Humans are part of a group of animals called vertebrates, which are all the animals with backbones. Broadly, all vertebrates have a similar body shape with a head at one end and a left and right side that are similar to each other. Although this is not very obvious in humans, vertebrate bodies are derived from pairs of segments arranged from the head to the tail. Each of these segments or somites originates early in embryonic development. Cells from each somite then divide, grow and specialize to form bones such as the vertebrae of the vertebral column, muscles, skin, and other tissues that make up each segment.

Studying different animals during embryonic development has provided insights into how somites form and grow, but it is technically difficult to do and only provides an approximate model of how somites develop in humans. Being able to make and study somites using human cells in the lab would help scientists learn more about how somite formation in humans is regulated.

Budjan et al. grew human stem cells in the lab as three-dimensional structures called organoids, and used chemical signals similar to the ones produced in the embryo during development to make the cells form somites. Various combinations of signals were tested to find the best way to trigger somite formation. Once the somites formed, Budjan et al. measured them and studied their structure and the genes they used. They found that these lab-grown somites have the same size and structure as natural somites and use many of the same genes.

This new organoid model provides a way to study human somite formation and development in the lab for the first time. This can provide insights into the development and evolution of humans and other animals that could then help scientists understand diseases such as the development of abnormal spinal curvature that affects around 1 in 10,000 newborns.

Paraxial mesoderm (PM) development involves the formation of embryonic segments called somites, which are produced sequentially from the presomitic mesoderm (PSM) and arranged periodically along the anterior-posterior (AP) axis of the vertebrate embryo. Somites give rise and contribute to a variety of tissues including skeletal muscle, dermis, cartilage, and bone (Chal and Pourquié, 2017). Somite formation is controlled by a conserved molecular oscillator, the segmentation clock (Dequéant et al., 2006; Hubaud and Pourquié, 2014; Oates et al., 2012). Previous efforts have focused on how this oscillator controls somite formation using a variety of model systems such as mouse, zebrafish, and chick because of ethical and technical limitations of culturing human embryos. Recently, researchers were able to recapitulate PM development using human and mouse pluripotent stem cells cultured as 2D monolayers (Chu et al., 2019; Diaz-Cuadros et al., 2020; Matsuda et al., 2020). These cells undergo species-characteristic oscillations similar to their in vivo counterparts. However, final stages of somite development and vertebra formation were not observed in currently published human cell culture systems (Palla and Blau, 2020), suggesting that certain aspects of in vivo development are not recapitulated in these 2D systems. We reasoned that a 3D cell culture system may exhibit all the stages of PM development including morphogenetic processes associated with somite formation.

Here, we describe an organoid system derived from human induced pluripotent stem cells (hiPSCs) called Somitoids, which faithfully recapitulates functional, morphological, and molecular features of PM development, including formation of somite-like structures in vitro. To identify the culture conditions that reproducibly yielded formation of somite-like structures, we developed a quantitative image-based screening platform for individual organoids. Our screening approach identified the optimal parameter values for the culture conditions such as the initial cell number and the concentration of the chemical modulators. We show using single-cell RNA-sequencing (scRNA-seq), immunofluorescence, and qRT-PCR that Somitoids resemble their in vivo counterparts both transcriptionally and morphologically and can be differentiated into somite derivatives such as sclerotome and dermomyotome in vitro.

Our Somitoid system is reproducible and scalable, allowing for systematic and quantitative analysis of PM development and somite formation to study human spine development and disease in a dish. Finally, our approach can be used to systematically screen organoid cultures for desired phenotypes and reproducibility.

Recently, protocols have been developed to differentiate mouse or human pluripotent stem cells (hPSCs) towards PM using a combination of the WNT agonist CHIR and BMP inhibitor LDN (Chal et al., 2015; Diaz-Cuadros et al., 2020). To adapt the protocol for a 3D model of human somitogenesis, we first optimized the initial conditions of our cultures by generating pluripotent spheroids of defined cell numbers. hiPSCs were allowed to aggregate for 24 hr as suspension cultures in pluripotency media (mTeSR1) in the presence of ROCK inhibitor and polyvinyl alcohol (PVA) to promote aggregation (Figure 1A). These pluripotent spheroids resemble cavity-stage epiblast embryos as previously described for kidney organoid cultures (Freedman et al., 2015; Figure 1B, Figure 1—figure supplement 1A). Next, spheroids were cultured in media containing CHIR and LDN (CL), similarly as done for the monolayer cultures (Chal et al., 2015; Diaz-Cuadros et al., 2020), but with CHIR at a higher concentration (10 µM). After 24 hr in CL media, epiblast-stage cells transition to a neuromesodermal progenitor (NMp) or primitive streak cell fate, characterized by co-expression of T/BRA and SOX2 (Tzouanacou et al., 2009; Figure 1B). By 48 hr, cells rapidly downregulate SOX2 and express PSM markers, including TBX6 and MSGN1 (Figure 1B). This PSM state persists from day 2 to day 4 and is also characterized by the expression of segmentation clock genes such as HES7 (Figure 1C). On day 5, organoids showed expression of marker genes associated with somite fate as characterized by qPCR (Figure 1C, Figure 1—figure supplement 1B). Taken together, the order of activation of marker genes in the PM organoids followed the expected stages of differentiation observed during PM development.

Figure 1 with 2 supplements see all

Download asset Open asset

Figure 1—source data 1

qPCR raw data of paraxial mesoderm (PM) organoid differentiation.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig1-data1-v2.txt

Download elife-68925-fig1-data1-v2.txt

We observed that following the above protocol resulted in a heterogeneous activation of somite marker genes across cells within the same organoid and across different replicates, as well as a low number of somite-like structures (Figure 1—figure supplement 2). To improve reproducibility, we set out to screen for the optimal initial cell number, the concentration of the signaling factors, and the timing of their additions during culture. For our primary screen, we compared organoids with an initial cell number of 500 and 1000 as well as combinations of modulators of candidate signaling pathways that have previously been involved in PM development and somite formation, including FGF, WNT, BMP, and TGF-β (Aulehla et al., 2008; Chal et al., 2015; Hubaud and Pourquié, 2014; Tonegawa et al., 1997; Xi et al., 2017). We primarily focused on the FGF and WNT signaling pathways since their critical role during somitogenesis is well established both in vivo (Aulehla et al., 2008; Aulehla et al., 2003; Delfini et al., 2005; Dubrulle et al., 2001; Dunty et al., 2008; Greco et al., 1996; Yamaguchi et al., 1999) and in vitro (Chal et al., 2015; Sakurai et al., 2012; Tan et al., 2013). Furthermore, dual inhibition of FGF and WNT signaling has been used with some success to generate PM derivatives in vitro (Loh et al., 2016; Matsuda et al., 2020). Finally, the BMP and TGF-β signaling pathways have been shown in vitro and in vivo to have a role in human somitogenesis (Loh et al., 2016; Xi et al., 2017).

PSM-stage organoids on day 3 were treated with signaling modulators for 24 and 48 hr, and somite fate and morphogenesis was assessed using PAX3, a somite fate marker, and F-ACTIN, a structural marker of somite formation (Figure 2A). We chose day 3 as a starting point for our systematic screen because PSM marker gene expression was more uniform compared to day 2 organoids (Figure 1B), and day 4 organoids are not significantly different from day 3 organoids based on a previous study and our own immunostaining and qPCR data (Figure 1B and C; Diaz-Cuadros et al., 2020). To quantitatively compare conditions, we developed an image analysis pipeline to determine organoid diameter and normalized average PAX3 expression intensity per organoid in an automated manner (Figure 2C and D, Figure 2—figure supplement 2B). We analyzed three organoids per condition. Strikingly, all organoids that were treated with any combination of FGF or WNT inhibitor reproducibly expressed PAX3 within 24 hr of treatment (Figure 2B and D). However, organoids that were initiated from 1000 cells displayed a higher fraction of PAX3-negative cells compared with organoids initiated from 500 cells, even though the average PAX3 expression levels across the entire organoid were comparable (Figure 2D, Figure 2—figure supplement 2A). Additional staining for SOX2, a neural marker, showed that PAX3-negative cells expressed SOX2, suggesting that our PM organoids derive from NMps (Figure 1B, Figure 2—figure supplement 2A). In addition, staining for F-ACTIN, together with nuclear expression of PAX3, more consistently revealed somite-like structures (radial arrangement of PAX3+ columnar cells with expression of apical F-ACTIN in the central cavity) in organoids made from 500 cells compared with organoids made from 1000 cells (Figure 2B, Figure 2—figure supplement 3A and B). Taken together, based on these observations, we used 500 cells as the initial cell number going forward.

