Asexual strain of S. mediterranea was maintained as previously described (Newmark and Sánchez Alvarado, 2000). For the experiments requiring regenerating fragments, 20 wild-type animals were cut into three fragments–head, trunk and tail – and equal numbers of fragments were collected at 0, 24, 72 and 120 h post amputation (hpa) and analysed further.

RNAi mediated gene knockdown experiments were separately performed by the introduction of artificial dsRNAs synthesized against H2B and Agat-1 respectively, using AmpliScribe T7-Flash Transcription Kit (Epicentre #ASF3507), via microinjections into the worms. A total of 15 animals unfed for a week were injected with 32 nL of 2 μg/μL dsRNA, using Nanoject III™ (Drummond Scientific Company) according to the following regime: injection for 3 consecutive days, a gap of 2 days and, again injection for next 2 consecutive days, for a total of 5 days. All injected worms were collected at 5 days post injection (dpi) and prepared alongside control worms for downstream analysis. Control samples were from the same number of wild-type worms without any injections. All experiments were performed in 3 independent biological replicates.

Primer sequences are provided in Supplementary Table S1 (S1).

A total of 20–30 whole worms were suspended in ice-cold CMFB (Calcium Magnesium-Free Buffer with 1% BSA) and dissociated into single-cells using the gentleMACS™ Octo Dissociator (Miltenyi Biotec). Each sample was subjected to the spleen-1 dissociating program (55 s pulse) for 3 times to achieve homogenous single-cell suspension with minimal debris. The dissociated cells were filtered through 50 µm CellTrics™ filters and centrifuged at 400 rcf and 4 °C to collect the cell pellet. Supernatant was discarded, leaving 50 µL. First, live cell cytoplasm was stained with 0.4 μg/mL Calcein AM dye (ThermoFisher Scientifc # 65-0853–78; Ex/Em 495/515 nm) in 500 µL CMFB solution, followed by a step of PBS wash. Wherever applicable, 1× CellBrite® Steady 550 (Biotium #30107-T; Ex/Em 550/570 nm) in PBS was then added to the cell pellet to stain the cell membrane. The dye was prepared and used according to the manufacturer’s instructions. After a 30-min incubation, no wash step was performed before proceeding to nuclear staining. Subsequently, nuclear staining was performed using 10 µM Draq5 (ThermoFisher Scientific #62251; Ex/Em 646/697 nm). At this stage, the cells were not washed and were directly proceeded for flow cytometry analysis.

To investigate whether MuNs produce AGAT-1 protein, a marker of late epidermal progenitors, a custom anti-AGAT-1 antibody conjugated with Alexa Fluor 488 (Ex/Em 490/525 nm) was used. This antibody was generously provided by Genotic. The antibody was designed using an AI-driven approach, based on the protein sequence predicted from SMEST051964001.1 transcript, obtained in an animal-free system, and has not been fully validated. Cells were fixed using IC Fixation Buffer (eBioscience #00-8222-49) and permeabilized with 1x Permeabilization Buffer (eBioscience #00-8333-56). Following permeabilization, around 1 × 10⁶ cells were stained with 2 µg of the anti-AGAT-1-Alexa Fluor 488 antibody, washed with 1x Permeabilization Buffer and stained with 10 µM Draq5 (ThermoFisher Scientific #62251) as described above.

Imaging flow cytometry was performed on the Cytek® Amnis® ImageStream®X Mk II cytometer equipped with two CCD camera detectors (Cytek Biosciences, USA). INSPIRE software (Cytek Biosciences) was used for data acquisition. The images were captured using a 40× objective at a low fluidics speed. Brightfield images were obtained in channels 1 and 9, side scatter (SSC) images were obtained in channel 6 (745–785 nm filter), using a 785 nm laser with power of 3.75 mW; Calcein AM and Alexa Fluor 488 conjugated to anti-AGAT-1 antibody were detected in channel 2 (480–560 nm filter) using a 488 nm laser with power of 4 mW and 120 mW, respectively; CellBrite Steady 550 was detected in channel 3 (560–595 nm filter) using a 561 nm laser power of 20 mW; Draq5 was detected in channel 11 (642–745 nm filter) using a 642 nm laser with power of 30 mW and, Hoechst 33342 was detected in channel 7 (435–505 nm filter) using a 405 nm laser ×40 with power of 20 mW. At least 60,000 events were acquired from each unsorted sample and a minimum of 2,000 events from each sorted sample. For AGAT-1-oriented experiments, a minimum of 5,000 events per sample were acquired. IDEAS v6.2 software (Cytek Biosciences) was used for data analysis. The analysis strategy has been described in relevant sections of the Results. For the measurement of cell and nuclear diameters, approximately 50–200 cells from each category, MuNs and MoNs (within individual cell populations), were analyzed from a representative dataset.

