The function of the musculoskeletal system relies on the proper assemblage of its components, namely skeletal tissues (bone, cartilage, and joints), muscles, and tendons. However, the attachment of tissues composed of materials with large differences in their mechanical properties is highly challenging. In the musculoskeleton, elastic tendons, which have a Young’s modulus (a measure of stiffness) in the order of 200 megapascal, are attached to the much harder bone, with a modulus in the order of 20 gigapascal. This disparity makes the connection between these two tissues a mechanical weak point, which is subject to higher incidence of tearing by both external and internal forces acting on the musculoskeleton during movement. The evolutionary solution to this problem is the enthesis, a transitional tissue that displays a gradual shift in cellular and extracellular properties from the tendon side through to the bone side (Genin et al., 2009; Liu et al., 2011; Liu et al., 2012a; Lu and Thomopoulos, 2013; Thomopoulos et al., 2003). Yet, despite its importance, the formation of this cellular gradient as well as the underlying molecular mechanism remain largely unknown.

In recent years, the initial events that lead to the formation of the embryonic attachment unit (AU), which serves as the primordium of the enthesis, have started to be investigated. These studies identified the progenitors of the AU and showed that they express both the chondrogenic and tenogenic transcription factors Sox9 and scleraxis (Scx), respectively (Blitz et al., 2013; Sugimoto et al., 2013). The patterning of the Sox9⁺/Scx⁺ progenitors along the skeleton is regulated by a genetic program that includes several transcription factors (Eyal et al., 2019). Next, Sox9⁺/Scx⁺ progenitors differentiate to chondrocytes, which form a bone eminence on one side, or to tenocytes, which form the tendon on the other side, whereas the cells in between differentiate into Gli1⁺ cells that eventually will form the enthesis (Felsenthal et al., 2018; Schwartz et al., 2017; Schwartz et al., 2015).

Both molecular and mechanical signals regulate the AU. TGFβ signaling regulates the specification of AU progenitors, whereas BMP and FGF signaling as well as mechanical signals determine their fate and differentiation (Blitz et al., 2013; Blitz et al., 2009; Roberts et al., 2019). Postnatal enthesis cells have been termed fibrocartilage cells based on their histological appearance, because they display morphological features that are shared with tenocytes and chondrocytes (Thomopoulos et al., 2010). In recent years, several studies have identified some of the genes that these cells express, including collagens type I, II, and X; Indian hedgehog (Ihh); parathyroid hormone-related peptide (Pthlh); patched 1 (Ptc1); runt-related transcription factor 2 (Runx2); tenascin C (Tnc); and biglycan (Bgn) (Liu et al., 2012b; Liu et al., 2013; Liu et al., 2018; Thomopoulos et al., 2010). Interestingly, these genes are also expressed by cells in the neighboring tissues, namely by chondrocytes or tenocytes. However, despite these advances, a comprehensive molecular signature of this tissue and the mechanism that enables its formation are still missing.

In this work, we aimed to decipher the identity of the fibrocartilage cells that form the attachment tissue between tendon and bone. Bulk and single-cell transcriptomic analyses of the attachment cells, which were validated by in situ hybridization (ISH) and single-molecule fluorescent ISH (smFISH), showed that these cells express a mix of the transcriptomes of chondrocytes and tenocytes. Chromatin analysis further verified the transcriptomic results and provided a mechanistic explanation for the bi-fated behavior of attachment cells, which share enhancers with their neighboring tenocytes or chondrocytes. Finally, we identify the transcription factors KLF2 and KLF4 as regulators of attachment cell differentiation. Overall, we provide the transcriptional as well as the epigenetic mechanism that allows attachment cells to activate a combination of cartilage and tendon transcriptomes and, thereby, the formation of the unique transitional tissue.

In this work, we describe the unique transcriptome that allows cells of the attachment between tendon and bone to act as a transitional tissue. The ability to activate a combination of chondrogenic and tenogenic transcriptomes is regulated by sharing enhancers with these cells. Finally, we identify the transcription factors KLF2/4 as regulators of these unique bi-fated cells.

The existence of borders between tissues that differ in cell type, extracellular matrix composition, structure, and function raises the question of how tissues are connected. The border can be sharp, as seen in blood vessels, where pericytes and endothelial cells are separated by a basement membrane, or between the esophagus and the stomach in the gastrointestinal tract (Bergers and Song, 2005; San Roman and Shivdasani, 2011). On the other hand, the border can be less defined histologically and molecularly, thus forming a transitional tissue. Examples for the latter are the borders between the sections of the small intestine and between tendon and bone (Liu et al., 2007; Romih et al., 2005; San Roman and Shivdasani, 2011). From a broader perspective, as all organs and systems are made of different tissues, understanding the biology of border tissues is imperative. Moreover, some of these border tissues are involved in various pathologies. For example, gastric cancers may emerge from distinct anatomical areas, such as the esophagus–stomach boundary (Chawengsaksophak, 2019; McDonald et al., 2015; San Roman and Shivdasani, 2011).

In the case of the attachment between tendon and bone, the significance of this tissue is demonstrated by enthesopathies, a collective name for injuries and pathologies of the enthesis. For example, over 30% of the population over the age of 60 will injure their shoulder’s rotator cuff (Lehman et al., 1995). Failure rates of surgical reattachment range from 20% for small tears to 94% for repair of massive tears (Galatz et al., 2004; Harryman et al., 1991). The high failure and recurrence rates of these procedures highlight the need for understanding the biology of this complex transitional tissue of the enthesis. This understanding may allow the development of new strategies to improve the treatment of enthesopathies.

There are two options to form a transitional tissue. The first strategy is by mixing cells from the two neighboring tissues, such as in the epithelia of the urinary tract (Romih et al., 2005). Alternatively, the border cells can express a mixture of the transcriptomes of the two neighboring cell types. As we show here, the attachment cells represent the latter strategy well, as they express a high number of genes that are differentially expressed by either tenocytes or chondrocytes. These cells display morphological features that are shared with tenocytes and chondrocytes (Thomopoulos et al., 2010). Our results therefore provide a molecular explanation for the age-old histological definition of enthesis cells as fibrocartilage, which was based on their morphology (Thomopoulos et al., 2010). Moreover, the finding of mixed matrix genes in the transcriptome of the attachment cells may provide a mechanism for the formation of a transitional tissue, which allows safe transfer of forces by the tendon between muscle and bone.