Figure 2 with 3 supplements see all

Download asset Open asset

Figure 2—source data 1

Quantification of organoid diameter from primary screen.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig2-data1-v2.csv

Download elife-68925-fig2-data1-v2.csv

Figure 2—source data 2

Quantification of average PAX3 levels per organoid from primary screen.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig2-data2-v2.csv

Download elife-68925-fig2-data2-v2.csv

Organoids that were treated for 48 hr with signaling pathway modulators showed an overall decrease of PAX3 expression compared with organoids treated for only 24 hr across replicates, indicating that prolonged signaling manipulation does not improve the somite phenotype (Figure 2D). Additionally, organoids initiated from 500 cells and then treated for 48 hr had a smaller diameter compared with organoids of the same initial cell number that were treated for only 24 hr (Figure 2C). This suggests that long-term inhibition of WNT and/or FGF, known mitogenic signaling pathways, has detrimental effects on proliferation or cell survival. These results indicate that treatment of PSM-stage organoids with pathway modulators for 24 hr is sufficient to induce somite fate.

Next, we set out to optimize the culture conditions to increase the number of somite-like structures in addition to the expression levels of the somite marker genes. We looked for morphological hallmarks of somite formation, specifically the formation of rosette-like structures consisting of radially arranged bottle-shaped PAX3+ epithelial cells with their NCAD+ apical surface facing a central cavity (Figure 3—figure supplement 1A and B, Figure 3—video 1).

We used 500-cell spheroids as an initial starting point and compared two different inhibitor doses for FGF and WNT in addition to the other pathway modulators applied to PSM-stage organoids on day 3. We characterized the organoids after treating them for 24 hr, 48 hr, and 24 hr followed by culture in basal media without any added factor for an additional 24 hr (five organoids per condition, Figure 3A). In addition to quantifying PAX3 levels (Figure 3D, Figure 3—figure supplement 2B), we also counted the number of somite-like structures per organoid (Figure 3B and E, Figure 3—figure supplement 2A and B; see Materials and methods for a description of the scoring criteria). Comparing PAX3 expression levels in our treated organoids, we observed that somite fate can be broadly induced across a range of treatment regimes, concentrations, and types of WNT and/or FGF inhibitors (Figure 3C and D, Figure 3—figure supplement 2C). However, the number of somite-like structures is not necessarily correlated with average PAX3 expression levels. For example, the numbers of somite-like structures in several conditions (FGFRi^hi/PD173, WNTi^hi/C59, WNT^hi/C59+ FGFRi^lo, WNTi^hi/XAV) were lower in organoids treated for 24 hr followed by culture in basal media for 24 hr compared with organoids that were treated with the same inhibitors for 48 hr, even though they exhibited higher PAX3 expression levels on average (Figure 3D and E, Figure 3—figure supplement 2A). This suggests that marker gene expression alone may not be the best predictor when screening for morphologically complex phenotypes such as somite formation.

Figure 3 with 4 supplements see all

Download asset Open asset

Figure 3—source data 1

Quantification of average PAX3 levels per organoid from secondary screen.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig3-data1-v2.csv

Download elife-68925-fig3-data1-v2.csv

Figure 3—source data 2

Quantification of somite-like structures of secondary screen.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig3-data2-v2.csv

Download elife-68925-fig3-data2-v2.csv

Surprisingly, organoids that were cultured for an additional 24 hr (day 3 to day 4) in FGF, WNT pathway agonist, and BMP inhibitor, considered a treatment control, followed by culture in basal media only for another 24 hr, consistently exhibited the highest number of somite-like structures across all organoid replicates (Figure 3E, Figure 3—figure supplement 1A and B, Figure 3—figure supplement 2A and B, Figure 3—video 1) as well as technical replicates (Figure 3—figure supplement 3A). Additionally, the average PAX3 expression was among the highest of all conditions tested (Figure 3D). This suggests that simply removing FGF/WNT pathway agonists and BMP inhibitor, which maintain cells in a PSM state, is sufficient to reproducibly induce somite fate and morphological formation of somite-like structures (Figure 3—figure supplement 2A, Figure 3—figure supplement 3A). Computing the variation of the number of somite-like structures across the five organoids confirmed that this phenotype was highly reproducible (coefficient of variation = 11.1%; Figure 3—figure supplement 2B). Finally, our optimized protocol reproducibly yielded efficient induction of somite-like structures in multiple genetically independent hiPSC lines (Figure 3—figure supplement 3B). Taken together, we determined that initiating the protocol with 500 cells and treating day 3 organoids with CL + FGF for 24 hr followed by culture in basal media for an additional 24 hr yield the most robust induction of somite-like structures while minimizing variation between experiments (technical variation) and different cell lines (biological variation). This optimized differentiation protocol was therefore used for all subsequent experiments.

To further characterize the developmental trajectory and transcriptional states of our Somitoid system, we collected 15,558 cells (after postprocessing) over the course of the optimized 5-day differentiation protocol at timepoints that capture the key transition steps (day 1, day 2, day 3, day 5) and performed scRNA-seq (Figure 4A). Multiple organoids were used to obtain the required number of cells at each timepoint (see Materials and methods). We first combined all the cells across the timepoints and clustered them using the Leiden clustering algorithm (Traag et al., 2019). Predominantly, the four major clusters corresponded to cells from the four different timepoints. Therefore, cells at each timepoint have transcriptional states that are different compared with cells from the other timepoints. In addition, within each timepoint, cells exhibit similar transcriptional profiles as indicated by the uniformity of the expression levels of marker genes across individual cells (Figure 4E, Figure 4—figure supplements 1–3).

Figure 4 with 5 supplements see all

Download asset Open asset

Cells collected on day 1 exhibited gene expression profiles similar to primitive streak or NMps, expressing genes such as SOX2, T/BRA, MIXL1, and NODAL (Figure 4C and E, Figure 4—figure supplement 1). Starting on day 2, cells expressed canonical PSM marker genes such as TBX6, MSGN1, WNT3A, RSPO3, and clock genes of the Notch signaling pathway including HES7, LFNG, DLL1, and DLL3 (Figure 4C and E, Figure 4—figure supplement 2). Day 5 cells expressed somite marker genes such as TCF15, PAX3, FOXC2, and MEOX2 (Figure 4C and E, Figure 4—figure supplement 3). Interestingly, a subset of the day 5 cells also expressed somite polarity markers, UNCX and TBX18, which suggests faithful recapitulation of somite patterning in Somitoids (Figure 4C, Figure 4—figure supplement 3). Furthermore, two of the subclusters (‘PSM-to-somite,’ ‘early somite’), which are characterized by co-expression of PSM and somite marker genes, comprised both day 3 and day 5 cells, indicating that the PSM-to-somite transition is captured in our in vitro system (Figure 4A and B). Interestingly, one somite subcluster (‘late somite’) was enriched for myogenic genes (MYL4/6/7/9, TROPONIN L1) and sclerotome genes (TWIST1, COL1A1, COL11A1, COL7A1, ACTA2), suggesting that these cells represent more downstream fates of somite-derived cells (Figure 4B, Figure 4—figure supplement 4A). Finally, we observed the expected sequential activation pattern of the HOX genes in our Somitoid system starting with HOXA1 on day 1, followed by other cervical and thoracic HOX genes on days 2–3, to HOXD9, a lumbosacral HOX gene, in the somite-stage organoids (day 5; Figure 4D, Figure 4—figure supplement 4B). Taken together, our scRNA-seq analyses show that our Somitoid system faithfully recapitulates the gene regulatory programs of human PM development and generates mature somite-like cells, which express the full repertoire of known marker genes. Furthermore, we did not detect cells of different developmental origins, suggesting that we are generating homogeneous organoids containing only PM derivatives.

We next assessed whether our in vitro-derived somite-like structures show similar spatial organization and size distribution to their in vivo counterparts. To independently confirm some of the somite marker genes that we identified in our scRNA-seq dataset, we measured expression levels of TCF15/PARAXIS, PAX3, and F-ACTIN in day 5 Somitoids using whole-mount immunostaining (Figure 5A). Day 5 cells co-expressed both somite marker genes TCF15 and PAX3 throughout the Somitoid, and somite-like structures displayed apical localization of F-ACTIN. To determine whether in vitro-derived somite-like structures were similar in size to human embryonic segments, we compared them with Carnegie stage 9–11 early human somites (Figure 5B; see Materials and methods for a description of how somite sizes were quantified). Organoid-derived somite-like structures were similar in size (median area = 8892 µm², interquartile range [IQR] = 7698–10682 µm²) to Carnegie stage 11 somites (median area = 9681 µm², IQR = 8262–11493 µm²) but larger than earlier-stage human somites (Carnegie stage 9 somites, median area = 4399 µm², IQR = 4089–4433 µm²; Carnegie stage 10 somites, median area = 4704 µm², IQR = 4477–5343 µm²; see Figure 5B). Together, these results suggest that our organoid-derived somite-like structures share spatial and molecular features as well as overall size with their in vivo counterparts.