FACS was performed with a high-speed flow cytometer BD FACSAria™ Fusion (Becton Dickinson). Calcein AM was excited with 488 nm laser and detected in FITC channel (BP 530/30). Draq5 was illuminated using 640 nm laser and detected in APC channel (BP 670/30). The cells were sorted using 100 μm nozzle and 4-way purity sorting mode. Approximately 50,000 cells were sorted per sample into Eppendorf® Protein LoBind 1.5 mL tubes, containing 300 µL PBS with 1% BSA for downstream imaging flow cytometry, or 350 µL TRIzol™ Reagent (Invitrogen, #15596018) for total RNA extraction. The former was processed immediately, and the latter were vigorously shaken once out of the sorter and kept on ice until the next step. Data were analyzed using FACSDiva 9.0.1 software (Becton Dickinson).

The experiments were made in three replicates, unless explicitly stated otherwise. For comparisons involving more than two groups or involving two or more groups, but exhibiting zero variance in any condition, statistical analyses were performed using one-way ANOVA followed by Tukey’s post hoc test. To account for multiple testing, the Holm–Bonferroni method was applied for p-value adjustment. For comparisons between two groups, a two-sample two-sided t-test was conducted with Holm–Bonferroni adjustment. A threshold of significance was set at p < 0.05 for all analyses.

Given the widespread occurrence of polyploid cells in metazoans, we hypothesized that such cells are also present in S. mediterranea. Thus, we undertook an exploratory study in search of multinucleated cells (MuNs), as their polyploid nature can be more definitively established compared to mononucleated cells (MoNs), for which ploidy assessment is more prone to errors. First, we aimed at devising a stepwise gating strategy to identify true MuNs on the IFC Platform, Cytek® Amnis® ImageStream®X Mk II (IS). The cells were stained with Calcein AM and Draq5, where Calcein AM stains cytoplasm of viable cells and Draq5 stains nuclei of live or fixed cells upon DNA binding. Nuclei-containing viable Calcein AM and Draq5 double positive cells were sequentially gated using unstained sample as control (Figure 1A, Steps 1-3). Next, these double positive cells were visualized on Calcein AM vs. Draq5 intensity plot, which allowed us to identify four distinct populations, denoted as B′, C′, D′ and E′, hereafter referred to as IS populations (Figure 1A, Step 4). Preliminary visual assessment of cell images on the Draq5 channel revealed the existence of cells with more than one Draq5 positive spots, considered to be MuNs. Majority of them were found in population E′, with only a few (∼3–4) present in the adjacent population D′. Characterized by large cells with high intensities of both Calcein AM and Draq5 (Figure 1A, blue dots on the dot plot from Step 4), population E′ primarily consisted of large mononucleated cells (MoNs) and a small fraction of MoN clumps. To accurately distinguish multinucleated cells (MuNs) from these other objects, we performed a detailed visual inspection of all objects within gates D′ and E′, identifying the rare MuNs, at a frequency of ∼2% in E′ population (Supplementary Table S2). This translated to a frequency of ∼1–2% in all cells double positive for Calcein AM and Draq5 (Calcein⁺Draq5⁺). This analysis compared images from three channels, brightfield (BF), Calcein AM and Draq5. While both MuNs and MoN clumps on the Draq5 channel displayed multiple nuclear spots, images from the BF and Calcein AM channels distinguished MuNs from MoNs clustered together (Figure 1B).

To address concerns that MuNs might be artifacts of Draq5 staining, we analyzed cell suspensions stained with an alternative nuclear dye, Hoechst 33342. MuNs were indeed observed in similar proportions (Supplementary Figure S1) as in Draq5-stained samples. However, the population containing MuNs in Hoechst 33342-stained samples was less distinct compared to Draq5-stained samples. Consequently, all subsequent studies were conducted using Calcein AM and Draq5 staining for greater clarity and accuracy. Overall, we identified presumed MuNs in wild-type planarians, exploiting the IFC platform. The Calcein AM and Draq5 staining intensity values were amongst the highest for these cells, within the total double positive cell pool.

