Mitotic Shape Changes Drive Spatiotemporally Segregated Cell-Contact Rearrangements in a Proliferating Epithelium

SUMMARY

In animals, epithelia tissue morphology is determined by the shape, position, and number of its component cells. Nevertheless, it remains incompletely understood how the shape transformations that accompany epithelial mitosis influence tissue packing. Combining an in silico vertex model with live imaging of murine intestinal organoids, we show that mitotic shape changes drive a spatiotemporally segregated sequence of cell-contact rearrangements: basal diminution of mitotic cell, transient multicellular rosette assembly, and reinsertion of the daughter cells. Strikingly, the reinsertion is accompanied by neighbor exchanges oriented perpendicular to the daughter–daughter contact in silico and in organoids, indicating that reinsertion geometry is sufficient to set the direction of rearrangement. Pharmacological arrest of mitotic progression reveals that basal diminution and daughter reinsertion independently drive distinct contact rearrangements. Together, our results establish that two distinct geometric asymmetries of mitosis, the apico-basal diminution of the dividing cell and the in-plane anisotropy of daughter reinsertion, drive a program of cell rearrangement that reorganizes basal cell-contact topology, providing an intrinsic route by which proliferating epithelia accommodate new cells.

Graphical Abstract

Blurb:

Ramanathan et al. show that the shape changes during mitosis actively reorganize epithelial packing. Withdrawal of the dividing cell from the basal region and the geometry of daughter reinsertion rearrange neighboring cells and their junctions, providing an intrinsic route by which proliferating epithelia accommodate new cells.

INTRODUCTION

Organogenesis requires precise control of cell packing and cell numbers, which are governed by cell position and proliferation, respectively. Being tightly organized and largely immotile, epithelial cells depend on distinct cell contact transformations and consequent neighbor exchanges to reposition individual cells^1–3. Distinct modes of epithelial cell reorganization are driven by characteristic contact transformations. For instance, neighbor cell exchanges are achieved by contracting existing cell contacts and extending new contacts (T1 transitions and rosettes), whereas cell shrinkage and elimination are accommodated by simultaneous contraction of all cell’s edges (T2 transitions)^1,4,5. Neighbor exchanges through T1 transitions are a fundamental driver of epithelial morphogenesis, mediating tissue elongation, remodeling, and the accommodation of new cells as tissues grow⁶.

New cells can be incorporated either by establishing de novo contacts between daughter cells after division or by forming multiple new contacts during radial migration events such as apical emergence^7–9. Most models of cell contact dynamics have been limited to analyzing the differential tension and adhesion in apical junctions and assume that the apical cell-contact architecture is recapitulated basally¹⁰. However, more recent work shows that cell contact topology and dynamics are strongly influenced by cell and tissue geometry^11,12. Consequently, differences in cell morphology between the apical and basal sides can give rise to distinct cell contact behaviors at each surface¹³. Nevertheless, an understanding of basal cell contacts and cell positioning remains limited, especially in proliferating tissues.

Mitotic cells adopt an asymmetric position along the apico-basal axis of epithelia. At mitotic entry, cells undergo a dramatic morphological transformation whereby their cellular mass translocates apically and the cell assumes a spherical shape¹⁴. This transformation leads to a near-complete loss of basal volume, a phenomenon we refer to as basal diminution. This mitotic shape change is evolutionarily conserved across proliferating epithelia¹⁵ and influences both development and disease in diverse tissue types^16–26. Furthermore, the geometry of nascent apical junctions between daughter cells directly influences epithelial cell packing to ensure a consistent cell shape distribution within tissues^7,27, and can also influence cell rearrangement^24,28,29. More recently, cell division and embryonic tissue compaction have been associated with apico-basal cell intercalations that accommodate newly formed cells in the proliferating sea star embryo³⁰, implicating proliferation in three-dimensional epithelial reorganization. Together, these observations highlight the well-established role of mitotic cell shape and cell-contact topology in development and disease. However, the cell-contact dynamics through which the mitotic shape transformations influence basal epithelial topology remain largely unknown.

RESULTS

Vertex model simulations of epithelial mitosis predict differential apico-basal junction topology accompanied by neighbor exchanges

The mechanical behaviors of mitotic rounding and apical contact remodeling are well characterized by force measurements and modeling^7,27,31,32, but the corresponding dynamics in the basal plane remain poorly understood. We therefore used 2D vertex-model simulations to contrast basal and apical contact dynamics during division^33–35 (Figure 1A). We first modeled apical division conventionally, by topological bifurcation of the mitotic cell (Figure S1A)^36,37. Across four regimes spanning the rigidity transition to fluid tissue (preferred perimeters $p_i^0=26.16-29.35\mu \text{m}$, realized shape indices ~3.81–4.15^38,39), bifurcation produced only a low baseline of neighbor exchange (0.66 T1s/mitosis on average; Figure S1B, C). Because mitotic cells in many epithelia round and expand apically¹⁵, we next doubled the mitotic-cell area at mitotic entry, approximating this footprint expansion (Figure S1D); this raised neighbor exchange events ~2.3-fold above baseline (Figure S1E, F). Mitotic apical area can expand fourfold in thickened pseudostratified epithelia, so area change likely contributes more than this conservative simulation indicates. However, faithfully modeling such highly rounded cells, with curved rather than straight interfaces, is beyond the current vertex model. Conversely, halving the mitotic-cell area, approximating the smaller cross-section of squamous epithelia or in 2D culture, produced a smaller increase in neighbor exchanges than area doubling (Figure S1G–I). These results indicate that in silico models of apical division that ignore cell geometry are likely to underrepresent neighbor exchanges.

Figure 1. Vertex-model simulations of mitosis predict basal contact remodeling and neighbor exchanges.

(A) Vertex model in which cells are polygons defined by vertices (green) and straight edges (cyan). Tissue mechanics follow an energy function, $W$, with area ($K_a$) and perimeter ($K_p$) elasticity about preferred values $a_i^0$ and $p_i^0$; vertices evolve to minimize $W$.

(B) Basal division simulated in two phases: (1) diminution, a gradual reduction in the mitotic cell’s preferred area and perimeter, removing it from the basal plane; and (2) reinsertion of two daughters at half the mother’s preferred area.

(C) Snapshots of basal contact rearrangements. The mitotic cell (0°) is white; neighbors (1°– 6°) are colored by topological proximity ($p_i^0=26.16\mu \text{m}$). The asterisk marks a representative neighbor exchange (a 2° cell intercalating between two 1° cells), and the thick line marks the new cell–cell contact formed during the exchange.