Figure 5 with 1 supplement see all

Download asset Open asset

Figure 5—source data 1

Comparative somite size quantification of in vitro somite-like structures and human somites from the Carnegie collection.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig5-data1-v2.csv

Download elife-68925-fig5-data1-v2.csv

Figure 5—source data 2

qPCR raw data of sclerotome differentiation.

https://cdn.elifesciences.org/articles/68925/elife-68925-fig5-data2-v2.txt

Download elife-68925-fig5-data2-v2.txt

Finally, we assessed whether Somitoids can give rise to downstream PM derivatives of sclerotome and dermomyotome. First, we differentiated Somitoids to sclerotome by exposing day 5 organoids to SHH agonist and WNT inhibitor to mimic the signaling environment of ventral somites in vivo (Fan et al., 1995; Fan and Tessier-Lavigne, 1994; Loh et al., 2016). After 3 days, the organoids showed a robust induction of canonical sclerotome marker genes such as PAX1, SOX9, and COL2A1 (Figure 5C). Additionally, day 5 Somitoids were differentiated towards dermomyotome by exposing them to WNT/BMP agonists and SHH inhibitor for 48 hr followed by dissociation and culture on Matrigel as a monolayer in muscle differentiation medium (Loh et al., 2016; Matsuda et al., 2020) to further differentiate them to skeletal muscle. Immunostaining for myosin heavy chain (MYH1, a myocyte/myotube marker) confirmed that our Somitoid-derived cells can generate skeletal muscle derivatives in vitro (Figure 5—figure supplement 1). These data demonstrate that our in vitro-induced somite-like structures maintain their ability to differentiate further into somitic mesoderm derivatives of the sclerotome and dermomytome lineages.

Here, we reported the generation of human PM organoids from hPSCs that reproduce important features of somitogenesis not previously captured in conventional monolayer differentiation cultures, most notably formation of somite-like structures. Using a simple suspension culture that does not require manual matrix embedding, we identified optimal differentiation conditions by systematically screening initial cell numbers and modulating the signaling factors. Importantly, our culture conditions are compatible with high-throughput screening approaches. Many established organoid protocols currently have limited applications because they are not reproducible. Therefore, we set out to identify the optimal differentiation conditions that minimized the variability between organoids as quantified by automated image analysis.

One critical parameter we identified in our screens was the initial cell number used for aggregation. Our results suggest that if the initial cell number is above a certain threshold then somite fate cannot be induced in a homogeneous manner in our organoid system. This result is in line with previous findings in 3D models such as gastruloids, multi-axial self-organizing aggregates of mouse ES cells, which exhibit a higher degree of variability and multiple elongations when the number of initial cells exceeds a threshold (Beccari et al., 2018; van den Brink et al., 2014). Another important finding of our screens was that simply removing FGF, WNT pathway agonist as well as BMP inhibitor yielded the most reproducible and efficient somite-like structure-forming organoids. This treatment regime does not necessarily follow from applying prior in vivo and in vitro knowledge of somitogenesis. Previous protocols have used FGF and WNT inhibitors (Matsuda et al., 2020) or inhibition of all four candidate signaling pathways (FGF, WNT, BMP, and TGF-β) to induce somite fate (Loh et al., 2016) in monolayer cultured hiPSCs. While these conditions similarly induced somite fate marker genes in our 3D system, removal of FGF/WNT agonists and BMP inhibitor overall performed better as indicated by larger organoid diameters, higher average PAX3 expression levels, and higher number of somite-like structures. In line with these findings, our scRNA-seq analysis revealed that day 5 somite-like cells from our optimized protocol autonomously downregulate WNT target genes (DKK1, AXIN2, WNT3A, RSPO3; Figure 4—figure supplement 2) and FGF target genes (FGF8, FGF17, SPRY4, DUSP6/MKP3, SEF/IL17RD; Figure 4—figure supplement 5). Finally, our screening results also suggest that focusing on marker gene induction as a phenotypic readout alone is not sufficient to optimize culture conditions of more complex organoid models such as somitogenesis.

The scRNA-seq analysis of our hPSC-derived Somitoids independently confirmed our immunostaining and qRT-PCR results and showed that all major paraxial mesodermal cell types are generated, consistent with the cell types observed during PM development. Comparing our single-cell dataset with previously published in vitro-generated human PM transcriptomic data of monolayer cultures (Diaz-Cuadros et al., 2020; Matsuda et al., 2020) reveals a similar pattern of activation of marker genes. Diaz-Cuadros and colleagues did not generate bona fide somitic cells as their final timepoint cell population does not express canonical somitic markers. Matsuda et al. indeed show expression of several somitic marker genes including TCF15, MEOX1, and PAX3 based on bulk RNA-seq data. Interestingly, our own analysis of day 5 somitic cells revealed multiple distinct subclusters, suggesting transcriptional heterogeneity within this population, which could have not been inferred from bulk data (Figure 4B and C, Figure 4—figure supplement 4). Importantly, neither of these papers report formation of somite-like structures, suggesting that transcriptional similarity alone is not sufficient to predict morphological features, in line with our screening results showing that average marker gene expression is not a good predictor of in vitro induction efficiency of somite-like structures (Figure 3D and E, Figure 3—figure supplement 2A–C).

While expression patterns of canonical somitic marker genes seem to be conserved in humans, it will be interesting to perform detailed gene expression analysis to identify putative human-specific genes of somite differentiation. Since our Somitoid system is reproducible, it could serve as a versatile platform to perform functional screens of human-specific or disease-relevant genes using CRISPR/Cas9 or small-molecule inhibitor libraries. Somitoids thus provide a powerful in vitro system for studying the regulation and dynamics of human somitogenesis, including somite formation.

One limitation of this work is that the current protocol does not produce an AP axis and therefore does not generate somite-like structures in a bilaterally symmetric fashion as in the vertebrate embryo. A similar phenotype was recently reported in mouse gastruloids grown under similar signaling conditions that were embedded in Matrigel to promote self-organization into trunk-like structures (TLS; Veenvliet et al., 2020). Chemical modulation of BMP and WNT signaling pathways in Matrigel-embedded gastruloids resulted in the formation of somite-like structures arranged like a bunch of grapes, similar to what we observed in our system. In standard culture conditions, gastruloids and TLS recapitulate the axial organization of the embryo, which is missing in our Somitoids (Beccari et al., 2018; Moris et al., 2020; Veenvliet et al., 2020). To expand the patterning and morphogenetic potential of our Somitoid system, our approach could be combined with a microfluidics setup to generate spatio-temporally controlled morphogen gradients (Manfrin et al., 2019). In summary, Somitoids provide a scalable, reproducible, and easy to manipulate platform to study molecular networks underlying the differentiation of PM, as well as the morphogenetic processes of somite formation. Furthermore, Somitoids represent a promising in vitro system to study congenital diseases that are linked to the human segmentation clock and somite formation, such as congenital scoliosis.

Sequencing data has been deposited in GEO under accession code GSE194214; All data used to generate the figures are included in the manuscript as Source Data files.

1. 1. Christoph B
   2. Shichen L
   3. Sahand H

   (2021) NCBI Gene Expression Omnibus

   ID GSE194214. Paraxial mesoderm organoids model development of human somites.

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE194214

1. 1. Aulehla A
   2. Wehrle C
   3. Brand-Saberi B
   4. Kemler R
   5. Gossler A
   6. Kanzler B
   7. Herrmann BG

   (2003) Wnt3a plays a major role in the segmentation clock controlling somitogenesis

   Developmental Cell 4:395–406.

   https://doi.org/10.1016/s1534-5807(03)00055-8

   • PubMed
   • Google Scholar
2. 1. Aulehla A
   2. Wiegraebe W
   3. Baubet V
   4. Wahl MB
   5. Deng C
   6. Taketo M
   7. Lewandoski M
   8. Pourquié O

   (2008) A beta-catenin gradient links the clock and wavefront systems in mouse embryo segmentation

   Nature Cell Biology 10:186–193.

   https://doi.org/10.1038/ncb1679

   • PubMed
   • Google Scholar
3. 1. Beccari L
   2. Moris N
   3. Girgin M
   4. Turner DA
   5. Baillie-Johnson P
   6. Cossy AC
   7. Lutolf MP
   8. Duboule D
   9. Arias AM

   (2018) Multi-axial self-organization properties of mouse embryonic stem cells into gastruloids

   Nature 562:272–276.

   https://doi.org/10.1038/s41586-018-0578-0

   • PubMed
   • Google Scholar
4. 1. Chal J
   2. Oginuma M
   3. Al Tanoury Z
   4. Gobert B
   5. Sumara O
   6. Hick A
   7. Bousson F
   8. Zidouni Y
   9. Mursch C
   10. Moncuquet P
   11. Tassy O
   12. Vincent S
   13. Miyanari A
   14. Bera A
   15. Garnier J-M
   16. Guevara G
   17. Hestin M
   18. Kennedy L
   19. Hayashi S
   20. Drayton B
   21. Cherrier T
   22. Gayraud-Morel B
   23. Gussoni E
   24. Relaix F
   25. Tajbakhsh S
   26. Pourquié O