The size and nuclearity of multinucleated cells vary widely (Brodbeck and Anderson, 2009; Vorobjev et al., 2023). Therefore, to morphologically characterize planarian MuNs we analyzed the cell images based on three criteria: nuclearity (the number of individual nuclei within the cell), cell diameter and nuclear diameter. The number of nuclei within the multinucleated cell pool varied between individual cells. Specifically, the planarian MuNs could be broadly grouped into two categories: 2-nucleated (2nuc) and >2-nucleated (>2nuc). Accordingly, while most MuNs had 2 nuclei (∼67%), a stable percentage was >2 nucleated (∼33%), ranging from 3 to 8 nuclei per cell (Figures 2A,B).

Further, to measure the cellular and nuclear diameters of MoNs and MuNs, we created custom BF and Draq5 image masks to optimally cover the cell and nuclear surface areas, respectively. To obtain cell diameter values, the BF mask was adjusted using the erode function in a way that it correctly fits the BF channel cell images. The nuclear diameters of individual nuclei of each multinucleated cell were measured using the combination of three IDEAS masks: LevelSet, Watershed and Component on Draq5 channel nuclear image. For setting the optimal nuclear mask, three main aspects were considered: (i) to cover a particular nuclear spot area adequately, (ii) to ignore stain halos extending beyond the actual spot area and (iii) to differentiate between two closely placed spots as two different nuclei. Component masks are capable of identifying user-defined objects (in this case, individual nuclei stained with Draq5) one at a time. When combined, the three masks (Component, LevelSet and Watershed) enable the accurate and separate analysis of one or more nuclei within a cell, distinguishing both diffusely stained and distinctly stained nuclei as separate nuclear entities within MuNs. To achieve optimal nuclear identification, individual Component masks (defining nuclear identity) were paired with user-defined constant values for LevelSet and Watershed masks (ensuring accurate nuclear coverage). This approach was designed to accommodate the maximum number of nuclei observed in a planarian MuN, which in this case was 8. Images depicting example masks correctly covering cell and nuclear areas are shown in Figure 2C. The images displaying nuclear masks include five Component masks (combined with LevelSet and Watershed masks), each corresponding to individual nuclei within MuNs containing 2–5 nuclei. In cells with less than 5 nuclei, such as a 2-nucleated MuN – specific nuclear diameter values were recorded for nuclei 1 and 2, while the values from component masks 3, 4, and 5 were zero. This demonstrated the accuracy of the mask development strategy. Using this strategy, we observed that although MuNs had significantly bigger diameters than MoNs, ∼16 µm vs. ∼8–10 μm, respectively (Figure 2D), the nuclear diameters for all cells were generally constant at ∼3–4 µm (Figure 2E) (Supplementary Table S2). These parameters, cell and nuclear diameters, served as indirect measurements for their respective sizes. Overall, MuNs could be described as larger cells containing 2-8 nuclei per cell and having comparable nuclear sizes as that of MoNs. The fact that the nuclear sizes of these 2 cell subsets were similar supported the notion that MuNs indeed contained several nuclei, rather than a single nucleus with multiple foci.

To further characterize the morphological features of MuNs, we visualized the cell membranes using CellBrite, a dye that binds to membrane proteins. We observed that CellBrite efficiently labeled the plasma membrane without significantly penetrating intracellular membranes (Figure 3A). As a result, each cell appeared surrounded by a clearly defined membrane signal, providing a distinct outline of individual cell borders. Interestingly, in some MuNs we observed structures that differed from typical single-cell morphology. These MuNs appeared as clusters of contiguous cells, but their appearance was clearly distinct from random clumps generated during tissue dissociation. In these cases, a single MuN was composed of two or more tightly connected cells forming a coherent structure. This unique membrane staining pattern revealed that individual nuclei of such MuNs were separated from each other via plasma membranes, indicating internal compartmentalization within the larger structure (Figure 3B). The morphology observed here, lacking regular patterns, was interpreted as a late stage of cell fusion, rather than cytokinesis. These findings suggested that, at least in some cases, MuNs biogenesis involves a cell fusion mechanism.