In addition to expression of chondrogenic and tenogenic genes, we identified genes that are uniquely expressed by attachment cells. These genes may provide another level of specificity to the regulation of the development of this unique tissue. Finally, our scRNA- seq of the attachment cells revealed heterogeneity, which was a consequence of varying levels of chondrogenic and tenogenic gene expression. This heterogeneity may represent the differentiation processes that the Sox9/Scx-positive progenitors undergo during development, ultimately leading to their terminal cell fate.

Our finding that attachment cells are bi-fated raises the question of the mechanism that underlies this fate. An immediate implication of our finding is that there must be an epigenetic mechanism that supports the bi-fated state. The observed chromatin accessibility at the sites of the promoters of most of the shared genes in all three cell types rules out the possibility of limited promoter accessibility as the main mechanism. By contrast, the high percentage of shared accessible intergenic sites between attachment cells and one group of flanking cells, that is chondrocytes or tenocytes, suggests that this is the main mechanism. Moreover, many of these shared sites correlated with putative enhancers in ENCODE datasets. Finally, we verified the ability of three different enhancers from that list to drive gene expression in attachment cells and in either tendon or cartilage. These findings strongly support our hypothesis that the regulatory mechanism is based on the ability of attachment cells to share enhancers with either chondrocytes or tenocytes in order to drive the mixed expression profile of these bi-fated cells. It is important to emphasize that we have also identified a group of intergenic elements that were accessible exclusively in attachment cells. These attachment-specific elements may act as enhancers that drive the expression of genes that are specific to attachment cells. It is, of course, possible that they participate in the regulation of shared genes as well. Overall, both sets of putative enhancers, namely shared and attachment cell-specific, may play important roles in the genetic program that regulates the development of this unique tissue.

Sharing enhancers is not the only possible strategy for the generation of a mixed transcriptome. A simple alternative would be a specific set of enhancers to be used by the attachment cells. A possible explanation for the sharing strategy is the common origin of all these cells, which is limb mesenchyme originating from lateral plate mesoderm (Johnson and Tabin, 1997). It is possible that during development, limb mesenchymal progenitors display highly accessible chromatin; yet, during differentiation, this accessibility is restricted to prevent the expression of genes from alternate lineages. In contrast to this restriction process, in the bi-fated attachment cells the shared sites are maintained accessible to allow the expression of the mixed transcriptome. A mechanism for silencing of genes of alternate lineages was previously described. For example, polycomb-repressed chromatin leads to silencing of genes of alternate lineages, leading to the commitment to a specific cell fate (Aldiri et al., 2017; Laugesen and Helin, 2014; Zhu et al., 2013). Interestingly, previous studies demonstrated the importance of chromatin repression in the developing limb, showing how deletion of Ezh2, which acts as the enzymatically active subunit of PRC2, leads to skeletal malformations (Deimling et al., 2018). This obviously raises the question of the mechanism that prevents this silencing in attachment cells.

It is clear that we cannot exclude the possibility that an active mechanism, such as the SWI/SNF remodeling complexes, opens the chromatin structure in bi-fated cells to allow attachment cell dual behavior (Hu et al., 2011). However, such a mechanism cannot explain why the strategy of shared enhancers was selected. Finally, a mixed transcriptome is one of the hallmarks of stem cell pluripotency, as progenitor cells eventually chose one fate over the other upon differentiation (Johnson et al., 2015; Soldatov et al., 2019). Our findings suggest a different strategy, where using both programs facilitates the establishment of a specific attachment tissue. Overall, our results reveal a novel function for chromatin state, which allows the activation of two sets of genes in a third cell type to create a new cell fate that forms a transitional tissue.

KLF2 and KLF4 are known to regulate several biological processes, such as promoting the differentiation of gut and skin (KLF4, [Katz et al., 2002; Segre et al., 1999]) as well as the immune system (KLF2, [Kuo et al., 1997]), maintaining pluripotency of embryonic stem cells (KLF2 and KLF4, [Jiang et al., 2008]), and, together with other factors, inducing pluripotency to generate iPSC by reprogramming (KLF4, [Takahashi and Yamanaka, 2006]). Several works describe the involvement of KLF2 and KLF4 in the musculoskeletal system. In bones, Klf4 over-expression in osteoblasts caused delayed bone development, in addition to impaired blood vessel invasion and osteoclast recruitment (Michikami et al., 2012). Another study showed that KLF2/4 are expressed during chick limb development in tendons and ligaments as part of the genetic program that regulates connective tissues (Orgeur et al., 2018). Previous studies showed that KLF2 and KLF4 display high similarity in protein sequences (Dang et al., 2000; Shields and Yang, 1997), suggesting that these factors could have common target sequences and may be functionally redundant. Indeed, loss of both KLF2 and KLF4 during embryogenesis led to abnormal blood vessel development and early lethality. This phenotype was more severe than what was observed in embryos that lost only KLF2 or KLF4 (Chiplunkar et al., 2013). Furthermore, previous work identified 313 target genes shared between KLF2 and KLf4, suggesting that they overlap in regulating gene expression (Orgeur et al., 2018). In this work, we show that KLF2/4 are central regulators of the attachment site. While the attachment did form initially in their absence, the subsequent differentiation failed, suggesting that KLF2/4 play a role at this stage. While we concentrated in this study on the attachment site, it is most likely that KLF2/4 play a role also in other musculoskeletal tissues such as the skeleton, tendon, and muscle. This possibility is supported by previous studies, where KLF2/4 were shown to be expressed in osteoblasts, chondrocytes, tenocytes, and muscle connective tissues (Michikami et al., 2012; Orgeur et al., 2018). Previous studies demonstrate the role of muscle-induced mechanical load in the development of attachment site (Blitz et al., 2009). In that context, our finding that KLF2/4 regulate the differentiation of attachment cells is interesting, because previous works have shown that these factors are mechanically regulated. It was shown in mice that shear stress on the vessels induced by blood flow leads to upregulation of Klf2 expression (Lee et al., 2006). Additional in vitro studies showed that KLF2 and KLF4 are influenced by shear stress (Chiplunkar et al., 2013; Dekker et al., 2005; Villarreal et al., 2010). It is therefore possible that these factors are regulated by muscle forces, leading to the proper differentiation and maturation of the attachment site.