   (2015) Differentiation of pluripotent stem cells to muscle fiber to model Duchenne muscular dystrophy

   Nature Biotechnology 33:962–969.

   https://doi.org/10.1038/nbt.3297

   • PubMed
   • Google Scholar
5. 1. Chal J
   2. Pourquié O

   (2017) Making muscle: skeletal myogenesis in vivo and in vitro

   Development (Cambridge, England) 144:2104–2122.

   https://doi.org/10.1242/dev.151035

   • Google Scholar
6. 1. Chu LF
   2. Mamott D
   3. Ni Z
   4. Bacher R
   5. Liu C
   6. Swanson S
   7. Kendziorski C
   8. Stewart R
   9. Thomson JA

   (2019) An In Vitro Human Segmentation Clock Model Derived from Embryonic Stem Cells

   Cell Reports 28:2247–2255.

   https://doi.org/10.1016/j.celrep.2019.07.090

   • PubMed
   • Google Scholar
7. 1. Delfini MC
   2. Dubrulle J
   3. Malapert P
   4. Chal J
   5. Pourquié O

   (2005) Control of the segmentation process by graded MAPK/ERK activation in the chick embryo

   PNAS 102:11343–11348.

   https://doi.org/10.1073/pnas.0502933102

   • PubMed
   • Google Scholar
8. 1. Dequéant M-L
   2. Glynn E
   3. Gaudenz K
   4. Wahl M
   5. Chen J
   6. Mushegian A
   7. Pourquié O

   (2006) A complex oscillating network of signaling genes underlies the mouse segmentation clock

   Science (New York, N.Y.) 314:1595–1598.

   https://doi.org/10.1126/science.1133141

   • PubMed
   • Google Scholar
9. 1. Diaz-Cuadros M
   2. Wagner DE
   3. Budjan C
   4. Hubaud A
   5. Tarazona OA
   6. Donelly S
   7. Michaut A
   8. Al Tanoury Z
   9. Yoshioka-Kobayashi K
   10. Niino Y
   11. Kageyama R
   12. Miyawaki A
   13. Touboul J
   14. Pourquié O

   (2020) In vitro characterization of the human segmentation clock

   Nature 580:113–118.

   https://doi.org/10.1038/s41586-019-1885-9

   • PubMed
   • Google Scholar
10. 1. Dubrulle J
    2. McGrew MJ
    3. Pourquié O

    (2001) FGF signaling controls somite boundary position and regulates segmentation clock control of spatiotemporal Hox gene activation

    Cell 106:219–232.

    https://doi.org/10.1016/s0092-8674(01)00437-8

    • PubMed
    • Google Scholar
11. 1. Dunty WC
    2. Biris KK
    3. Chalamalasetty RB
    4. Taketo MM
    5. Lewandoski M
    6. Yamaguchi TP

    (2008) Wnt3a/beta-catenin signaling controls posterior body development by coordinating mesoderm formation and segmentation

    Development (Cambridge, England) 135:85–94.

    https://doi.org/10.1242/dev.009266

    • PubMed
    • Google Scholar
12. 1. Fan CM
    2. Tessier-Lavigne M

    (1994) Patterning of mammalian somites by surface ectoderm and notochord: evidence for sclerotome induction by a hedgehog homolog

    Cell 79:1175–1186.

    https://doi.org/10.1016/0092-8674(94)90009-4

    • PubMed
    • Google Scholar
13. 1. Fan CM
    2. Porter JA
    3. Chiang C
    4. Chang DT
    5. Beachy PA
    6. Tessier-Lavigne M

    (1995) Long-range sclerotome induction by sonic hedgehog: direct role of the amino-terminal cleavage product and modulation by the cyclic AMP signaling pathway

    Cell 81:457–465.

    https://doi.org/10.1016/0092-8674(95)90398-4

    • PubMed
    • Google Scholar
14. 1. Freedman BS
    2. Brooks CR
    3. Lam AQ
    4. Fu H
    5. Morizane R
    6. Agrawal V
    7. Saad AF
    8. Li MK
    9. Hughes MR
    10. Werff RV
    11. Peters DT
    12. Lu J
    13. Baccei A
    14. Siedlecki AM
    15. Valerius MT
    16. Musunuru K
    17. McNagny KM
    18. Steinman TI
    19. Zhou J
    20. Lerou PH
    21. Bonventre JV

    (2015) Modelling kidney disease with CRISPR-mutant kidney organoids derived from human pluripotent epiblast spheroids

    Nature Communications 6:8715.

    https://doi.org/10.1038/ncomms9715

    • PubMed
    • Google Scholar
15. 1. Greco TL
    2. Takada S
    3. Newhouse MM
    4. McMahon JA
    5. McMahon AP
    6. Camper SA

    (1996) Analysis of the vestigial tail mutation demonstrates that Wnt-3a gene dosage regulates mouse axial development

    Genes & Development 10:313–324.

    https://doi.org/10.1101/gad.10.3.313

    • PubMed
    • Google Scholar
16. 1. Hafemeister C
    2. Satija R

    (2019) Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression

    Genome Biology 20:296.

    https://doi.org/10.1186/s13059-019-1874-1

    • PubMed
    • Google Scholar
17. 1. Hubaud A
    2. Pourquié O

    (2014) Signalling dynamics in vertebrate segmentation

    Nature Reviews. Molecular Cell Biology 15:709–721.

    https://doi.org/10.1038/nrm3891

    • PubMed
    • Google Scholar
18. 1. Li CH
    2. Tam PKS

    (1998) An iterative algorithm for minimum cross entropy thresholding

    Pattern Recognition Letters 19:771–776.

    https://doi.org/10.1016/S0167-8655(98)00057-9

    • Google Scholar
19. 1. Loh KM
    2. Chen A
    3. Koh PW
    4. Deng TZ
    5. Sinha R
    6. Tsai JM
    7. Barkal AA
    8. Shen KY
    9. Jain R
    10. Morganti RM
    11. Shyh-Chang N
    12. Fernhoff NB
    13. George BM
    14. Wernig G
    15. Salomon REA
    16. Chen Z
    17. Vogel H
    18. Epstein JA
    19. Kundaje A
    20. Talbot WS
    21. Beachy PA
    22. Ang LT
    23. Weissman IL

    (2016) Mapping the Pairwise Choices Leading from Pluripotency to Human Bone, Heart, and Other Mesoderm Cell Types

    Cell 166:451–467.

    https://doi.org/10.1016/j.cell.2016.06.011

    • PubMed
    • Google Scholar
20. 1. Manfrin A
    2. Tabata Y
    3. Paquet ER
    4. Vuaridel AR
    5. Rivest FR
    6. Naef F
    7. Lutolf MP

    (2019) Engineered signaling centers for the spatially controlled patterning of human pluripotent stem cells

    Nature Methods 16:640–648.

    https://doi.org/10.1038/s41592-019-0455-2

    • PubMed
    • Google Scholar
21. 1. Matsuda M
    2. Yamanaka Y
    3. Uemura M
    4. Osawa M
    5. Saito MK
    6. Nagahashi A
    7. Nishio M
    8. Guo L
    9. Ikegawa S
    10. Sakurai S
    11. Kihara S
    12. Maurissen TL
    13. Nakamura M
    14. Matsumoto T
    15. Yoshitomi H
    16. Ikeya M
    17. Kawakami N
    18. Yamamoto T
    19. Woltjen K
    20. Ebisuya M
    21. Toguchida J
    22. Alev C

    (2020) Recapitulating the human segmentation clock with pluripotent stem cells

    Nature 580:124–129.

    https://doi.org/10.1038/s41586-020-2144-9

    • PubMed
    • Google Scholar
22. 1. Moris N
    2. Anlas K
    3. van den Brink SC
    4. Alemany A
    5. Schröder J
    6. Ghimire S
    7. Balayo T
    8. van Oudenaarden A
    9. Martinez Arias A

    (2020) An in vitro model of early anteroposterior organization during human development

    Nature 582:410–415.

    https://doi.org/10.1038/s41586-020-2383-9

    • PubMed
    • Google Scholar
23. 1. Oates AC
    2. Morelli LG
    3. Ares S

    (2012) Patterning embryos with oscillations: structure, function and dynamics of the vertebrate segmentation clock

    Development (Cambridge, England) 139:625–639.

    https://doi.org/10.1242/dev.063735

    • PubMed
    • Google Scholar
24. 1. Palla A
    2. Blau H

    (2020) The clock that controls spine development modelled in a dish

    Nature 580:32–34.