Our next objective was to isolate the cell population containing MuNs for molecular characterization. For this, Calcein AM and Draq5 stained cells were analyzed using FACSAria™ Fusion cell sorter and FACSDiva™ 9.0.1 software as follows. Cells were gated out from the debris, followed by doublet discrimination (Figure 4A, Steps 1 and 2, respectively). After identifying the Calcein AM- and Draq5-positive cells (Figure 4A, Steps 3-4), the double positive cells were plotted on a Calcein AM vs. Draq5 intensity plot (Figure 4A, Step 5). Clearly, due to the intrinsic instrument-related differences, the obtained dot plot presented the acquired events in a distinct way than previously seen in the case of the imaging cytometer. Accordingly, here we gated 5 cell populations, denoted A-E, hereafter referred to as FACS populations (Figure 4A, Step 5). Based on the high staining intensities of both dyes, population E was postulated to contain the MuNs.

To verify this assumption, we visualized the cells from each of the FACS gates. To this end, ∼50,000 cells from each population were sorted and immediately visualized separately on IS. Other than MoNs, we found multiple MuNs within population E, as expected (Figure 4B). MuNs constituted ∼2% of population E, similar to what was observed for population E′ of non-sorted IS samples. This result confirmed our hypothesis that population E contains MuNs and that cell sorting does not alter their distribution within population E. By inspecting the other sorted populations on IS individually, we additionally confirmed that these populations were comprised of MoNs of varying sizes, having different staining intensities for Calcein AM and Draq5 (Supplementary Figure S2). Lastly, we wanted to ensure that MuNs, which are larger than MoNs, were not being substantially excluded by the singlet gate, originally set to remove larger entities such as doublets and/or clumps. Accordingly, we skipped the step of gating singlets (Figure 4A, Step 2) while preserving all the downstream steps to eventually obtain double positive cells. We named the new population E as E1, whose position on the Calcein AM vs. Draq5 dot plot was identical to that of the original population E. Using IS, we observed that sorted population E1 also contained a similar number of MuNs with respect to other cells in the population. Indeed, few larger MuNs were obtained, but most of them were similar in size and morphology to those obtained from population E (Supplementary Figure S3).

To molecularly characterize all the five FACS populations, we again sorted cells and performed qPCR profiling of selected marker genes. A sorted sample of singlets, containing a mix of cells from all targeted populations and untargeted regions of the dot plot, was used as a proxy for the suspension of all cell types. The selected marker genes represented three classes of cells at various stages of differentiation – Smedwi1 (neoblast); Agat-1 (late epidermal progenitor) and Glutathione S transferase-1 or GST-1 (differentiated phagocyte - dd_Smed_v6_20_0_1). As all of the sorted samples were heterogenous mixtures of cells, the relative expression level of each marker in a particular sample reflected the approximate proportion of the corresponding cell type within that population. Accordingly, populations A, B, and C majorly consisted of differentiated cells. Population D was likely dominated by neoblasts, inferred from the highest proportion of the neoblast marker expression. Population E exhibited the highest proportion of the late epidermal progenitor marker gene expression compared to all other populations, suggesting a predominance of progenitor cell states. Additionally, populations D and E showed the lowest expression of the differentiated cell marker, further supporting their classification as undifferentiated cell populations (Figure 4C). However, it is important to note that none of these individually sorted populations represented homogenous cell types. Altogether, despite the heterogeneity within each population, the undifferentiated marker gene expression signatures suggested that the MuN-containing population E likely represents undifferentiated subtypes.

Given the possible undifferentiated state of MuNs, we next studied them along a defined time frame of regeneration. To this end, planarians were cut, and cell dissociates were analyzed using IS at 0, 24, 72 and 120 h post amputation (hpa). A dissociate of non-amputated worms was used as control. The percentage of MuNs in the pool of all Calcein⁺Draq5⁺ cells fluctuated between various regenerative time points and non-amputated worms, but the changes did not reach statistical significance. This means that irrespective of the regenerative state, the total proportion of MuNs was stable (Figure 5A). To analyze the MuNs further, we calculated the proportions of 2nuc and >2nuc cells in all the above conditions. We found that there was no significant difference in proportions between the samples (Figure 5B). The consistent presence of MuNs in both regenerative and non-regenerative life stages suggest they are a consistent component of the basal homeostatic state of planarians.

From the qPCR analysis of sorted cell populations, we had initial indication suggesting that planarian MuNs could be undifferentiated. To further investigate this, we aimed to examine the effects of targeted neoblast depletion on MuNs and the MuN-containing cell population.