(D) Temporal distribution of T1 transitions per mitosis (all neighbor orders, excluding the mitotic cell and daughters; $n=324$ divisions); negative and positive times denote diminution and reinsertion.

(E) T1 transitions per mitosis versus topological distance from the mitotic cell (plotting convention in Methods).

See also, Figure S1.

Next, we sought to determine the influence of mitotic cell shape dynamics on cell contact topology at the basal plane. To simulate mitosis at the basal plane, cells were progressively removed and subsequently reinserted as two co-equal daughter cells (Figure 1B). Strikingly, the removal and reinsertion of mitotic cells frequently triggered basal neighbor exchanges and local cell mixing (Figures 1C and S1J). Notably, 92.4% of T1 transitions, across all neighbor orders, excluding the mitotic cell and its daughters, occurred during daughter cell reinsertion, rather than during the initial basal diminution phase (Figure 1D). Furthermore, the spatial distribution of mitosis-driven T1s within the tissue was heavily skewed toward cell contacts closer to the mitotic cell (Figure 1E). Compared to topological bifurcation, basal division elevated T1 transitions ~5.4-fold above baseline (to 3.58 T1s/mitosis) across both solid-like and floppy tissue regimes (Figure S1B, C vs. S1K, L). In all cases, the basal T1 transitions concentrated around daughter cell reinsertion in both time (0.16 vs. 1.95 T1 transitions/mitosis during diminution vs. reinsertion; Figure 1D) and space (1.91 T1 transitions/mitosis among the immediate neighbors; Figure 1E). Unlike previous in silico models that simulate mitotic division by bifurcating a single mother cell, our approach captures the unique topology of basal division, predicting sequential events: basal cell diminution/elimination (T2-like transitions), daughter cell reinsertion, and mitosis-driven neighbor exchanges (T1s). Collectively, these results reveal that apical and basal mitotic cell geometry engage cell-contact topology differently. This difference prompted us to validate the model in living tissue.

Mitotic cell morphology and dynamics in intestinal organoid epithelium

In most tissues, the basal surface of a polarized epithelium is obscured by cells or a thick basement membrane, making basal cell dynamics difficult to image. Mouse intestinal organoids overcome this: the apical plane faces the interior lumen (anti-ZO-1; Figure S2A), leaving the basal surface exposed and accessible (anti-Laminin; Figure S2B). Apical displacement and rounding of mitotic cells followed by basal reinsertion of daughters, a hallmark mitotic behavior of intact intestine^16,40, were recapitulated in organoids (Figure S2C). Fixed and live imaging confirmed a columnar epithelial architecture with interphase nuclei near the basal surface (Figure S2C, 2A), from which mitotic cells translocated apically at nuclear-envelope breakdown. After division, the daughters returned to the basal domain (Figure 2A). Staining for the proliferation marker Ki-67 showed apically translocating mitotic cells throughout the epithelium (Figure 2B, top). We restricted analysis to regions where basal nuclei lay in a single plane, excluding the high-curvature crypt base and its resident crypt-base columnar and Paneth cells⁴¹. In these regions, nuclei (outlined by Lamin A) were almost uniformly Ki-67-positive, indicating that the analyzed domains comprise proliferative, rapidly dividing transit-amplifying cells and lack terminally differentiated types (Figure 2B, bottom). These flat, cylinder-like and sphere-like domains permit basal junctions to be tracked at high spatiotemporal resolution, establishing organoids as a tractable model for basal topology during mitosis.

Figure 2. Basal diminution and daughter reinsertion coincide with cell shape changes, basal rosettes, and neighbor exchanges in vivo.

(A) Cross-sectional timelapse of organoid epithelium showing a mitotic cell (‘M’) and daughters (‘D’) through diminution, rounding, cytokinesis, and reinsertion; arrowheads mark diminution and reinsertion sites. Cell membranes: mTomato (green); nuclei: H2B-GFP (magenta). Scale bar, $10\mu \text{m}$.

(B) Fixed tissue stained for Ki-67 (green) and Lamin A (magenta). Top: cross-section; arrows mark mitotic bodies repositioning apically. Bottom: basal view; dashed outlines mark flat cylinder-like (yellow) and sphere-like (cyan) domains used for junction tracking. Scale bar, $20\mu \text{m}$.

(C) Basal cell-contact topology of a mitotic cell (‘M’), its 1° neighbors (‘1–8’), and daughters (‘D’); Cells 1 and 2 lose contact as the 2° neighbor 9 intercalates between them forming the new cell-cell contact marked in red. Yellow circle, a transient five-cell rosette. Scale bar, $10\mu \text{m}$.

(D) Rosette initiation at basal vs. apical regions ($n=33$ mitoses).

(E) Basal outcomes: Unchanged topology, daughter separation, or local exchange ($n=33$ mitoses); exchange in 58% of mitoses (0.70 T1s/mitosis).

(F) Daughter-separation initiation at basal vs. apical regions.

(G) First-order neighbor exchanges during diminution vs. reinsertion.

(H) T1 initiation at basal vs. apical regions.

(I) Basal T1 transitions in cylinder-like ($n=16$) vs. sphere-like ($n=16$) regions.

(J) Circularity of the mother (late interphase) vs. combined daughters (**:$p=1.2\times {10}^{-6}$).

(K) Schematic of the angle between the daughter–daughter contact (orange) and the new neighbor contact (teal).

(L) Polar histogram of that angle; radial line, median.

See also, Figure S2, Figure S3, and Figure S4.

Cell-contact transformations during mitosis in organoid epithelia

We next imaged dividing cells to investigate mitotic topological dynamics at the basal epithelial surface. At mitotic entry, the basal cell-contact dynamics resembled T2-like events, characterized by the progressive diminution and eventual withdrawal of the mitotic cell from the basal plane (Figure 2C). This was frequently (>90% of divisions) accompanied by the formation of transient, basally-initiated multicellular rosette-like contact configurations (Figure 2D, $n=33$ cell divisions). In most cases (67% of 33 cell divisions, Figure 2E), the subsequent reinsertion of daughter cells was accompanied by either daughter cell separation (9%) or neighbor exchanges and local cell mixing (58%). Across all divisions, neighbor exchange occurred at a rate of 0.70 T1 transitions per mitosis (23 transitions in 33 mitoses). Consistent with previous reports^42,43, daughter cells occasionally separated during reinsertion (Figure S3, Figure 2E). As with rosette assembly, daughter cell separation was basally initiated (Figure S3, Figure 2F). Interestingly, the majority of daughter cell reinsertion events were associated with contact exchanges between cells in the immediate neighborhood (Figure 2C, E). Furthermore, 91% of first-order (1°) neighbor exchanges occurred during daughter cell reinsertion, rather than basal diminution (Figure 2G). The neighbor exchange first registered at the apical surface, coincident with cytokinesis, and propagated to the basal plane at later times (Figure 2C, H). The rosettes accompany basal diminution, whereas this neighbor mixing was temporally coupled to daughter cell reinsertion (Figure 2D, G, H).