The ability of KLF2/4 to regulate gene expression in the attachment site is supported by our finding that many of the genes that were expressed by attachment cells had in their regulatory region KLF2/4 binding sites. Yet, it is clear that not all of them share this property, suggesting that these two factors are part of a larger transcriptional network. For example, our bioinformatic analysis identified other TF families such as GLI’s, RUNX’s, and NFI’s as regulators of the attachment sites. Gli1 was previously reported as a marker for enthesis cells (Dyment et al., 2015; Felsenthal et al., 2018; Liu et al., 2012a; Schwartz et al., 2015). Since gene expression by attachment cells is regulated by sharing enhancers with chondrocytes or tenocytes, it is reasonable to assume that some regulators of these cells might be part of the network that regulates the attachment cells. Indeed, loss of the tendon regulator Scx in mice led to failure of attachment cells to differentiate. Additionally, loss of the chondrogenic regulator Sox9 in Scx-expressing cells led to failure in attachment site formation (Blitz et al., 2013; Blitz et al., 2009).

To conclude, by characterizing the transcriptome and chromatin landscape of tendon-to-bone attachment cells, we provide a molecular understanding of the bi-fated identity of these cells. Moreover, by identifying the transcription factors KLF2/4 as central regulators and the strategy of sharing enhancers with either tenocytes or chondrocytes, we provide a mechanism that regulates these bi-fated cells (Figure 6). These findings present a new concept for the formation of a border tissue, which is based on the simultaneous expression of a mixed transcriptome of the two flanking cell types by the intermediate cells. This strategy allows the formation of a unique transitional tissue without developing de novo a dedicated genetic program that regulates a third, new cell fate.

Figure 6

Download asset Open asset

For Key Resources Table, see Appendix.

The data was submitted to GEO (GSE144306). Additional data of scRNA-seq was submitted to GEO (GSE160090).

1. 1. Kult S
   2. Olender T
   3. Osterwalder M
   4. Markman S
   5. Leshkowitz D
   6. Krief S
   7. Blecher-Gonen R
   8. Ben-Moshe S
   9. Farack L
   10. Keren-Shaul H
   11. Meir Salame T
   12. Capellini TD
   13. Itzkovitz S
   14. Amit I
   15. Visel A
   16. Zelzer E

   (2020) NCBI Gene Expression Omnibus

   ID GSE144306. Bi-fated tendon-to-bone attachment cells are regulated by shared enhancers and KLF transcription factors.

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE144306
2. 1. Kult S
   2. Olender T
   3. Osterwalder M
   4. Markman S
   5. Leshkowitz D
   6. Krief S
   7. Blecher-Gonen R
   8. Ben-Moshe S
   9. Farack L
   10. Keren-Shaul H
   11. Meir Salame T
   12. Capellini TD
   13. Itzkovitz S
   14. Amit I
   15. Visel A
   16. Zelzer E

   (2020) NCBI Gene Expression Omnibus

   ID GSE160090. Bi-fated tendon-to-bone attachment cells are regulated by shared enhancers and KLF transcription factors.

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE160090

1. 1. Ren B
   2. UCSD

   (2015) ENCODE

   ID ENCSR905FFU. H3K27ac ChIP-seq on embryonic 13.5 day mouse limb.

   https://www.encodeproject.org/experiments/ENCSR905FFU/
2. 1. Ren B
   2. UCSD

   (2015) ENCODE

   ID ENCSR426EZM. H3K4me1 ChIP-seq on embryonic 13.5 day mouse limb.

   https://www.encodeproject.org/experiments/ENCSR426EZM/
3. 1. Ren B
   2. UCSD

   (2015) ENCODE

   ID ENCSR416OYH. H3K4me3 ChIP-seq on embryonic 13.5 day mouse limb.

   https://www.encodeproject.org/experiments/ENCSR416OYH/
4. 1. Ren B
   2. UCSD

   (2015) ENCODE

   ID ENCSR022DED. H3K9me3 ChIP-seq on embryonic 13.5 day mouse limb.

   https://www.encodeproject.org/experiments/ENCSR022DED/

1. 1. Aldiri I
   2. Xu B
   3. Wang L
   4. Chen X
   5. Hiler D
   6. Griffiths L
   7. Valentine M
   8. Shirinifard A
   9. Thiagarajan S
   10. Sablauer A
   11. Barabas M-E
   12. Zhang J
   13. Johnson D
   14. Frase S
   15. Zhou X
   16. Easton J
   17. Zhang J
   18. Mardis ER
   19. Wilson RK
   20. Downing JR
   21. Dyer MA

   (2017) The dynamic epigenetic landscape of the retina during development, reprogramming, and tumorigenesis

   Neuron 94:550–568.

   https://doi.org/10.1016/j.neuron.2017.04.022

   • Google Scholar
2. 1. Anders S
   2. Pyl PT
   3. Huber W

   (2015) HTSeq--a Python framework to work with high-throughput sequencing data

   Bioinformatics 31:166–169.

   https://doi.org/10.1093/bioinformatics/btu638

   • PubMed
   • Google Scholar
3. 1. Andrey G
   2. Schöpflin R
   3. Jerković I
   4. Heinrich V
   5. Ibrahim DM
   6. Paliou C
   7. Hochradel M
   8. Timmermann B
   9. Haas S
   10. Vingron M
   11. Mundlos S

   (2017) Characterization of hundreds of regulatory landscapes in developing limbs reveals two regimes of chromatin folding

   Genome Research 27:223–233.