    https://doi.org/10.1038/d41586-020-00322-y

    • PubMed
    • Google Scholar
25. 1. Sakurai H
    2. Sakaguchi Y
    3. Shoji E
    4. Nishino T
    5. Maki I
    6. Sakai H
    7. Hanaoka K
    8. Kakizuka A
    9. Sehara-Fujisawa A

    (2012) In vitro modeling of paraxial mesodermal progenitors derived from induced pluripotent stem cells

    PLOS ONE 7:e47078.

    https://doi.org/10.1371/journal.pone.0047078

    • PubMed
    • Google Scholar
26. 1. Schindelin J
    2. Arganda-Carreras I
    3. Frise E
    4. Kaynig V
    5. Longair M
    6. Pietzsch T
    7. Preibisch S
    8. Rueden C
    9. Saalfeld S
    10. Schmid B
    11. Tinevez JY
    12. White DJ
    13. Hartenstein V
    14. Eliceiri K
    15. Tomancak P
    16. Cardona A

    (2012) Fiji: an open-source platform for biological-image analysis

    Nature Methods 9:676–682.

    https://doi.org/10.1038/nmeth.2019

    • PubMed
    • Google Scholar
27. 1. Stuart T
    2. Butler A
    3. Hoffman P
    4. Hafemeister C
    5. Papalexi E
    6. Mauck WM
    7. Hao Y
    8. Stoeckius M
    9. Smibert P
    10. Satija R

    (2019) Comprehensive Integration of Single-Cell Data

    Cell 177:1888–1902.

    https://doi.org/10.1016/j.cell.2019.05.031

    • PubMed
    • Google Scholar
28. 1. Tan JY
    2. Sriram G
    3. Rufaihah AJ
    4. Neoh KG
    5. Cao T

    (2013) Efficient derivation of lateral plate and paraxial mesoderm subtypes from human embryonic stem cells through GSKi-mediated differentiation

    Stem Cells and Development 22:1893–1906.

    https://doi.org/10.1089/scd.2012.0590

    • PubMed
    • Google Scholar
29. 1. Tonegawa A
    2. Funayama N
    3. Ueno N
    4. Takahashi Y

    (1997) Mesodermal subdivision along the mediolateral axis in chicken controlled by different concentrations of BMP-4

    Development (Cambridge, England) 124:1975–1984.

    https://doi.org/10.1242/dev.124.10.1975

    • PubMed
    • Google Scholar
30. 1. Traag VA
    2. Waltman L
    3. van Eck NJ

    (2019) From Louvain to Leiden: guaranteeing well-connected communities

    Scientific Reports 9:5233.

    https://doi.org/10.1038/s41598-019-41695-z

    • PubMed
    • Google Scholar
31. 1. Tzouanacou E
    2. Wegener A
    3. Wymeersch FJ
    4. Wilson V
    5. Nicolas JF

    (2009) Redefining the progression of lineage segregations during mammalian embryogenesis by clonal analysis

    Developmental Cell 17:365–376.

    https://doi.org/10.1016/j.devcel.2009.08.002

    • PubMed
    • Google Scholar
32. 1. van den Brink SC
    2. Baillie-Johnson P
    3. Balayo T
    4. Hadjantonakis A-K
    5. Nowotschin S
    6. Turner DA
    7. Martinez Arias A

    (2014) Symmetry breaking, germ layer specification and axial organisation in aggregates of mouse embryonic stem cells

    Development (Cambridge, England) 141:4231–4242.

    https://doi.org/10.1242/dev.113001

    • PubMed
    • Google Scholar
33. 1. Veenvliet JV
    2. Bolondi A
    3. Kretzmer H
    4. Haut L
    5. Scholze-Wittler M
    6. Schifferl D
    7. Koch F
    8. Guignard L
    9. Kumar AS
    10. Pustet M
    11. Heimann S
    12. Buschow R
    13. Wittler L
    14. Timmermann B
    15. Meissner A
    16. Herrmann BG

    (2020) Mouse embryonic stem cells self-organize into trunk-like structures with neural tube and somites

    Science (New York, N.Y.) 370:eaba4937.

    https://doi.org/10.1126/science.aba4937

    • PubMed
    • Google Scholar
34. 1. Xi H
    2. Fujiwara W
    3. Gonzalez K
    4. Jan M
    5. Liebscher S
    6. Van Handel B
    7. Schenke-Layland K
    8. Pyle AD

    (2017) In Vivo Human Somitogenesis Guides Somite Development from hPSCs

    Cell Reports 18:1573–1585.

    https://doi.org/10.1016/j.celrep.2017.01.040

    • PubMed
    • Google Scholar
35. 1. Yamaguchi TP
    2. Takada S
    3. Yoshikawa Y
    4. Wu N
    5. McMahon AP

    (1999) T (Brachyury) is a direct target of Wnt3a during paraxial mesoderm specification

    Genes & Development 13:3185–3190.

    https://doi.org/10.1101/gad.13.24.3185

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Paraxial mesoderm organoids model development of human somites" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by Marianne Bronner as the Senior and Reviewing Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted the letter below. As you will see, the reviewers raise some significant concerns, particularly regarding the morphological resemblance of somitoids to somites in vivo. If you feel you can address these concerns, we would be open to receiving a significantly revised version for full in depth review. Equally, we realize that you may prefer to go to a different journal at this juncture rather than make extensive revisions.

Essential revisions:

One of the great utilities of organoids is their self-organising ability which enables them to adopt morphological and molecular features that closely resemble their in vivo counterparts. While we appreciate that challenges of utilizing human cell lines, a major concern with the current work it that the human somitoids do not morphologically resemble somites. To remedy this:

1. In order to distinguish these from differentiating mesoderm, the authors need to provide solid morphological evidence that somitoids can generate somite-like structures. This would include demonstrating the presence of pseudo-stratified, bottle shaped epithelial cells, with apical/basal inside/out polarity.

The full reviews are included to help with a possible revision.

Reviewer #1 (Recommendations for the authors):

1) Three human iPSC lines are mentioned in the methods and replicates are mentioned throughout the text. The authors should make clear for each data set/figure presented which iPSC line(s) was used. Moreover, they should comment any between-line variability in the differentiation protocol. I can only find inter-organoid (so within-experiment) variability in the supplemental data. The extent of technical (between experiment) and biological variation is not clear.

2) It is not indicated clearly anywhere exactly what the authors are scoring as a somite. The clearest depiction is Figure 5A where rosettes of TCF15 with apical F actin can be observed. Is each of these scored as a somite? A higher magnification should be provided in Figure 3 showing clearly how somites are scored.

3) The combinatorial FGFRi/Wnti treatments using various small molecules mentioned in Figure 2C, D, 3B, C, D does not impact the overall understanding of the differentiation protocol. Moreover, data suggests that inhibition of FGF and Wnt signalling using small molecules is not required to derive PAX3+ somites. Just extending the CL+ FGF by 24hrs (day3 to day4) and then changing media to basal condition (no growth factors) for 24hrs is sufficient for differentiation. This condition also results in significantly higher PAX3+ cells and somite number (Figure 3C, D). Why was this condition not utilised further? Why is the use of FGFRi/Wnti essential?

4) d5 Somites are TCF15+ (Figure 4E, 5A) how different are these cells from the Somatic mesoderm population published in Diaz-Cuadros et al., 2020; Matsuda et al., 2020?

5) Figure1A and 2A: What is the rationale behind choosing D3 for modulating the FGFi/Wnti treatments. D4 aPSM (Figure 1A) are further committed towards somitogenesis and wouldn't modulating small molecule treatment provide a better output?

6) Figure 2C and D: 500 cells was chosen for all downstream experiments based on f-actin arrangement (line111). Authors need to significantly elaborate on this rationale. Does this relate to the definition of an in vitro somite? The data seem to suggest that using 500 cells (in 24hour condition) generates smaller organoids with no significant difference in PAX3 expression intensity.

7) Figure 4: Which differentiation condition was eventually used for sc-RNA seq analysis (Figure 4). Was it the CL+FGF condition or the FGFi/Wnti condition? This information is absolutely critical to determine the ultimate conclusion from Figure 1-3.

8) Figure 5: For sclerotome differentiation and other comparisons in this figure CL+FGF condition has been used (with no use of FGFi/Wnti for somite formation). With significant focus given on timed use of FGFi/Wnti for acquiring somite fate in Figure 1,2,3 why has the alternate method been used in this section?

9) Figure 5B and line 202, 2019-210: The in vitro organoid size is significantly larger than in vivo human stage9-10. Why has this difference been overlooked to claim "our organoid-derived somites share spatial and molecular features as well as overall size with their in vivo counterparts".

Reviewer #2 (Recommendations for the authors):

It seems therefore that despite the evident amount of work already placed in this study, this is rather incremental addition to existing knowledge rather than a breakthrough in the field.