We next asked whether the mitosis-associated neighbor exchanges depend on tissue curvature. In silico and in vivo studies report apico-basally distinct cell-contact topology in cylinder-like epithelial domains, which possess unequal principal curvatures. In contrast, sphere-like domains tend to have matched apical and basal topology¹². Comparing the relatively flat basal domains of cylinder-like (crypt) and sphere-like (body) regions (Figure 2B, bottom), we found that basal T1 transitions were more frequent in sphere-like (75%) than cylinder-like (56%) regions (Figure 2I, $\text{n}=32$). We interpret this difference as curvature modulated the apical-to-basal propagation of mitosis-associated rearrangements: in sphere-like regions, exchanges propagate more completely to the basal plane and register as basal T1 transitions, whereas in cylinder-like, anisotropically curved regions, propagation to the base is less complete, producing fewer basal T1 transitions. This curvature-dependence is consistent with prior work showing that cell-contact topology differs between apical and basal surfaces in anisotropically curved epithelia¹².

Next, we examined the morphological differences between the mitotic cell eliminated from the basal plane and the combined shape of reinserting daughter cells. While the combined basal area of daughter cells was comparable to that of their mitotic mother in late interphase (Figure S4A), we observed a consistent increase in the basal perimeter of reinserted daughter cells relative to that measured in the corresponding late interphase cells (Figure S4B). Indeed, the circularity of the two daughters combined was 30% lower than that of the progenitor in late interphase, indicating that their combined shape was more elongated (Figure 2J). Interestingly, changes in contact topology and neighbor exchanges were more likely to occur perpendicular to the shared junction between the daughter cells (Figure 2K, L).

Together, these observations reveal that at the basal plane, mitosis involves a sequence of morphological and topological changes: basally initiated cell body diminution (T2-like transitions), apical division, and basal reinsertion of daughter cells accompanied by apico-basal neighbor exchanges (T1 transitions). These findings closely align with predictions from our vertex model simulations (Figure 1). Additionally, basally initiated rosette-like contact patterns were a consistent feature of divisions in vivo but were not observed in silico.

Incomplete elimination of mitotic cells from the basal plane generates rosette-like cell contacts in silico

Multicellular rosettes are widely observed apical structures that bring together four or more epithelial cells at a single vertex, typically by elevating contractility at the adherens junctions⁴. These configurations are inherently unstable and serve as transient intermediate topological arrangements in a range of organs and organisms^44,45. Intriguingly, we observed rosette-like cell assemblages at the basal regions of proliferating organoid epithelia (Figure 2C, D). In diverse tissues, including the mammalian intestine, apically displaced mitotic cells often retain attachment to the basement membrane through thin, elongated cellular tethers^14,16,42,46. To explore the potential contribution of mitotic cell tethers in stabilizing basal rosettes and inducing neighbor exchanges, we used vertex model simulations. Specifically, we modeled a mitotic cell that was incompletely eliminated from the basal plane, leaving a residual basal footprint representing an in vivo tether (Figure 3A, B). When mitotic cell remnants were vanishingly small (one-millionth the original area), basal contact dynamics were indistinguishable from those observed with complete elimination. However, when larger remnants (one-hundredth the original area) were retained, rosette-like arrangements of neighboring cells consistently observed (Figure 3B). Thus, rosettes can arise passively from basal mitotic remnants without elevated junctional contractility, recapitulating what we observed in organoids. Notably, the presence of these rosettes did not lead to increased neighbor exchanges (Figure 3B, C). In fact, simulations across a range of mechanical regimes, spanning solid-like to floppy tissues ($p_i^0=26.16\mu \text{m},27.22\mu \text{m},28.28\mu \text{m}$, and $29.35\mu \text{m}$), consistently showed a reduction in T1 transitions during daughter cell reinsertion when rosettes were present (Figure 3D). These results suggest that basal rosette formation suppresses, rather than promotes, neighbor exchanges. This in silico observation contrasts with the established role of apical rosettes as drivers of large-scale intercalation in other systems, such as the Drosophila germband⁴, and likely reflects the distinct, passive origin of these basal rosettes; their functional role in neighbor exchange remains to be tested in vivo.

Figure 3. Simulated mitotic tethers generate basal rosettes but suppress contact rearrangements.

(A) A dividing cell (white) in a tissue with $p_i^0=27.22\mu m$.

(B) Complete ($a^0=0$) or partial (${\text{a}}^0=5\times {10}^{-5},5\times {10}^{-1}\mu m^2$) displacement of the mitotic cell from the basal plane; incomplete diminution ($5\times {10}^{-1}$) produces rosette-like structures (teal circle).

(C) Daughter reinsertion; magenta asterisks mark exchanges separating 1° neighbors (red).

(D) T1 transitions per division during diminution and reinsertion, with complete elimination (no rosette) vs. incomplete elimination (~1% remnant, rosette-forming), across $p_i^0=26.16\mu m-29.35\mu m$; fewer T1s occur in the rosette-forming condition ($n=324$ divisions per condition).

Daughter cell reinsertion drives oriented neighbor displacement in silico

Having established that basal diminution and rosette formation are temporally separated from the majority of neighbor exchanges, we next examined the role of daughter cell insertion in influencing cell contact topology. Both in silico simulations and live imaging of organoid epithelia revealed that most neighbor exchanges were temporally coupled with the reinsertion of post-mitotic daughter cells into the basal plane (Figure 1D; Figure 2C, G). Moreover, these exchanges were predominantly oriented orthogonally to the newly formed cell contact between daughter cells (Figure 2K, L). To test whether reinsertion orientation influenced the directionality of neighbor exchanges, we analyzed vertex model simulations across tissues with varying mechanical properties, spanning both solid-like and floppy regimes ($p_i^0=26.16\mu \text{m},27.22\mu \text{m},28.28\mu \text{m}$, and $29.35\mu \text{m}$; Figure 4A–C). Across all simulations, fewer than 8% of neighbor exchanges occurred within 30° of the axis aligned with the nascent daughter cell contact, confirming a strong bias for orthogonal reorganization. Importantly, this bias was reproduced both in silico and in organoids (Figure 4C), linking the geometry of cell division to the direction of the resulting neighbor exchange.