   https://doi.org/10.1101/gr.213066.116

   • Google Scholar
4. 1. Barone LM
   2. Owen TA
   3. Tassinari MS
   4. Bortell R
   5. Stein GS
   6. Lian JB

   (1991) Developmental expression and hormonal regulation of the rat matrix gla protein (MGP) gene in chondrogenesis and osteogenesis

   Journal of Cellular Biochemistry 46:351–365.

   https://doi.org/10.1002/jcb.240460410

   • PubMed
   • Google Scholar
5. 1. Bergers G
   2. Song S

   (2005) The role of pericytes in blood-vessel formation and maintenance

   Neuro-Oncology 7:452–464.

   https://doi.org/10.1215/S1152851705000232

   • PubMed
   • Google Scholar
6. 1. Bhattacherjee V
   2. Mukhopadhyay P
   3. Singh S
   4. Roberts EA
   5. Hackmiller RC
   6. Greene RM
   7. Pisano MM

   (2004) Laser capture microdissection of fluorescently labeled embryonic cranial neural crest cells

   Genesis 39:58–64.

   https://doi.org/10.1002/gene.20026

   • PubMed
   • Google Scholar
7. 1. Blitz E
   2. Viukov S
   3. Sharir A
   4. Shwartz Y
   5. Galloway JL
   6. Pryce BA
   7. Johnson RL
   8. Tabin CJ
   9. Schweitzer R
   10. Zelzer E

   (2009) Bone ridge patterning during musculoskeletal assembly is mediated through SCX regulation of Bmp4 at the Tendon-Skeleton junction

   Developmental Cell 17:861–873.

   https://doi.org/10.1016/j.devcel.2009.10.010

   • Google Scholar
8. 1. Blitz E
   2. Sharir A
   3. Akiyama H
   4. Zelzer E

   (2013) Tendon-bone attachment unit is formed modularly by a distinct pool of scx- and Sox9-positive progenitors

   Development 140:2680–2690.

   https://doi.org/10.1242/dev.093906

   • PubMed
   • Google Scholar
9. 1. Buenrostro JD
   2. Wu B
   3. Chang HY
   4. Greenleaf WJ

   (2015) ATAC-seq: a method for assaying chromatin accessibility Genome-Wide

   Current Protocols in Molecular Biology 109:21–29.

   https://doi.org/10.1002/0471142727.mb2129s109

   • PubMed
   • Google Scholar
10. 1. Butler A
    2. Hoffman P
    3. Smibert P
    4. Papalexi E
    5. Satija R

    (2018) Integrating single-cell transcriptomic data across different conditions, technologies, and species

    Nature Biotechnology 36:411–420.

    https://doi.org/10.1038/nbt.4096

    • PubMed
    • Google Scholar
11. 1. Casper J
    2. Zweig AS
    3. Villarreal C
    4. Tyner C
    5. Speir ML
    6. Rosenbloom KR
    7. Raney BJ
    8. Lee CM
    9. Lee BT
    10. Karolchik D
    11. Hinrichs AS
    12. Haeussler M
    13. Guruvadoo L
    14. Navarro Gonzalez J
    15. Gibson D
    16. Fiddes IT
    17. Eisenhart C
    18. Diekhans M
    19. Clawson H
    20. Barber GP
    21. Armstrong J
    22. Haussler D
    23. Kuhn RM
    24. Kent WJ

    (2018) The UCSC genome browser database: 2018 update

    Nucleic Acids Research 46:D762–D769.

    https://doi.org/10.1093/nar/gkx1020

    • Google Scholar
12. 1. Chawengsaksophak K

    (2019) Cdx2 animal models reveal developmental origins of cancers

    Genes 10:928.

    https://doi.org/10.3390/genes10110928

    • Google Scholar
13. 1. Chiplunkar AR
    2. Curtis BC
    3. Eades GL
    4. Kane MS
    5. Fox SJ
    6. Haar JL
    7. Lloyd JA

    (2013) The Krüppel-like factor 2 and Krüppel-like factor 4 genes interact to maintain endothelial integrity in mouse embryonic vasculogenesis

    BMC Developmental Biology 13:40.

    https://doi.org/10.1186/1471-213X-13-40

    • PubMed
    • Google Scholar
14. 1. Dang DT
    2. Pevsner J
    3. Yang VW

    (2000) The biology of the mammalian Krüppel-like family of transcription factors

    The International Journal of Biochemistry & Cell Biology 32:1103–1121.

    https://doi.org/10.1016/S1357-2725(00)00059-5

    • PubMed
    • Google Scholar
15. 1. Deimling SJ
    2. Lau K
    3. Hui CC
    4. Hopyan S

    (2018) Genetic interaction between Gli3 and Ezh2 during limb pattern formation

    Mechanisms of Development 151:30–36.

    https://doi.org/10.1016/j.mod.2018.05.002

    • PubMed
    • Google Scholar
16. 1. Dekker RJ
    2. van Thienen JV
    3. Rohlena J
    4. de Jager SC
    5. Elderkamp YW
    6. Seppen J
    7. de Vries CJ
    8. Biessen EA
    9. van Berkel TJ
    10. Pannekoek H
    11. Horrevoets AJ

    (2005)  Endothelial KLF2 links local arterial shear stress levels to the expression of vascular tone-regulating genes

    The American Journal of Pathology 167:609–618.

    https://doi.org/10.1016/S0002-9440(10)63002-7

    • Google Scholar
17. 1. Dyment NA
    2. Breidenbach AP
    3. Schwartz AG
    4. Russell RP
    5. Aschbacher-Smith L
    6. Liu H
    7. Hagiwara Y
    8. Jiang R
    9. Thomopoulos S
    10. Butler DL
    11. Rowe DW

    (2015) Gdf5 progenitors give rise to fibrocartilage cells that mineralize via hedgehog signaling to form the zonal enthesis

    Developmental Biology 405:96–107.

    https://doi.org/10.1016/j.ydbio.2015.06.020

    • PubMed
    • Google Scholar
18. 1. Eden E
    2. Navon R
    3. Steinfeld I
    4. Lipson D
    5. Yakhini Z