Essential revisions:

One of the great utilities of organoids is their self-organising ability which enables them to adopt morphological and molecular features that closely resemble their in vivo counterparts. While we appreciate that challenges of utilizing human cell lines, a major concern with the current work it that the human somitoids do not morphologically resemble somites. To remedy this:

1. In order to distinguish these from differentiating mesoderm, the authors need to provide solid morphological evidence that somitoids can generate somite-like structures. This would include demonstrating the presence of pseudo-stratified, bottle shaped epithelial cells, with apical/basal inside/out polarity.

We thank the reviewers and editor for the feedback on our manuscript. We agree that it is important to clearly show the morphological features of our in vitro generated somites because it is one the main advances of our 3D Somitoid system over previously published 2D models. In the revised manuscript, we have included high-magnification images of somite-like structures, which clearly show the shape of the somitic cells and the polarity of these epithelial cells by staining for multiple polarity markers (new Figure 3-Supplemental Figure 1; new Figure 3-Video 1). Consistent with the morphology and polarity of in vivo somites, we observe PAX3+/TCF15+ bottle-shaped cells radially arranged around a central cavity, which form rosette-like structures that are approximately the same size as the Carnegie stage 11 in vivo somites (Figure 5B). This can be observed in multiple new images added to our revised manuscript (Figure 3-Supplemental Figure 1A-A’’, Figure 3-Supplemental Figure 1B-B’’, Figure 2-Supplemental Figure 3A,B). Additionally, apical surface markers F-ACTIN (Figure 5A) and N-CADHERIN (Figure 3-Supplemental Figure 1) are both expressed around the central cavity, suggesting that the apical side of the somitic cells is facing the inside of the somite structure, again consistent with their in vivo counterparts. This is particularly evident in the newly added high-magnification images (Figure 3-Supplemental Figure 1) and accompanying video showing a full confocal z-stack through the in vitro somites (Figure 3- Video 1). Together with our protein (Figure 3B,C and Figure 5A) and gene expression data (both in bulk (Figure 1C, Figure 5C, Figure 1-Supplemental Figure 1B) and at the single-cell level (Figure 4, Figure 4-Supplemental Figure 1-5)), and directed differentiation experiments of Somitoid-derived cells towards sclerotome (Figure 5C) and dermomyotome (newly added Figure 5-Supplemental Figure 1), we conclude that our in vitro somites are molecularly, morphologically, and functionally equivalent to in vivo somites.

The full reviews are included to help with a possible revision.

Reviewer #1 (Recommendations for the authors):

1) Three human iPSC lines are mentioned in the methods and replicates are mentioned throughout the text. The authors should make clear for each data set/figure presented which iPSC line(s) was used. Moreover, they should comment any between-line variability in the differentiation protocol. I can only find inter-organoid (so within-experiment) variability in the supplemental data. The extent of technical (between experiment) and biological variation is not clear.

The reviewer brings up a very important point. To clarify which cell lines were used for each figure, we have added the information to each figure caption in the revised manuscript. We have also specified the number of organoids used to quantify variation for each experiment in the figure captions. As described below, we have conducted new experiments to further quantify the technical variability across experiments, as well as quantify the efficacy of our optimized differentiation protocol when applied to genetically independent cell lines.

For our initial screens (Figures 2 and 3), we only used one cell line (NCRM1 hiPSCs). As mentioned by the reviewer, we measured variability across individual organoids for each condition of the screen to help identify the conditions that minimized variability and produced the most reproducible organoids (Figure 2-Supplemental Figure 2B; Figure 3-Supplemental Figure 1B).

To analyze technical variability of our optimized protocol, we have repeated our optimized protocol two more times and quantified the number of somites across 10 organoids for each experiment (see newly added Figure 3-Supplemental Figure 3). Inter-organoid variability was comparable to our initial results from the secondary screen (CV for Experiment 1 = 17% and CV for Experiment 2 = 9.8%; see also Figure 3-Supplemental Figure 2B). Furthermore, mean and median are very comparable between the two additional experiments (Experiment 1: 39+/-8 (mean+/-std), median=40; Experiment 2: 43+/-4 (mean+/-std), median=41, p-val = 0.16). Please note that the absolute number of somites in our new experiments has increased compared to our initial screen (see Figure 3). This is a result of improvements in both our immunostaining protocol as well as the image acquisition workflow. The two new experiments both used the same improved staining and image acquisition workflow and are therefore comparable with each other.

To extend our variability analysis to other cell lines, we tested the following cell lines: our original cell line (NCRM1), the WTC cell line released by the Conklin Laboratory at the Gladstone Institute, and a reporter cell line, ACTB-GFP, from the Allen Cell Collection. We tested our optimized protocol alongside another high scoring condition in all three cell lines:

– CL+FGF2 for 24h, basal media for 24h (our optimized protocol)

– WNTi^hi(C59, 2 µM) for 48h

Applying our optimized protocol to two other cell lines confirmed that our protocol is reproducible across different genetic backgrounds/cell lines. NCRM1, average somite number = 43+/-4 (mean+/-std); ACTB-GFP, average somite number = 40+/-6; WTC, average somite number = 33+/-4. We also tested one additional top scoring condition (C59, 2µM for 48 hours) in all three cell lines. Notably, for this condition, the ACTB-GFP derived Somitoids (32+/-4 (mean+/-std)) showed a higher average number of somites compared with the Somitoids derived from the other cell lines (NCRM1, 20+/-3; WTC, 17+/-4). However, our optimized protocol resulted in the highest number of somites across all cell lines.

2) It is not indicated clearly anywhere exactly what the authors are scoring as a somite. The clearest depiction is Figure 5A where rosettes of TCF15 with apical F actin can be observed. Is each of these scored as a somite? A higher magnification should be provided in Figure 3 showing clearly how somites are scored.

We thank the reviewer for an insightful comment and question. Following the suggestion of the reviewer, we generated new organoids with somite structures and imaged them at a higher magnification (shown in the newly added Figure 3-Supplement Figure 1). The higher magnification clearly shows the ‘bottle-shaped’ morphology of the cells that comprise the somites in that the apical surface of these columnar cells is typically smaller than their basal side. These higher-magnification images also more clearly show the rosette structure (radial arrangement of columnar cells) with a central cavity in which NCAD is highly expressed. Nuclear expression of PAX3 is also clearly visible in these cells. These empirical observations were the criteria used to manually identify somites in the images acquired from the organoid screen. We have clearly described these scoring criteria in a new paragraph added to the Methods section (section titled “Quantification of per organoid somite numbers for secondary screen”).

To further illustrate the scoring criteria for identifying somites, we have also included a new figure panel, Figure 3B. In this figure, we show examples of somite quantification using our scoring criteria. Two z-sections are used as examples and the manually scored somites are highlighted using dashed circles.

Additionally, we generated videos of z-stacks acquired on a confocal microscope showing examples of organoids with and without somite-like structures based on our scoring criteria using PAX3 and NCAD staining (Newly added Figure 3-Video 1).

3) The combinatorial FGFRi/Wnti treatments using various small molecules mentioned in Figure 2C, D, 3B, C, D does not impact the overall understanding of the differentiation protocol. Moreover, data suggests that inhibition of FGF and Wnt signalling using small molecules is not required to derive PAX3+ somites. Just extending the CL+ FGF by 24hrs (day3 to day4) and then changing media to basal condition (no growth factors) for 24hrs is sufficient for differentiation. This condition also results in significantly higher PAX3+ cells and somite number (Figure 3C, D). Why was this condition not utilised further? Why is the use of FGFRi/Wnti essential?

We apologize to the reviewer if this important point was not clear in the main text. We indeed end up using CL+FGF2 from Day 3 to Day 4 followed by culture in basal media as our preferred differentiation protocol for all subsequent experiments (shown in Figure 4-5), including our single-cell RNA-seq experiment. This optimized protocol is a direct conclusion from our systematic screens and was not obvious prior to performing these screens. We have revised the manuscript so that this point is clear to the reader (see line numbers 176-183 on page 6).

Our screens focused on modulating the timing and duration of WNT and FGF signaling pathways (in addition to BMP and TGF-β) based on long-established in vivo evidence (Aulehla et al., 2008, 2003; Delfini et al., 2005; Dubrulle et al., 2001; Dunty et al., 2008; Greco et al., 1996; Yamaguchi et al., 1999)(Loh et al., 2016; Xi et al., 2017)(Aulehla et al., 2008, 2003; Delfini et al., 2005; Dubrulle et al., 2001; Dunty et al., 2008; Greco et al., 1996; Yamaguchi et al., 1999) and recently published in vitro differentiation protocols (Chal et al., 2015; Loh et al., 2016; Matsuda et al., 2020; Sakurai et al., 2012; Tan et al., 2013). To make the rationale for designing the screens more clear, we describe it in the revised manuscript and include citations for in vivo evidence and in vitro differentiation protocols using hPSCs (see line numbers 97-105 on page 4).