Figure 4. Daughter reinsertion generates anisotropic displacements and orients neighbor exchange in silico.

(A) Schematic of the angle between the daughter–daughter contact (blue) and the new T1 contact (orange).

(B) Example in a simulated tissue ($p_i^0=27.22\mu \text{m}$): after reinsertion, two exchange events form angles of 53.9° and 16.7° with the daughter–daughter contact.

(C) Orientation of T1 contacts in silico ($p_i^0=26.16\mu m-29.35\mu m$) and in organoids (grey); $n=324$ and 22 divisions, respectively.

(D, E) Basal-plane centroid displacement fields during (D) diminution and (E) reinsertion; arrow length and direction give displacement magnitude and orientation. Blue segment, collinear with the daughter–daughter contact.

To investigate the geometric basis of this correlation, we quantified the displacement of cell centroids during the mitotic cycle. Starting from a defect-free honeycomb lattice, we tracked centroid displacements during basal diminution and subsequent reinsertion. During diminution, displacements were greatest in the immediate vicinity of the mitotic cell but remained isotropic in orientation (Figure 4D). In contrast, reinsertion of daughter cells generated anisotropic displacements, oriented perpendicular to the cell contact shared by the daughters (Figure 4E). We therefore conclude that the orientation of neighbor exchanges arises passively from reinsertion geometry alone without additional active or directed force.

Basal neighbor exchanges in proliferating organoid epithelia require daughter reinsertion

Having shown that daughter reinsertion correlated spatially and temporally with T1 transitions, we next tested whether it was required. We blocked mitotic cells from returning to the basal plane using S-trityl-L-cysteine (STC), an inhibitor of the spindle motor kinesin-5 that arrests cells in mitosis⁴⁷. Critically, prior atomic force measurements show that STC-arrested cells retain the cortical mechanics of normally rounding cells, without confounding cytoskeletal perturbation^48,32. In organoids, STC held mitotic bodies apically, depleting them indefinitely from the basal plane while they retained their rounded shape for >1 h (Figure 5A–C).

Figure 5. Blocking daughter reinsertion abolishes neighbor exchange in organoid epithelium.

(A) Timelapse of an STC-treated organoid; cell that entered mitosis (‘M’) at 0, 56, 112, and 232 min are labeled red, green, magenta, and cyan. Membrane, mTomato (yellow); all scale bars, $10\mu \text{m}$.

(B) Circularity of apical mitotic cells at 0 and 60 min after rounding ($n=15$).

(C) Basal topology of an arrested mitotic cell (‘M’) and neighbors (‘1–7’); red asterisk, basal rosette.

(D) Circularity of untreated (blue) vs. STC-treated (red) cells before mitosis, during diminution, and after reinsertion; STC data end at late interphase, as arrested cells do not reinsert.

(E, F) Circularity (E) and neighbor number (F) of untreated (blue) and STC-treated (red) pre-mitotic cells and their neighbors ($n=31$ untreated, 23 STC-treated).

(G) Local neighbor-exchange frequency in untreated vs. STC-treated organoids.

See also, Figure S4 and Figure S5.

STC treatment did not perturb the surrounding epithelium. Late-interphase cell and neighbor shapes, overall crypt and body morphology, and neighbor numbers were all unchanged (Figure 5D–F, Figure S5A). Together with the spindle-specific action of STC and the prior single-cell mechanical measurements, this indicates that geometry and packing are preserved under treatment, though tissue-level mechanics were not measured directly. Under these conditions, blocking reinsertion reduced first-order neighbor exchange by 93%, even as transient basal rosettes still formed during apical rounding (Figure 5C, G). We thus infer that daughter reinsertion is required for neighbor exchange in proliferating intestinal epithelia. These results agree closely with our simulations, indicating that basal diminution drives T2-like elimination and transient rosettes while daughter reinsertion drives apico-basal neighbor exchange (T1 transitions).

DISCUSSION

Tissue geometry and mechanics are well known to influence mitotic progression and daughter cell dynamics in epithelia: neighbor topology can orient the mitotic cleavage plane across a wide variety of systems, including plants and animals²⁷. Similarly, local mechanical constraints can determine whether daughter cells remain adjacent or are separated during reinsertion, as demonstrated in the chick epiblast²⁸ and mammalian intestinal organoids^42,43. Our findings demonstrate that the converse is also true: beyond the well-established topological consequences of proliferation for epithelial polygon distributions²⁷, the geometric transformation of the mitotic cell itself actively drives a sequence of cell-contact rearrangements that remodel basal epithelial topology. Mitotic cells trigger a sequence of transformations in their neighbors, including basally initiated T2-like diminution, basally initiated rosette formation, and apically initiated T1 neighbor exchanges. Most neighbor exchanges occur not during the early diminution phase but during the basal reinsertion of daughters, marked by the basal emergence of anisotropic daughter cell shapes, accompanied by anisotropic displacement of neighboring cells oriented perpendicular to the daughter–daughter contact in silico. Our simulations show that geometric asymmetry, basal diminution of the mitotic cell and anisotropic daughter cell insertion within the plane of the epithelium, is sufficient to drive local packing transformations, while the apical-to-basal propagation linking them is resolved only in vivo. Mitosis is thus not a topologically passive event that merely yields two daughters, but a geometrically active process that sculpts epithelial topology.

Both the direction and the timing of these events distinguish their origins. The apical neighbor exchange coincides with the apical mechanical events of mitosis, namely rounding and cytokinesis, and propagates apical-to-basal, reaching the basal plane only later (Figure 2C, H). The following events are plausible drivers: cytokinetic ring contraction exerts force on neighboring cells⁴⁹ and mitotic rounding generates substantial force³² to alter cell cross-sectional area, and therefore, epithelial topology⁵⁰. The rosettes and daughter separations, by contrast, are decoupled from rounding and cytokinesis in both space and time: they initiate basally and propagate apically, outside the apical-event window, with rosettes forming earlier, during basal diminution, and separations later, during daughter reinsertion. These basally initiating events are therefore unlikely to be driven by rounding or cytokinesis. The daughter separations propagate in the opposite direction to the co-occurring reinsertion exchanges and reflect low cortical contractility rather than tissue geometry²⁸, indicating a distinct origin. Propagation direction alone cannot localize a driving force, since rearrangements can travel either way along the apico-basal axis^12,28,51,52; however, direction and timing together can.