    (2009) GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists

    BMC Bioinformatics 10:48.

    https://doi.org/10.1186/1471-2105-10-48

    • PubMed
    • Google Scholar
19. 1. ENCODE Project Consortium

    (2012) An integrated encyclopedia of DNA elements in the human genome

    Nature 489:57–74.

    https://doi.org/10.1038/nature11247

    • PubMed
    • Google Scholar
20. 1. Eyal S
    2. Kult S
    3. Rubin S
    4. Krief S
    5. Felsenthal N
    6. Pineault KM
    7. Leshkowitz D
    8. Salame TM
    9. Addadi Y
    10. Wellik DM
    11. Zelzer E

    (2019) Bone morphology is regulated modularly by global and regional genetic programs

    Development 146:dev167882.

    https://doi.org/10.1242/dev.167882

    • PubMed
    • Google Scholar
21. 1. Felsenthal N
    2. Rubin S
    3. Stern T
    4. Krief S
    5. Pal D
    6. Pryce BA
    7. Schweitzer R
    8. Zelzer E

    (2018) Development of migrating tendon-bone attachments involves replacement of progenitor populations

    Development 145:dev165381.

    https://doi.org/10.1242/dev.165381

    • PubMed
    • Google Scholar
22. 1. Galatz LM
    2. Ball CM
    3. Teefey SA
    4. Middleton WD
    5. Yamaguchi K

    (2004) The outcome and repair integrity of completely arthroscopically repaired large and massive rotator cuff tears

    The Journal of Bone & Joint Surgery 86:219–224.

    https://doi.org/10.2106/00004623-200402000-00002

    • PubMed
    • Google Scholar
23. 1. Genin GM
    2. Kent A
    3. Birman V
    4. Wopenka B
    5. Pasteris JD
    6. Marquez PJ
    7. Thomopoulos S

    (2009) Functional grading of mineral and collagen in the attachment of tendon to bone

    Biophysical Journal 97:976–985.

    https://doi.org/10.1016/j.bpj.2009.05.043

    • PubMed
    • Google Scholar
24. 1. Guo M
    2. Liu Z
    3. Willen J
    4. Shaw CP
    5. Richard D
    6. Jagoda E
    7. Doxey AC
    8. Hirschhorn J
    9. Capellini TD

    (2017) Epigenetic profiling of growth plate chondrocytes sheds insight into regulatory genetic variation influencing height

    eLife 6:e29329.

    https://doi.org/10.7554/eLife.29329

    • PubMed
    • Google Scholar
25. 1. Harryman DT
    2. Mack LA
    3. Wang KY
    4. Jackins SE
    5. Richardson ML
    6. Matsen FA

    (1991) Repairs of the rotator cuff. Correlation of functional results with integrity of the cuff

    The Journal of Bone & Joint Surgery 73:982–989.

    https://doi.org/10.2106/00004623-199173070-00004

    • Google Scholar
26. 1. Heinz S
    2. Benner C
    3. Spann N
    4. Bertolino E
    5. Lin YC
    6. Laslo P
    7. Cheng JX
    8. Murre C
    9. Singh H
    10. Glass CK

    (2010) Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities

    Molecular Cell 38:576–589.

    https://doi.org/10.1016/j.molcel.2010.05.004

    • PubMed
    • Google Scholar
27. 1. Hu G
    2. Schones DE
    3. Cui K
    4. Ybarra R
    5. Northrup D
    6. Tang Q
    7. Gattinoni L
    8. Restifo NP
    9. Huang S
    10. Zhao K

    (2011) Regulation of nucleosome landscape and transcription factor targeting at tissue-specific enhancers by BRG1

    Genome Research 21:1650–1658.

    https://doi.org/10.1101/gr.121145.111

    • Google Scholar
28. 1. Itzkovitz S
    2. Lyubimova A
    3. Blat IC
    4. Maynard M
    5. van Es J
    6. Lees J
    7. Jacks T
    8. Clevers H
    9. van Oudenaarden A

    (2012) Single-molecule transcript counting of stem-cell markers in the mouse intestine

    Nature Cell Biology 14:106–114.

    https://doi.org/10.1038/ncb2384

    • Google Scholar
29. 1. Jaitin DA
    2. Kenigsberg E
    3. Keren-Shaul H
    4. Elefant N
    5. Paul F
    6. Zaretsky I
    7. Mildner A
    8. Cohen N
    9. Jung S
    10. Tanay A
    11. Amit I

    (2014) Massively parallel single-cell RNA-seq for marker-free decomposition of tissues into cell types

    Science 343:776–779.

    https://doi.org/10.1126/science.1247651

    • PubMed
    • Google Scholar
30. 1. Jiang J
    2. Chan YS
    3. Loh YH
    4. Cai J
    5. Tong GQ
    6. Lim CA
    7. Robson P
    8. Zhong S
    9. Ng HH

    (2008) A core klf circuitry regulates self-renewal of embryonic stem cells

    Nature Cell Biology 10:353–360.

    https://doi.org/10.1038/ncb1698

    • PubMed
    • Google Scholar
31. 1. Johnson MB
    2. Wang PP
    3. Atabay KD
    4. Murphy EA
    5. Doan RN
    6. Hecht JL
    7. Walsh CA

    (2015) Single-cell analysis reveals transcriptional heterogeneity of neural progenitors in human cortex

    Nature Neuroscience 18:637–646.

    https://doi.org/10.1038/nn.3980

    • PubMed
    • Google Scholar
32. 1. Johnson RL
    2. Tabin CJ

    (1997) Molecular models for vertebrate limb development

    Cell 90:979–990.

    https://doi.org/10.1016/S0092-8674(00)80364-5

    • PubMed
    • Google Scholar
33. 1. Katz JP
    2. Perreault N
    3. Goldstein BG
    4. Lee CS
    5. Labosky PA
    6. Yang VW
    7. Kaestner KH

    (2002)

    The zinc-finger transcription factor Klf4 is required for terminal differentiation of goblet cells in the Colon

    Development 129:2619–2628.

