radially symmetric spherical shape, resulting in little or no movement. Accordingly, inherent noise in the system (such as small imbalances in the cilia distribution on the spheroid surface or how the bot happened to be oriented in the plate) may be sufficient to have these bots generate small amounts of movement, causing them to wiggle, but not enough movement to become a displacing bot.