To further characterize the driver of the basal rearrangement, we used the vertex model, which isolates each (apical vs. basal) perturbation. Although it is a 2D reduction of the organoid’s three-dimensional tissue, the model reproduces the topology and orientation of mitosis-associated T1 transitions using only passive area and perimeter elasticity, with no active contractile or cytokinetic forces. The measured organoid shape index (~3.90–4.01; Figure S5D) falls within the simulated range of realized shape indices (~3.81–4.15), indicating that the regimes we explore are physiologically relevant and that the basal rearrangements do not depend on a particular fluidity. Modeled separately, apical area expansion (approximating rounding) and topological division produced neighbor exchanges at ~2.3- and ~5.4-fold lower frequency, respectively, than basal division (Figure S1). The decisive evidence is orientation: reinsertion of anisotropic daughters reproduced the basal rearrangement perpendicular to the daughter–daughter contact, with an orientation distribution closely matching organoid (Figure 4C). This directional bias arises from the anisotropy of daughter reinsertion and cannot arise from apical rounding, which is symmetric. This division-axis orientation is consistent with independent observations that dividing cells deform their neighbors along the division axis across diverse epithelia⁴⁹; at the basal plane, we find that this oriented deformation resolves into topological neighbor exchange. Our simulations find that the oriented rearrangement emerges from passive relaxation around the reinserted daughters without additional active force, and that the reinsertion geometry is sufficient to account for it. We propose that, while the apical mechanical events of mitosis may contribute to the apical exchange as discussed above, the basal rearrangement is driven predominantly by daughter reinsertion geometry. A synergistic in vivo contribution cannot be excluded and merits future force-resolved study; we anticipate it would be context-specific, whereas the reinsertion-driven basal rearrangement should be broadly conserved.

The prominence of these basal rearrangements is expected to vary across epithelial architectures. Basal diminution is conserved across columnar epithelia¹⁵, but in squamous and cuboidal epithelia, where mitotic cells present a smaller basal cross-section, the rearrangements should be less pronounced. This is consistent with our simulations, in which reducing the mitotic cell cross-section lowered neighbor exchange ~1.9-fold relative to apical area expansion (Figure S1G–I vs. S1D–F). In pseudostratified epithelia, wherein the nuclear diameter is typically greater than average cell width, nuclear dynamics would also shape the outcome.

Our findings connect to recent work on division-associated three-dimensional packing. In the sea star embryo, apico-basal intercalations form shortly after division to accommodate new cells, but cell density and geometry explain only part of them, implicating a division-associated factor that static models did not capture³⁰. Our results identify a cell-biological basis for that factor: the mitotic shape transformation drives the basal neighbor exchanges that reorganize local packing, with reinsertion geometry as a proximate driver. Tissue curvature further shapes this process with basal T1 transitions being more frequent in sphere-like than cylinder-like regions (Figure 2I). We interpret this as more complete apical-to-basal propagation in sphere-like domains, consistent with reports that apical and basal topology diverge in anisotropically curved epithelia¹².

Neighbor exchange is central to how epithelia build and remodel tissue: beyond driving convergent extension¹, it allows epithelia to flow while maintaining barrier integrity^6,53 and to relieve the packing constraints that accumulate as cells proliferate⁷. In a growing epithelium, each division adds a cell the tissue must accommodate, and our findings indicate that mitosis itself meets this demand and reorganizes local packing. Although we analyze individual divisions, such rearrangements may accumulate over successive divisions to remodel steady-state packing, as they do in the compacting sea star embryo³⁰. However, demonstrating a cumulative effect that will require tracking topology across division rounds. Whether a given mitosis remodels basal topology depends on the local geometry of reinsertion and on whether transient rearrangements stabilize, making mitosis-driven neighbor exchange a frequent but conditional, rather than obligatory, route to basal remodeling. Mitosis-driven neighbor exchange may thus be an intrinsic, broadly conserved route by which proliferating epithelia absorb new cells and remodel their architecture. This work extends the established link between proliferation and epithelial topology⁷ from apical polygon statistics to the geometry of basal cell-contact rearrangement.

We find that rosette formation, a hallmark of apical remodeling, emerges transiently at the basal surface during mitosis. Multicellular rosettes organize morphogenesis across species and can initiate basally⁵¹. Although the precise initiation mechanisms remain unclear, mitotic cell shape changes are increasingly implicated in triggering rosette formation^54–56. However, in contrast to the contractility-driven apical rosettes that drive intercalation in convergent extension⁴, basal rosettes stabilized by incomplete cell elimination did not facilitate neighbor exchanges and tended to reduce them.

Several questions remain open. A central frontier concerns the tissue’s three-dimensional, curved geometry. Our analysis was restricted to relatively flat basal domains where junctions can be reliably segmented; resolving the highly curved basal junctions of the crypt–villus axis, where apico-basal effects are likely most pronounced, will require advances in volumetric imaging and new segmentation methods for live tissue. The organoid’s three curvature domains offer a way to study curvature’s influence on packing within a single system once such methods are available.

Our model takes a step toward geometry-aware modeling of division, incorporating the mitotic cell’s area change during rounding and diminution rather than treating division as a purely topological bifurcation. Numerous available frameworks for simulating epithelia include Voronoi tessellations^57,58, cellular Potts models^59–61, phase-field models⁶², and particle-based methods^63,64. However, the vertex model is the natural choice for the mechanically driven topological rearrangements we study, representing cell–cell junctions and their contraction explicitly and capturing T1 transitions and division directly^36,65. Extending it to the organoid’s three-dimensional, curved geometry, however, is a two-front challenge. Representing the curved tissue is feasible with 3D vertex models, but representing a rounded mitotic cell with convex cell contacts within such a tissue is beyond current straight-edged approaches. More broadly, a complete description of mitosis-associated remodeling requires three ingredients that no single framework yet combines: curved interfaces, active division, and apico-basal polarity.

Our study resolves the topology and geometry of cell contact rearrangements in mitosis; however, their underlying mechanics remain to be measured. Actomyosin and adhesion force-resolved measurement is complicated by cytoskeletal inhibitors that perturb both mitotic mechanics and baseline tissue properties, and attributing causality to apically initiated T1 transitions is difficult given likely long-range effects in a proliferating tissue. How local mechanics, basement-membrane properties, and epithelial subtype shape cell contact dynamics, and whether the rearrangements we describe are general or specific to geometrically constrained tissues such as organoids, testable against a flat epithelial system, all remain to be addressed. In conclusion, mitosis is a potent driver of topological transformation at the basal epithelial surface, positioning proliferation as not merely additive but morphogenetic. Revisiting other cellular transitions through this geometric and topological lens may clarify how tissues self-organize during development, regeneration, and disease.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Subramanian P. Ramanathan (sramanathan@unmc.edu).