    • PubMed
    • Google Scholar
34. 1. Keren-Shaul H
    2. Kenigsberg E
    3. Jaitin DA
    4. David E
    5. Paul F
    6. Tanay A
    7. Amit I

    (2019) MARS-seq2.0: an experimental and analytical pipeline for indexed sorting combined with single-cell RNA sequencing

    Nature Protocols 14:1841–1862.

    https://doi.org/10.1038/s41596-019-0164-4

    • PubMed
    • Google Scholar
35. 1. Kuo CT
    2. Veselits ML
    3. Leiden JM

    (1997) LKLF: a transcriptional regulator of single-positive T cell quiescence and survival

    Science 277:1986–1990.

    https://doi.org/10.1126/science.277.5334.1986

    • PubMed
    • Google Scholar
36. 1. Kvon EZ
    2. Zhu Y
    3. Kelman G
    4. Novak CS
    5. Plajzer-Frick I
    6. Kato M
    7. Garvin TH
    8. Pham Q
    9. Harrington AN
    10. Hunter RD
    11. Godoy J
    12. Meky EM
    13. Akiyama JA
    14. Afzal V
    15. Tran S
    16. Escande F
    17. Gilbert-Dussardier B
    18. Jean-Marçais N
    19. Hudaiberdiev S
    20. Ovcharenko I
    21. Dobbs MB
    22. Gurnett CA
    23. Manouvrier-Hanu S
    24. Petit F
    25. Visel A
    26. Dickel DE
    27. Pennacchio LA

    (2020) Comprehensive in Vivo Interrogation Reveals Phenotypic Impact of Human Enhancer Variants

    Cell 180:1262–1271.

    https://doi.org/10.1016/j.cell.2020.02.031

    • PubMed
    • Google Scholar
37. 1. Langdon WB

    (2015) Performance of genetic programming optimised Bowtie2 on genome comparison and analytic testing (GCAT) benchmarks

    BioData Mining 8:1.

    https://doi.org/10.1186/s13040-014-0034-0

    • PubMed
    • Google Scholar
38. 1. Laugesen A
    2. Helin K

    (2014) Chromatin repressive complexes in stem cells, development, and Cancer

    Cell Stem Cell 14:735–751.

    https://doi.org/10.1016/j.stem.2014.05.006

    • PubMed
    • Google Scholar
39. 1. Lee JS
    2. Yu Q
    3. Shin JT
    4. Sebzda E
    5. Bertozzi C
    6. Chen M
    7. Mericko P
    8. Stadtfeld M
    9. Zhou D
    10. Cheng L
    11. Graf T
    12. MacRae CA
    13. Lepore JJ
    14. Lo CW
    15. Kahn ML

    (2006) Klf2 is an essential regulator of vascular hemodynamic forces in vivo

    Developmental Cell 11:845–857.

    https://doi.org/10.1016/j.devcel.2006.09.006

    • PubMed
    • Google Scholar
40. 1. Lehman C
    2. Cuomo F
    3. Kummer FJ
    4. Zuckerman JD

    (1995)

    The incidence of full thickness rotator cuff tears in a large cadaveric population

    Bulletin (Hospital for Joint Diseases 54:30–31.

    • PubMed
    • Google Scholar
41. 1. Li Y
    2. Lacerda DA
    3. Warman ML
    4. Beier DR
    5. Yoshioka H
    6. Ninomiya Y
    7. Oxford JT
    8. Morris NP
    9. Andrikopoulos K
    10. Ramirez F

    (1995) A fibrillar collagen gene, Col11a1, is essential for skeletal morphogenesis

    Cell 80:423–430.

    https://doi.org/10.1016/0092-8674(95)90492-1

    • PubMed
    • Google Scholar
42. 1. Liu S
    2. Li J
    3. Tan DT
    4. Beuerman RW

    (2007) The eyelid margin: a transitional zone for 2 epithelial phenotypes

    Archives of Ophthalmology 125:523–532.

    https://doi.org/10.1001/archopht.125.4.523

    • PubMed
    • Google Scholar
43. 1. Liu Y
    2. Birman V
    3. Chen C
    4. Thomopoulos S
    5. Genin GM

    (2011) Mechanisms of bimaterial attachment at the interface of tendon to bone

    Journal of Engineering Materials and Technology 133:e011006.

    https://doi.org/10.1115/1.4002641

    • Google Scholar
44. 1. Liu YX
    2. Thomopoulos S
    3. Birman V
    4. Li J-S
    5. Genin GM

    (2012a) Bi-material attachment through a compliant interfacial system at the tendon-to-bone insertion site

    Mechanics of Materials 44:83–92.

    https://doi.org/10.1016/j.mechmat.2011.08.005

    • Google Scholar
45. 1. Liu CF
    2. Aschbacher-Smith L
    3. Barthelery NJ
    4. Dyment N
    5. Butler D
    6. Wylie C

    (2012b) Spatial and temporal expression of molecular markers and cell signals during normal development of the mouse patellar tendon

    Tissue Engineering Part A 18:598–608.

    https://doi.org/10.1089/ten.tea.2011.0338

    • PubMed
    • Google Scholar
46. 1. Liu CF
    2. Breidenbach A
    3. Aschbacher-Smith L
    4. Butler D
    5. Wylie C

    (2013) A role for hedgehog signaling in the differentiation of the insertion site of the patellar tendon in the mouse

    PLOS ONE 8:e65411.

    https://doi.org/10.1371/journal.pone.0065411

    • PubMed
    • Google Scholar
47. 1. Liu ES
    2. Martins JS
    3. Zhang W
    4. Demay MB

    (2018) Molecular analysis of enthesopathy in a mouse model of hypophosphatemic rickets

    Development 145:dev163519.

    https://doi.org/10.1242/dev.163519

    • PubMed
    • Google Scholar
48. 1. Logan M
    2. Martin JF
    3. Nagy A
    4. Lobe C
    5. Olson EN
    6. Tabin CJ

    (2002) Expression of cre recombinase in the developing mouse limb bud driven by a prxl enhancer

    Genesis 33:77–80.