Importantly, one main conclusion from the screen and our in-depth single-cell RNA-seq analysis is that while it is not necessary to actively inhibit WNT and/or FGF signaling to induce somite fate in vitro, day 5 somite-like cells from our optimized protocol autonomously downregulate WNT target genes (DKK1, AXIN2, WNT3A, RSPO3; Figure 4-Supplemental Figure 2) and FGF target genes (FGF8, FGF17, SPRY4, DUSP6/MKP3, SEF/IL17RD) in addition to NOTCH target genes (DLL1, DLL3, HES7). To illustrate this, we have generated a heatmap of single-cell gene expression of canonical target genes for these signaling pathways (see newly added Figure 4-Supplemental Figure 5) over the course of the differentiation protocol.

4) d5 Somites are TCF15+ (Figure 4E, 5A) how different are these cells from the Somatic mesoderm population published in Diaz-Cuadros et al., 2020; Matsuda et al., 2020?

The reviewer raises a valid and important point. To address the reviewer’s question, we compared the single-cell RNA-seq data from Diaz-Cuadros et al., with our own single-cell RNA-seq data. The 2D differentiated cells in Diaz-Cuadros et al., at the final time point of their experiment do not show a clear somitic cell state signature. PAX3, MEOX1, MEOX2, FOXC2, UNCX, and TBX18 are not expressed (compared for example to FOXC1). FOXC1 does not appear to be a somite-specific marker as it is expressed in early and late PSM-like cells as seen in our own data, starting on day 2 (Figure 4-Supplemental Figure 3). In the Diaz-Cuadros et al., dataset, TCF15 is not expressed uniformly in all cells nor specifically in the late-stage cells (see UMAP plots and single-cell heatmap in the newly generated figure). Conversely, TCF15 is specifically expressed in day 5 somitic cells in our dataset both in a uniform manner and at high levels (as shown in Author response image 1 and Figure 4C).

Author response image 1

Download asset Open asset

We also analyzed the bulk RNA-seq data in Matsuda et al., 2020 as shown in Figure 1B of their paper. They show the expression of 4 somite markers (TCF15, MEOX1, PAX3, and RIPPLY1). As can be seen in Author response image 1 (and Figure 4-Supplemental Figure 3), all markers highlighted in their paper are specifically expressed in our day 5 somitic cell population, including RIPPLY1 (see also updated Figure 4-Supplemental Figure 3). Beyond comparison of marker gene expression, it is difficult to assess the similarities and differences with the Matsuda dataset since their data lacks single-cell resolution. Thus, heterogeneity and efficiency of somitic fate induction within their cell population is unclear.

Finally, neither of these two papers report the formation of somite-like structures. To make this comparison more clear, we have added it to the Discussion section of the revised manuscript (see line numbers 291-304 on page 10).

5) Figure1A and 2A: What is the rationale behind choosing D3 for modulating the FGFi/Wnti treatments. D4 aPSM (Figure 1A) are further committed towards somitogenesis and wouldn't modulating small molecule treatment provide a better output?

This is a valid point by the reviewer which we considered when designing the screen. Based on the 2D data from Diaz-Cuadros et al., D3 and D4 cells have similar gene expression profiles. Their own analysis of clustering the single-cell gene expression profile of these two populations of cells does not distinguish D3 cells from D4 cells. This is not surprising because D3 and D4 cells are kept in the same media containing FGF and WNT agonists. In vivo, down-regulation of FGF and WNT signaling is correlated with differentiation to somite fate. In addition, in their data both D3 and D4 cells show HES7 oscillations, indicating that D4 cells are not significantly more differentiated than D3 cells. Finally, the comparison in Diaz-Cuadros et al., of the gene expression profiles of D3 and D4 cells to E9.5 mouse in vivo data shows that both populations closely resemble the posterior PSM and that D4 cells are no closer to an anterior PSM fate than D3 cells. Taken together, we chose to start our differentiation protocol with D3 cells, which also reduces the duration of the protocol by one day. Our earlier version of Figure 1 was therefore misleading and we have changed the labels on Figure 1A to ‘PSM’ to not falsely suggest that cells progress to a more anterior PSM fate under these conditions.

We have also considered whether D2 organoids could already be efficiently differentiated towards somite fate. Although D2 organoids expressed PSM markers, such as TBX6 and MSGN1 (Figure 1B), the expression was not as uniform as that observed in D3 organoids. This observation is consistent with our single-cell RNA-seq data. Although both D2 and D3 organoids express PSM makers, D2 organoids also still express primitive streak markers at higher levels such as T/BRA, MIXL1, POU5F1. Additionally, WNT target genes such as WNT3A and DKK1, which are markers of posterior PSM cells, are more highly expressed in D2 organoids compared with D3 organoids (Figure 4C, Figure 4-Supplemental Figure 2). In conclusion, based on both previously published data and our own single-cell RNA-seq and qPCR data, we concluded that day 3 of our differentiation protocol constitutes the best time window to optimize somite differentiation in our organoid system.

6) Figure 2C and D: 500 cells was chosen for all downstream experiments based on f-actin arrangement (line111). Authors need to significantly elaborate on this rationale. Does this relate to the definition of an in vitro somite? The data seem to suggest that using 500 cells (in 24hour condition) generates smaller organoids with no significant difference in PAX3 expression intensity.

We thank the reviewer for raising an important point. One of the main goals of our primary screen was to investigate whether initial cell number has any effect on the ability of organoids to robustly form somites without the generation of undesired (i.e. non-PSM derived) cell types. We concluded that organoids made from 500 cells were preferable over organoids made from 1000 cells based on the following observations:

1. We used F-ACTIN expression as a marker of somite-structure in the primary screen comparing organoids made from 500 cells with organoids made from 1000 cells. Qualitative assessment of F-ACTIN staining showed more consistent somite structures (radial arrangements of PAX3+ columnar cells with high expression of F-ACTIN in the central cavity) in organoids made from 500 cells. To make this more clear to the reader, we have included representative images showing differences of F-ACTIN expression across organoids made from 500 cells and 1000 cells in the newly added Figure 2-Supplemental Figure 3.

2. Although average PAX3 intensity is similar between organoids made from 500 cells and organoids made from 1000 cells, a higher fraction of PAX3-negative cells were observed in the organoids made from 1000 cells. We only made this assessment using qualitative observation by eye because single-cell segmentation of our images was not possible. To do so, additional nuclear or membrane markers would be required. Figure 2-Supplemental Figure 2A shows examples of organoids generated from 1000 cells which show patches of PAX3-negative cells. By contrast, organoids generated from 500 cells and treated with the same small-molecule inhibitors show more uniform PAX3 induction across the organoids.

3. Finally, we observed sox2⁺/PAX3- cells in organoids made from 1000 cells but did not observe these contaminating cells in organoids made from 500 cells. Figure 2-Supplemental Figure 2A shows examples of organoids from 1000 cells with patches of PAX3-negative cells. These cells are positive for SOX2, suggesting neural contaminants in organoids generated from 1000 cells. Organoids generated from 500 cells using the same treatment conditions, do not show sox2⁺ patches of cells.

Taken together, we concluded that organoids made from 500 cells were a better choice than organoids made from 1000 cells. To make this clear in the revised manuscript, we carefully outline our reasoning on page 4-5 (line numbers 118-129).

7) Figure 4: Which differentiation condition was eventually used for sc-RNA seq analysis (Figure 4). Was it the CL+FGF condition or the FGFi/Wnti condition? This information is absolutely critical to determine the ultimate conclusion from Figure 1-3.

We thank the reviewer for pointing out that the original version of the manuscript did not explicitly state which protocol was used for our scRNA-seq experiment. In the revised manuscript, we have added an explicit statement saying which protocol we employed for our single-cell RNA-seq analysis. Based on the number of somites generated, the low variation of somite numbers between organoids and across experiments, as well as high average PAX3+ levels, we chose the CL+FGF2 and basal media condition for all subsequent experiments following the screen, including the single-cell RNA-seq analysis (Figure 4) and the sclerotome/dermomyotome differentiation protocols (Figure 5 and Figure 5-Supplemental Figure 1).

8) Figure 5: For sclerotome differentiation and other comparisons in this figure CL+FGF condition has been used (with no use of FGFi/Wnti for somite formation). With significant focus given on timed use of FGFi/Wnti for acquiring somite fate in Figure 1,2,3 why has the alternate method been used in this section?

Please refer to the previous response on why the CL+FGF/basal condition was used for sclerotome differentiation.

9) Figure 5B and line 202, 2019-210: The in vitro organoid size is significantly larger than in vivo human stage9-10. Why has this difference been overlooked to claim "our organoid-derived somites share spatial and molecular features as well as overall size with their in vivo counterparts".