Materials availability

This study did not generate new, unique reagents. All biological materials, including mouse lines, are commercially available.

Data and code availability

• Due to the multi-terabyte scale of the confocal image stacks, the imaging datasets have not been deposited in a public repository. The complete raw multi-terabyte datasets are securely archived on institutional servers and are available from the lead contact upon request.

• Original code underlying these simulations is archived as a citable static release on Zenodo (https://doi.org/10.5281/zenodo.22236021) and is publicly available on GitLab (https://gitlab.com/tanmoy_jsi/cell_diminution_protocols).

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR METHODS

EXPERIMENTAL MODEL AND SUBJECT DETAILS

In silico vertex model

A two-dimensional vertex model was used as the computational experimental system, representing the epithelial monolayer as a confluent tiling of polygonal cells whose mechanics derive from area and perimeter elasticity.

Mouse intestinal organoid model

Mouse small-intestinal organoids served as the experimental tissue model for all imaging and perturbation experiments. Organoids were derived from murine duodenal crypts of both male and female mice at 6–12 weeks. Organoids were generated from reporter mouse lines: mT/mG (JAX: 007676) for membrane labeling, crossed with R26-M2rtTA; TetOP-H2B-GFP (JAX: 016836) for nuclear labeling. All mouse work was approved by the IACUC of the Stowers Institute (2017–0179) and UNMC (22-078-11-EP).

METHOD DETAILS

Vertex model

In two-dimensional (2D) vertex models, tissues are represented as planar tilings of polygonal cells, where tissue mechanics emerge from the geometry and connectivity of the cellular network. The formulation used here follows the standard vertex model framework and relies on the following assumptions^33,34,66.

First, the epithelium is assumed confluent: cells tile the tissue without gaps or overlaps, so the packing fraction is unity. Consequently, every edge is shared by two neighboring cells, and every vertex is a junction of three or more cells. Under this assumption, the geometry of the tissue is completely determined by the positions of such vertices, which also define the shape of each cell. Additionally, it has been shown that tissue mechanics are governed by cell shape and topology rather than by density³⁸.

Second, the tissue is assumed to be well approximated by a reduced two-dimensional (2D) description, appropriate when cells have approximately uniform height³³.

Third, the cellular cytoplasm is treated as effectively incompressible on the timescales relevant to tissue remodeling⁶⁷. Together with the approximation of constant cell height, this implies that the cell area fluctuates around a preferred value, represented by an area-elasticity term in the potential energy (1^st term, Equation 2).

Fourth, cell-boundary mechanics arise from the competition between cortical actomyosin contractility, which tends to minimize the cell perimeter, and cell–cell adhesion, which stabilizes intercellular contacts. In the vertex model, this competition is represented phenomenologically by a perimeter-elasticity term, with the preferred perimeter serving as an effective parameter that captures the balance between cortical tension and adhesion (2^nd term in Equation 2).

Finally, the dynamics of the constituent vertices are assumed to be overdamped, where friction dominates inertia. Therefore, vertex motion is governed by the balance between the mechanical forces derived from the potential energy (Equation 2) and viscous dissipation (Equation 1). Neighbor exchanges (T1 transitions) and cell division are implemented through standard topological update rules that preserve confluence.

The shape of each polygonal cell is parametrized by the location of its vertices ${\mathbf{r}}_{\mathbf{v}}=\left(x_v,y_v\right)$, where $v$ stands for the $v$-th vertex of the system. Any cellular scale dynamics, such as deformations or displacements, are captured through the motion of these vertices:

$$\eta \frac{d{\mathbf{r}}_{\mathbf{v}}}{d\mathbf{t}}={\mathbf{F}}_{\mathbf{v}}$$ (Equation 1)

where $\eta$ is the friction coefficient of vertices against the substrate, and $F_v$ is the net force acting on the vertex $v$. The force on a vertex is derived from a potential energy that always draws the system towards its local minima. The potential energy contains contributions from area and perimeter elasticity as follows

$$W={\sum }_{i\in \text{cells}}\left[K_{a_i}{\left(a_i-a_i^0\right)}^2+K_{p_i}{\left(p_i-p_i^0\right)}^2\right].$$ (Equation 2)

Here, $a_i$ and $p_i$ are the actual area and perimeter of the $i$-th cell, $a_i^0$ and $p_i^0$ are the corresponding preferred values, and $K_{a_i}$ and $K_{p_i}$ are the moduli associated with area and perimeter elasticity, respectively. The force on each vertex is given by ${\mathit{F}}_v=-{\nabla }_v\text{W}$, where ${\nabla }_v=\left(\frac{\partial }{\partial x_v},\frac{\partial }{\partial y_v}\right)$. We consider that $a_0$ is the preferred area of any normal cell, and $a_0^{1/2}$ is the unit of length. Additionally, we assume identical elastic properties for all cells, i.e., $K_{a_i}=K_a$ and $K_{p_i}=K_p$.

We nondimensionalize the system by defining dimensionless variables: $\frac{a_i}{a_0}\to a_i,\frac{p_i}{a_0^{\frac{1}{2}}}\to p_i,\frac{{\mathbf{r}}_{\mathbf{v}}}{a_0^{\frac{1}{2}}}={\mathbf{r}}_{\mathbf{v}}$, and $\frac{a_i^0}{a_0}\to a_i^0$. The resulting dimensionless equation of motion for the $v$-th vertex reads

$$\frac{d{\mathbf{r}}_{\mathbf{v}}}{dt}=-{\sum }_{i\in \text{cells}}\left[2k\left(a_i-a_i^0\right){\nabla }_v{a}_i+2\left(p_i-p_i^0\right){\nabla }_v{p}_i\right],$$ (Equation 3)

where $k=\frac{K_a{a}_0}{K_p}$ is the dimensionless area elasticity modulus, whereas ${\tau }_0=\frac{\eta }{K_p}$ is the characteristic time. Furthermore, we assume that the cells are incompressible and set $k=100$.

Using a forward finite-difference scheme, we use a time step $\Delta t={10}^{-3}$ to solve Eq. (3).

Initial configuration preparation of in silico tissue

We take a pre-generated disordered 2D tissue with $N_c=324$ equal-sized cells ($a_i^0=a^0=1$) under periodic boundary conditions. From this base configuration, we generate $N_c$ configurations by designating one arbitrarily chosen cell from each as a large mitotic cell with preferred area $a_{\text{mt}}^0~1.33$. To accommodate this increase, we adjust the system size accordingly.