    https://doi.org/10.1002/gene.10092

    • PubMed
    • Google Scholar
49. 1. Love MI
    2. Huber W
    3. Anders S

    (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2

    Genome Biology 15:550.

    https://doi.org/10.1186/s13059-014-0550-8

    • PubMed
    • Google Scholar
50. 1. Lu HH
    2. Thomopoulos S

    (2013) Functional attachment of soft tissues to bone: development, healing, and tissue engineering

    Annual Review of Biomedical Engineering 15:201–226.

    https://doi.org/10.1146/annurev-bioeng-071910-124656

    • PubMed
    • Google Scholar
51. 1. Lun ATL
    2. Riesenfeld S
    3. Andrews T
    4. Dao TP
    5. Gomes T
    6. Marioni JC
    7. participants in the 1st Human Cell Atlas Jamboree

    (2019) EmptyDrops: distinguishing cells from empty droplets in droplet-based single-cell RNA sequencing data

    Genome Biology 20:63.

    https://doi.org/10.1186/s13059-019-1662-y

    • PubMed
    • Google Scholar
52. 1. Lyubimova A
    2. Itzkovitz S
    3. Junker JP
    4. Fan ZP
    5. Wu X
    6. van Oudenaarden A

    (2013) Single-molecule mRNA detection and counting in mammalian tissue

    Nature Protocols 8:1743–1758.

    https://doi.org/10.1038/nprot.2013.109

    • PubMed
    • Google Scholar
53. 1. Madisen L
    2. Zwingman TA
    3. Sunkin SM
    4. Oh SW
    5. Zariwala HA
    6. Gu H
    7. Ng LL
    8. Palmiter RD
    9. Hawrylycz MJ
    10. Jones AR
    11. Lein ES
    12. Zeng H

    (2010) A robust and high-throughput cre reporting and characterization system for the whole mouse brain

    Nature Neuroscience 13:133–140.

    https://doi.org/10.1038/nn.2467

    • PubMed
    • Google Scholar
54. 1. McDonald SA
    2. Lavery D
    3. Wright NA
    4. Jansen M

    (2015) Barrett oesophagus: lessons on its origins from the lesion itself

    Nature Reviews Gastroenterology & Hepatology 12:50–60.

    https://doi.org/10.1038/nrgastro.2014.181

    • PubMed
    • Google Scholar
55. 1. McLean CY
    2. Bristor D
    3. Hiller M
    4. Clarke SL
    5. Schaar BT
    6. Lowe CB
    7. Wenger AM
    8. Bejerano G

    (2010) GREAT improves functional interpretation of cis-regulatory regions

    Nature Biotechnology 28:495–501.

    https://doi.org/10.1038/nbt.1630

    • PubMed
    • Google Scholar
56. 1. Michikami I
    2. Fukushi T
    3. Tanaka M
    4. Egusa H
    5. Maeda Y
    6. Ooshima T
    7. Wakisaka S
    8. Abe M

    (2012) Krüppel-like factor 4 regulates membranous and endochondral ossification

    Experimental Cell Research 318:311–325.

    https://doi.org/10.1016/j.yexcr.2011.12.013

    • PubMed
    • Google Scholar
57. 1. Nakamura E
    2. Nguyen MT
    3. Mackem S

    (2006) Kinetics of tamoxifen-regulated cre activity in mice using a cartilage-specific CreER(T) to assay temporal activity windows along the proximodistal limb skeleton

    Developmental Dynamics 235:2603–2612.

    https://doi.org/10.1002/dvdy.20892

    • PubMed
    • Google Scholar
58. 1. Orgeur M
    2. Martens M
    3. Leonte G
    4. Nassari S
    5. Bonnin MA
    6. Börno ST
    7. Timmermann B
    8. Hecht J
    9. Duprez D
    10. Stricker S

    (2018) Genome-wide strategies identify downstream target genes of chick connective tissue-associated transcription factors

    Development 145:dev161208.

    https://doi.org/10.1242/dev.161208

    • PubMed
    • Google Scholar
59. 1. Ovchinnikov DA
    2. Deng JM
    3. Ogunrinu G
    4. Behringer RR

    (2000) Col2a1-directed expression of cre recombinase in differentiating chondrocytes in transgenic mice

    Genesis 26:145–146.

    https://doi.org/10.1002/(SICI)1526-968X(200002)26:2<145::AID-GENE14>3.0.CO;2-C

    • PubMed
    • Google Scholar
60. 1. Pazin DE
    2. Gamer LW
    3. Cox KA
    4. Rosen V

    (2012) Molecular profiling of synovial joints: use of microarray analysis to identify factors that direct the development of the knee and elbow

    Developmental Dynamics 241:1816–1826.

    https://doi.org/10.1002/dvdy.23861

    • PubMed
    • Google Scholar
61. 1. Pryce BA
    2. Brent AE
    3. Murchison ND
    4. Tabin CJ
    5. Schweitzer R

    (2007) Generation of transgenic tendon reporters, ScxGFP and ScxAP, using regulatory elements of the scleraxis gene

    Developmental Dynamics 236:1677–1682.

    https://doi.org/10.1002/dvdy.21179

    • PubMed
    • Google Scholar
62. 1. Quinlan AR

    (2014) BEDTools: the Swiss-Army tool for genome feature analysis

    Current Protocols in Bioinformatics 47:11–34.

    https://doi.org/10.1002/0471250953.bi1112s47

    • PubMed
    • Google Scholar
63. Software

    1. R Development Core Team

    (2013) R: A language and environment for statistical computing

    R Foundation for Statistical Computing, Vienna, Austria.

    http://www.R-project.org
64. 1. Raj A
    2. van den Bogaard P
    3. Rifkin SA
    4. van Oudenaarden A
    5. Tyagi S

    (2008) Imaging individual mRNA molecules using multiple singly labeled probes

    Nature Methods 5:877–879.

    https://doi.org/10.1038/nmeth.1253

    • PubMed
    • Google Scholar
65. 1. Riddle RD
    2. Johnson RL
    3. Laufer E
    4. Tabin C