Neoblast depletion was achieved by knocking down the H2B gene (H2B KD) using RNAi. At first, gene expression analyses were conducted on H2B KD vs. wild-type (WT) worms, focusing on three groups of marker genes representing neoblasts, progenitors, and differentiated cells. Neoblasts and early progenitors were marked by Smedwi1 and NB21.11e (SMEST024707001.1), respectively, late progenitors by Agat-1, and differentiated cells by the phagocytic marker GST-1 and the neural marker prohormone convertase-2 (PC2, dd_Smed_v6_1566_0_1). To this end, we used qPCR to analyze marker gene expression in H2B KD worms relative to WT. As expected, we observed a sharp decrease in the expression of neoblast and early progenitor markers. Notably, there was also a significant reduction in the expression of the late epidermal progenitor marker. These changes were accompanied by an increase in the expression of differentiated cell markers, reflecting a compensatory enrichment in the number of differentiated cells due to the absence of neoblasts and progenitors in the constant cell pool (Figure 6A).

To check if H2B KD directly affected the MuNs, we next calculated the percentage of these cells under RNAi vs. WT conditions. To this end, we analyzed cell dissociates for both sample types using IS. We found that the proportion of MuNs was marginally reduced in the H2B KD samples compared to WT, reaching approximately 0.8 relative to the WT level (normalized to 1). (Figure 6B). This led us to further analyze all the individual populations identified earlier using IS and FACS. Based on the IS plots, we found that there were two populations whose percentage contributions to the total pool of all Calcein⁺Draq5⁺ cells significantly decreased as a result of H2B KD. Percentage of population D′ in H2B KD was 0.2 times the WT value (3.1% in H2B KD and 16% in WT) and for E′ containing MuNs, the percentage in H2B KD was 0.6 times the WT value (34% in H2B KD and 54.8% in WT). This helped us conclude that these two populations contained undifferentiated cell types and based on the levels of reduction of the two, it was concluded that population D′ contained mostly neoblasts and the lesser affected population E′ was majorly comprised of progenitors at various stages (Figure 6C). As for the FACS populations, we again obtained two populations whose percentage contributions to the total cell pool were particularly reduced. Population percentages of E (containing the MuNs) and D were 0.4 times (9.1% in H2B KD and 25.3% in WT) and 0.6 times (9% in H2B KD and 14.7% in WT) of their respective WT values (Figure 6D). As the reduction in population D was much more profound than that of E, it could be concluded that D majorly contains neoblasts and E, progenitors. This was in line with the observations presented above (Figure 6A) that the progenitor marker expression was reduced to a much lesser extent than the neoblast marker, and populations D and E have the highest expressions of the neoblast and progenitor markers (Figure 4C), respectively.

Given that the MuNs-containing population E appeared to represent progenitor cells, we examined how the proportion of MuNs would be affected upon the knockdown of Agat-1. After verification of the knockdown efficiency (Figure 7A), we analyzed cell dissociates from Agat-1 KD and WT worms using IS, and quantified MuNs in both sample types. We observed a marked reduction in the abundance of MuNs in Agat-1 KD samples compared to WT, reaching approximately 0.4 relative to the WT level (normalized to 1) (Figure 7B). This was clearly more pronounced than the reduction seen upon H2B depletion. We then analyzed particular cell populations, this time focusing solely on the IS-based classification. Notably, no significant changes were observed in the percentages of any cell population, including population E’ (Figure 7C). These results indicated that Agat-1 knockdown exerted a preferential effect on MuNs, without broadly affecting the composition of major cell populations.

The observed reduction in MuNs numbers following Agat-1 knockdown implied that AGAT-1 is important for the formation or maintenance of MuNs. However, this result did not directly address whether MuNs themselves produce the AGAT-1 protein. To investigate this, we again employed imaging flow cytometry, performing a single experiment using Draq5 nuclear dye and an anti-AGAT-1 antibody to stain cell suspensions prepared from WT planarians. The antibody was custom-produced and has not been extensively validated. The Draq5 and anti-AGAT-1 double positive cells were sequentially gated using unstained sample as control (Figure 8A) The analysis revealed that ∼25–30% of Draq5⁺ cells were positive for AGAT-1. Further, majority of MuNs (16 out of 23 events, 70%) were found to be a part of the AGAT-1⁺ pool of cells suggesting that these cells are likely to represent a pool of late epidermal progenitors (Figure 8B). However, it is important to emphasize that AGAT-1 is not exclusive to MuNs and, therefore, cannot be considered a MuN-specific marker. Additionally, given the characteristics of the antibody used, the above results warrant further verification.