Additionally, we choose four values for the preferred perimeters of normal cells: $p_i^0=p^0=3.70,3.85,4.00$, and 4.15, resulting in four distinct sets of configurations, each containing $N_c$ samples. These dimensionless values correspond to physical preferred cell areas ($a^0$) of approximately $50\mu m^2$, and perimeters of $26.16\mu m,27.22\mu m,28.28\mu m$, and $29.35\mu m$ corresponding to $p^0=3.70,3.85,4.00$, and 4.15, respectively.

Each configuration is then melted using an active-tension-fluctuations scheme, described by the Ornstein-Uhlenbeck model^68,69. The dynamics include a fluctuating line tension term added to Eq. (3):

$$\frac{d{\mathbf{r}}_{\mathbf{v}}}{dt}=-{\sum }_{i\in \text{cells}}\left[2k\left(a_i-a_i^0\right){\nabla }_v{a}_i+2\left(p_i-p_i^0\right){\nabla }_v{p}_i\right]-{\sum }_{j\in \text{cell-contacts}}\Delta {\gamma }_j(t){\nabla }_v{l}_j,$$ (Equation 4)

where $l_j$ is the length of the cellular contact $j$, and the fluctuating line tension on contact edge $j$ evolves according to the Ornstein-Uhlenbeck scheme as

$$\Delta {\dot{\gamma }}_j(t)=-\frac{1}{{\tau }_m}\Delta {\gamma }_j(t)+\sqrt{\frac{2{\sigma }^2}{{\tau }_m}}{\xi }_j(t).$$

Here, the noise $\xi (t)$ obeys $⟨{\xi }_j(t)⟩=0$ and $⟨{\xi }_j(t){\xi }_l\left(t^{\prime }\right)⟩={\delta }_{jl}\delta \left(t-t^{\prime }\right)$ with ${\tau }_m$ and ${\sigma }^2$ being its relaxation time scale and long-time variance, respectively.

We melt each configuration for 3000 time units using $\sigma =0.5$, which is followed by a sharp quench ($\sigma =0$) and minimization over an additional 7000 time units. During these two steps, we allow T1 transitions. After that, all four-fold vertices are resolved into small edges of length ${10}^{-3}$, followed by another 1000 time units of minimization without T1 transitions using a smaller integration time step $\Delta t={10}^{-4}$. Each relaxed configuration is then used to initialize a pair of simulations representing the corresponding apical and basal tissues, thereby preserving the one-to-one apico-basal correspondence between the two layers. Unless otherwise stated, all subsequent simulations are performed with $\sigma =0$.

Cell diminution and reinsertion in silico

Each prepared configuration is used to initialize a corresponding pair of basal and apical simulations. Both simulations follow the same sequence of steps: (i) initial energy minimization for $t_{\text{eq}}=300$ time units, (ii) gradual change in the preferred area of the mitotic cell over a duration $t_{\text{mod}}=100$ time units, (iii) post-modification energy minimization for 100 time units, (iv) cell division followed by a gradual change in the preferred areas of the daughter cells over a duration, $t_{\text{rg}}=100$, and (v) final energy minimization for an additional 300 time units.

The only difference between the basal and apical simulations lies in the choice of the target preferred area of the mitotic cell. We therefore introduce a generic target preferred area, $a_{\text{tar}}^0$, which takes different values for the two simulations.

During the modification stage, the preferred area of the mitotic cell evolves linearly according to

$$a_{\text{mt}}^0(t)=\left(1.33-a_{\text{tar}}^0\right)\frac{t_{\text{mod}}-t}{t_{\text{mod}}}+a_{tar}^0,$$

while its preferred perimeter is updated as $p_{\text{mt}}^0(i)=p^0\sqrt{a_{\text{mt}}^0}(t)$, where $p^0$ is the preferred perimeter of a normal cell.

After a subsequent relaxation period of 100 time units, the mitotic cell is divided along an axis passing through its centroid. For each pair of simulations, the apical and basal cells are divided along the same axis. The time of division defines the reinsertion time.

Following division, the preferred area of each daughter cell evolves linearly as $a_{\text{d}}^0(t)=\frac{a_{\mathit{\text{tar}}}^0}{2}+\frac{t}{t_{\text{rg}}}\left(\frac{1.33}{2}-\frac{a_{\text{tar}}^0}{2}\right)$, with the preferred perimeter updated according to $p_{\text{d}}^0(t)=p^0\sqrt{a_{\text{d}}^0}(t)$.

The values of $a_{\text{tar}}^0$ used in the two simulations are summarized below:

Simulation	Target preferred area ${\text{a}}_{\text{tar}}^0$	Interpretation
Basal	${\text{a}}_{\text{tar}}^0={\text{a}}_{\text{min}}^0\in \left\{0,{10}^{-4},{10}^{-6}\right\}$	The mitotic cell undergoes diminution prior to division, and the daughter cells subsequently grow to half the original mitotic-cell size.
Apical	${\text{a}}_{\text{tar}}^0={\text{a}}_{\text{max}}^0\in \{0.5\times 1.33,1.33,2\times 1.33\}$	Apical division is simulated at three mitotic-cell cross-sections approximating the reduced and expanded mitotic footprints of squamous and columnar epithelia, respectively, with the unchanged case as a no-area-change control. Daughters then relax to half the original mitotic-cell area.

Displacement profile calculation in silico

Mitotic activity-induced displacement profiles are computed using the following steps:

We prepare one ordered 2D tissue (honeycomb lattice) by packing $N_c=1024$ identical hexagonal cells, each with preferred area $a^0=1$ and cell perimeter $p^0=3.722$, with periodic boundary conditions. The system is minimized for 300 time units, and the location of cell centroids (pre-deletion centroids) is recorded.

A single cell is arbitrarily chosen as the mitotic cell, which is then deleted using the same protocol described above with ${\text{a}}_{\text{min}}^0=0$. Following post-deletion minimization, new centroids are computed. The deletion-induced displacement map is obtained by subtracting the pre-deletion centroids from these newly computed centroids. Up to this step, we do not allow any T1 transition.

Next, we perform an additional 100 time units of minimization, allowing T1 transition. This is followed by a division of the tiny cell such that the common cell contact between two daughter cells lies approximately parallel to the x-axis. Subsequently, the daughter cells are regrown, and then the system is minimized using the same “division and reinsertion” and “final minimization” protocol described earlier without T1 transitions. The final centroid locations are again recorded, and the reinsertion-induced displacement map is computed by subtracting centroid positions immediately after division from those after final minimization.