In the manuscript, we compare the in vitro somite size to its in vivo counterpart in the Carnegie collection. The reviewer is absolutely correct that we do observe differences in the size of the in vitro generated somites and their stage 9-10 in vivo counterparts (as quantified in Figure 5B). To further expand our in vivo comparison, we quantified somite size in the later stages of the Carnegie collection up to stage 11. We were not able to quantify somite size beyond stage 11 because later stages already showed further differentiation of somites into sclerotome and dermomyotome and therefore did not have a well-defined rosette structure to quantify. Comparing our in vitro generated somites to stage 11 human somites, we observed that they are very similar in size (see updated Figure 5B). It is therefore conceivable that our somites correspond to later stages of the Carnegie collection, at least based on their relative sizes. We have updated Figure 5 and the corresponding description of the data in the main text to reflect the updated analysis.

Reviewer #2 (Recommendations for the authors):

It seems therefore that despite the evident amount of work already placed in this study, this is rather incremental addition to existing knowledge rather than a breakthrough in the field.

We are glad the reviewer found our study interesting and a step towards formation of complex structures in vitro. We believe that the reviewer has misunderstood the structure of the somites that we observe in vitro. In the revised manuscript, we have included high magnification images that more clearly show the shape of the cells and the locations at which the epithelial polarity markers are expressed (new Figure 3-Supplemental Figure 1; new Figure 3-Video 1). Consistent with in vivo somites, we do observe bottle-shaped cells radially arranged around a central cavity forming rosette-like structures that are the same size as their in vivo counterparts (Figure 5B). This can be observed in multiple images shown in our manuscript (Figure 3-Supplemental Figure 1A-A’’, Figure 3-Supplemental Figure 1B-B’’, Figure 2-Supplemental Figure 3A,B). Importantly, both F-ACTIN (Figure 5A) and N-CADHERIN (Figure 3-Supplemental Figure 1) are indeed expressed around the central cavity, suggesting that the apical side of the PAX3+/TCF15+ somitic cells is facing the inside of the somite structures. This is especially evident in the newly added high-magnification images (Figure 3-Supplemental Figure 1) and accompanying video showing a full confocal z-stack through the in vitro somites (Figure 3-Video 1). Taken together with our protein (Figure 3B,C and Figure 5A) and gene expression data (both in bulk (Figure 1C, Figure 5C, Figure 1-Supplemental Figure 1B)) and at the single-cell level (Figure 4, Figure 4-Supplemental Figure 1-5) and our directed differentiation experiments of our Somitoid cells to dermomyotome (Figure 5-Supplemental Figure 1) and sclerotome fates (Figure 5C), we believe that the somite structures generated by our optimized in vitro protocol are indeed equivalent to their in vivo counterparts.

References

Aulehla A, Wehrle C, Brand-Saberi B, Kemler R, Gossler A, Kanzler B, Herrmann BG. 2003. Wnt3a plays a major role in the segmentation clock controlling somitogenesis. Dev Cell 4:395–406. doi:10.1016/S1534-5807(03)00055-8

Aulehla A, Wiegraebe W, Baubet V, Wahl MB, Deng C, Taketo M, Lewandoski M, Pourquié O. 2008. A β-catenin gradient links the clock and wavefront systems in mouse embryo segmentation. Nat Cell Biol 10:186–193. doi:10.1038/ncb1679

Chal J, Oginuma M, Al Tanoury Z, Gobert B, Sumara O, Hick A, Bousson F, Zidouni Y, Mursch C, Moncuquet P, Tassy O, Vincent S, Miyanari A, Bera A, Garnier J-M, Guevara G, Hestin M, Kennedy L, Hayashi S, Drayton B, Cherrier T, Gayraud-Morel B, Gussoni E, Relaix F, Tajbakhsh S, Pourquié O. 2015. Differentiation of pluripotent stem cells to muscle fiber to model Duchenne muscular dystrophy. Nat Biotechnol 33:962–969. doi:10.1038/nbt.3297

Delfini M-C, Dubrulle J, Malapert P, Chal J, Pourquié O. 2005. Control of the segmentation process by graded MAPK/ERK activation in the chick embryo. Proceedings of the National Academy of Sciences 102:11343–11348. doi:10.1073/pnas.0502933102

Dubrulle J, McGrew MJ, Pourquié O. 2001. FGF Signaling Controls Somite Boundary Position and Regulates Segmentation Clock Control of Spatiotemporal Hox Gene Activation. Cell 106:219–232. doi:10.1016/S0092-8674(01)00437-8

Dunty WC Jr, Biris KK, Chalamalasetty RB, Taketo MM, Lewandoski M, Yamaguchi TP. 2008. Wnt3a/β-catenin signaling controls posterior body development by coordinating mesoderm formation and segmentation. Development 135:85–94. doi:10.1242/dev.009266

Greco TL, Takada S, Newhouse MM, McMahon JA, McMahon AP, Camper SA. 1996. Analysis of the vestigial tail mutation demonstrates that Wnt-3a gene dosage regulates mouse axial development. Genes Dev 10:313–324. doi:10.1101/gad.10.3.313

Loh KM, Chen A, Koh PW, Deng TZ, Sinha R, Tsai JM, Barkal AA, Shen KY, Jain R, Morganti RM, Shyh-Chang N, Fernhoff NB, George BM, Wernig G, Salomon REA, Chen Z, Vogel H, Epstein JA, Kundaje A, Talbot WS, Beachy PA, Ang LT, Weissman IL. 2016. Mapping the Pairwise Choices Leading from Pluripotency to Human Bone, Heart, and Other Mesoderm Cell Types. Cell 166:451–467. doi:10.1016/j.cell.2016.06.011

Matsuda M, Yamanaka Y, Uemura M, Osawa M, Saito MK, Nagahashi A, Nishio M, Guo L, Ikegawa S, Sakurai S, Kihara S, Maurissen TL, Nakamura M, Matsumoto T, Yoshitomi H, Ikeya M, Kawakami N, Yamamoto T, Woltjen K, Ebisuya M, Toguchida J, Alev C. 2020. Recapitulating the human segmentation clock with pluripotent stem cells. Nature 580:124–129. doi:10.1038/s41586-020-2144-9

Sakurai H, Sakaguchi Y, Shoji E, Nishino T, Maki I, Sakai H, Hanaoka K, Kakizuka A, Sehara-Fujisawa A. 2012. in vitro modeling of paraxial mesodermal progenitors derived from induced pluripotent stem cells. PLoS One 7:e47078. doi:10.1371/journal.pone.0047078

Tan JY, Sriram G, Rufaihah AJ, Neoh KG, Cao T. 2013. Efficient derivation of lateral plate and paraxial mesoderm subtypes from human embryonic stem cells through GSKi-mediated differentiation. Stem Cells Dev 22:1893–1906.

Xi H, Fujiwara W, Gonzalez K, Jan M, Liebscher S, Van Handel B, Schenke-Layland K, Pyle AD. 2017. in vivo Human Somitogenesis Guides Somite Development from hPSCs. CellReports 18:1573–1585.

Yamaguchi TP, Takada S, Yoshikawa Y, Wu N, McMahon AP. 1999. T (Brachyury) is a direct target of Wnt3a during paraxial mesoderm specification. Genes & Development. doi:10.1101/gad.13.24.3185

Author details

1. Christoph Budjan

   1. Department of Systems Biology, Harvard Medical School, Boston, United States
   2. Department of Data Science, Dana-Farber Cancer Institute, Boston, United States

   Contribution

   designed, optimized and performed the experiments; analyzed the experimental and single-cell RNA-sequencing data, Conceptualization, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1563-9750

2. Shichen Liu

   1. Department of Systems Biology, Harvard Medical School, Boston, United States
   2. Department of Data Science, Dana-Farber Cancer Institute, Boston, United States

   Contribution

   Methodology, assisted with single-cell sample collection and sequencing library preparation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0964-6559

3. Adrian Ranga

   Laboratory of Bioengineering and Morphogenesis, Biomechanics Section, Department of Mechanical Engineering, Leuven, Belgium

   Contribution

   Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6400-9472

4. Senjuti Gayen

   1. Department of Systems Biology, Harvard Medical School, Boston, United States
   2. Department of Data Science, Dana-Farber Cancer Institute, Boston, United States

   Contribution

   Formal analysis

   Competing interests

   No competing interests declared

5. Olivier Pourquié

   1. Department of Genetics, Harvard Medical School, Boston, United States
   2. Department of Pathology, Brigham and Women's Hospital, Boston, United States

   Contribution

   Funding acquisition, Supervision

   Competing interests

   No competing interests declared

6. Sahand Hormoz

   1. Department of Systems Biology, Harvard Medical School, Boston, United States
   2. Department of Data Science, Dana-Farber Cancer Institute, Boston, United States
   3. Broad Institute of MIT and Harvard, Cambridge, United States

   Contribution

   Funding acquisition, Supervision, Writing – original draft, Writing – review and editing

   For correspondence

   sahand_hormoz@hms.harvard.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4384-4428

• 3,066

  Page views

• 461

  Downloads

• 13

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.