Organoid culture

The organoid cultures were derived from murine duodenal crypts. The intestinal crypts were dissociated by treating duodenal tissue with Gentle Cell Dissociation Reagent (Stemcell Technologies) at ~25°C, followed by repeated pipetting in PBS. Crypts were enriched by passing the supernatant through a $70\mu \text{m}$ filter to exclude larger pieces of the intestinal tissue. The isolated crypts were seeded in Matrigel (Corning) and cultured in IntestiCult Organoid Growth Medium (Stemcell Technologies) at 37°C and 5% CO₂. The organoids were passaged every 7 days by dissociating and reseeding the intestinal crypts from the 3D culture. For tracking the nuclei in organoids, the expression of H2B-GFP in R26-M2rtTA; TetOP-H2B-GFP; mT/mG was induced by treating with Doxycycline ($2\mu \text{g}/\text{ml}$) for 24 h.

The mouse strains used in this study were a gift from Linheng Li’s lab. In all experiments involving membrane visualization with cell membrane-localized tdTomato (mT), mT/mG (JAX: 007676)⁷⁰ was used. In experiments where the nuclei were visualized in addition to the cell membrane, mT/mG mice were crossed with R26-M2rtTA; TetOP-H2B-GFP (JAX: 016836)⁷¹. All mouse experiments were approved by the Institutional Animal Care and Use Committee of the Stowers Institute for Medical Research (Protocol ID: 2017–0179) and the University of Nebraska Medical Center (Protocol ID: 22-078-11-EP).

Imaging organoid epithelium

Organoid epithelia were fixed with 4% paraformaldehyde for 30 min and washed with PBST (1X PBS with 0.1% TritonX-100) before the primary antibody was incubated ~12 h at 4°C. The secondary antibody (Goat anti-Mouse Alexa Fluor 568 or Goat anti-Rabbit Alexa Fluor 647) was incubated for 2 h at ~25°C. The samples were mounted in SlowFade Diamond Antifade.

An LSM 900 Airyscan confocal microscope with Plan Apo 40x/1.4 Oil DIC (UV) V-IR and C-Apo 40x/1.2 W objectives was used for point scanning applications. Spinning disk image acquisition was performed using a Nikon Ti Eclipse with a Yokogawa CSU-W1 spinning disk head equipped with a Hamamatsu Flash 4.0 sCMOS and Plan Apo 40X 1.15 NA LWD. Additionally, the organoids were maintained at 37°C and 5% CO2 during live imaging using stage-top incubation chambers from PeCon for Zeiss and OkoLab for Nikon. Timelapse images were acquired at 4 min intervals and with z-sections of $1\mu \text{m}$. For imaging the basal regions of the organoid, the range of z started at the basal-most part of the organoid and ended at the apical-most region. In experiments involving mitotic arrest, organoids were treated with $5\mu \text{g}/\text{ml}$ STC.

Image Processing

Image contrast, brightness, and size were adjusted by linear interpolation and applied to the entire image in FIJI. Drift correction in time-lapse images was performed using the StackReg FIJI plugin. Images were compiled with Adobe Illustrator CC.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification

All quantifications including, cell perimeter, area, cell-contact angles, and neighbor number, were determined by manually segmenting cells in FIJI. Circularity was calculated using the following formula: $C=4\pi A/P^2$, where $A$ and $P$ are cell area and perimeter, respectively. Unless otherwise mentioned, images three timeframes prior to the initiation of basal diminution of mitotic cells were used for all quantifications relating to late interphase cell shape.

Statistical analyses

Sample sizes were based on field standards. At least 20 mitotic events from at least five different organoids were tracked for each experimental condition. Unless otherwise mentioned, $n$ represents the number of mitotic events. In the boxplots, circles represent mean values from individual cells, unless specified. The diamond box contains the 25^th–75^th% percentiles of the data, and the bar denotes the median. The Wilcoxon signed ranks test was used to compare matched samples (e.g., the same cell before and after mitosis), and the Mann-Whitney U test was used to compare two independent samples; NS, $p>0.05$. In topological-distance plots (Figures 1E, S1), same-order exchanges are plotted at whole-number positions and cross-order exchanges at the intervening fractional position. All statistical significance tests were carried out using Origin (OriginLab).

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-ZO-1	Thermo Fisher	40-2200
Mouse anti-LaminA/C	Cell Signaling	4777
Rabbit anti-Ki67	Abcam	ab15580
Goat anti-Mouse Alexa Fluor 568	Thermo Fisher	A-11004
Goat anti-Rabbit Alexa Fluor 647	Thermo Fisher	A-21244
Chemicals, peptides, and recombinant proteins
Hoechst 33342	Thermo Fisher	PI62249
SlowFade Diamond Antifade	ThermoFisher	S36963
Paraformaldehyde	Electron microscopy services	15710
TritonX-100	Sigma	T8787
Gentle Cell Dissociation Reagent	Stemcell Technologies	100-0485
IntestiCult Organoid Growth Medium	Stemcell Technologies	06005
Matrigel	Corning	356231
DMEM/F-12 with 15 mM HEPES	Stemcell Technologies	36254
Experimental models: Organisms/strains
mT/mG mouse	The Jackson Laboratory	JAX: 007676
R26-M2rtTA; TetOP-H2B-GFP mouse	The Jackson Laboratory	JAX: 016836
Software and algorithms
FIJI	Schindelin et al., 2012	https://imagej.net/software/fiji/
Origin	OriginLab	https://www.originlab.com/
Illustrator	Adobe	https://www.adobe.com/
C++ (GNU Compiler Collection, g++); GCC 8.5.0 (Red Hat distribution)	GNU Project	https://gcc.gnu.org/
Python 3	Python Software Foundation	https://www.python.org/
NetworkX (Version: 3.1)	NetworkX Developers	https://networkx.org/
Bash (GNU Bourne Again Shell)	GNU Project	https://www.gnu.org/software/bash/
Visual Studio Code	Microsoft	https://code.visualstudio.com
NumPy (Numerical Python)	NumPy Developers	https://numpy.org/
Matplotlib	Matplotlib Development Team	https://matplotlib.org/
Other
macOS Ventura	Apple	Version 13.7.8
Rocky Linux	Rocky Enterprise Software	8.8 (Green Obsidian)

Highlights:

• In-plane anisotropy of reinsertion orients neighbor exchange in silico and in vivo

• Neighbor exchanges initiate apically and propagate to the basal plane

• Diminution triggers rosettes that initiate basally and propagate apically

• Blocking reinsertion suppresses neighbor exchange but not rosettes in organoids