    (1993) Sonic hedgehog mediates the polarizing activity of the ZPA

    Cell 75:1401–1416.

    https://doi.org/10.1016/0092-8674(93)90626-2

    • PubMed
    • Google Scholar
66. 1. Roberts RR
    2. Bobzin L
    3. Teng CS
    4. Pal D
    5. Tuzon CT
    6. Schweitzer R
    7. Merrill AE

    (2019) FGF signaling patterns cell fate at the interface between tendon and bone

    Development 146:dev170241.

    https://doi.org/10.1242/dev.170241

    • PubMed
    • Google Scholar
67. 1. Romih R
    2. Korosec P
    3. de Mello W
    4. Jezernik K

    (2005) Differentiation of epithelial cells in the urinary tract

    Cell and Tissue Research 320:259–268.

    https://doi.org/10.1007/s00441-004-1005-4

    • PubMed
    • Google Scholar
68. 1. San Roman AK
    2. Shivdasani RA

    (2011) Boundaries, junctions and transitions in the gastrointestinal tract

    Experimental Cell Research 317:2711–2718.

    https://doi.org/10.1016/j.yexcr.2011.07.011

    • PubMed
    • Google Scholar
69. 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
70. 1. Schwartz AG
    2. Long F
    3. Thomopoulos S

    (2015) Enthesis fibrocartilage cells originate from a population of Hedgehog-responsive cells modulated by the loading environment

    Development 142:196–206.

    https://doi.org/10.1242/dev.112714

    • PubMed
    • Google Scholar
71. 1. Schwartz AG
    2. Galatz LM
    3. Thomopoulos S

    (2017) Enthesis regeneration: a role for Gli1+ progenitor cells

    Development 144:1159–1164.

    https://doi.org/10.1242/dev.139303

    • PubMed
    • Google Scholar
72. 1. Segre JA
    2. Bauer C
    3. Fuchs E

    (1999) Klf4 is a transcription factor required for establishing the barrier function of the skin

    Nature Genetics 22:356–360.

    https://doi.org/10.1038/11926

    • PubMed
    • Google Scholar
73. 1. Shields JM
    2. Yang VW

    (1997) Two potent nuclear localization signals in the Gut-enriched Krüppel-like factor define a subfamily of closely related krüppel proteins

    Journal of Biological Chemistry 272:18504–18507.

    https://doi.org/10.1074/jbc.272.29.18504

    • Google Scholar
74. 1. Shwartz Y
    2. Zelzer E

    (2014) Nonradioactive in situ hybridization on skeletal tissue sections

    Methods in Molecular Biology 1130:203–215.

    https://doi.org/10.1007/978-1-62703-989-5_15

    • PubMed
    • Google Scholar
75. 1. Soeda T
    2. Deng JM
    3. de Crombrugghe B
    4. Behringer RR
    5. Nakamura T
    6. Akiyama H

    (2010) Sox9-expressing precursors are the cellular origin of the cruciate ligament of the knee joint and the limb tendons

    Genesis 48:635–644.

    https://doi.org/10.1002/dvg.20667

    • PubMed
    • Google Scholar
76. 1. Soldatov R
    2. Kaucka M
    3. Kastriti ME
    4. Petersen J
    5. Chontorotzea T
    6. Englmaier L
    7. Akkuratova N
    8. Yang Y
    9. Häring M
    10. Dyachuk V
    11. Bock C
    12. Farlik M
    13. Piacentino ML
    14. Boismoreau F
    15. Hilscher MM
    16. Yokota C
    17. Qian X
    18. Nilsson M
    19. Bronner ME
    20. Croci L
    21. Hsiao WY
    22. Guertin DA
    23. Brunet JF
    24. Consalez GG
    25. Ernfors P
    26. Fried K
    27. Kharchenko PV
    28. Adameyko I

    (2019) Spatiotemporal structure of cell fate decisions in murine neural crest

    Science 364:eaas9536.

    https://doi.org/10.1126/science.aas9536

    • PubMed
    • Google Scholar
77. 1. Subramanian A
    2. Tamayo P
    3. Mootha VK
    4. Mukherjee S
    5. Ebert BL
    6. Gillette MA
    7. Paulovich A
    8. Pomeroy SL
    9. Golub TR
    10. Lander ES
    11. Mesirov JP

    (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles

    PNAS 102:15545–15550.

    https://doi.org/10.1073/pnas.0506580102

    • PubMed
    • Google Scholar
78. 1. Sugimoto Y
    2. Takimoto A
    3. Akiyama H
    4. Kist R
    5. Scherer G
    6. Nakamura T
    7. Hiraki Y
    8. Shukunami C

    (2013) Scx+/Sox9+ progenitors contribute to the establishment of the junction between cartilage and tendon/ligament

    Development 140:2280–2288.

    https://doi.org/10.1242/dev.096354

    • Google Scholar
79. 1. Takahashi K
    2. Yamanaka S

    (2006) Induction of Pluripotent Stem Cells from Mouse Embryonic and Adult Fibroblast Cultures by Defined Factors

    Cell 126:663–676.

    https://doi.org/10.1016/j.cell.2006.07.024

    • Google Scholar
80. 1. Tasic B
    2. Hippenmeyer S
    3. Wang C
    4. Gamboa M
    5. Zong H
    6. Chen-Tsai Y
    7. Luo L

    (2011) Site-specific integrase-mediated transgenesis in mice via pronuclear injection

    PNAS 108:7902–7907.

    https://doi.org/10.1073/pnas.1019507108

    • PubMed
    • Google Scholar
81. 1. Thomopoulos S
    2. Williams GR
    3. Gimbel JA
    4. Favata M
    5. Soslowsky LJ

    (2003) Variation of biomechanical, structural, and compositional properties along the tendon to bone insertion site

    Journal of Orthopaedic Research 21:413–419.

    https://doi.org/10.1016/S0736-0266(03)0057-3

    • PubMed
    • Google Scholar
82. 1. Thomopoulos S
    2. Genin GM
    3. Galatz LM

    (2010)

    The development and morphogenesis of the tendon-to-bone insertion - what development can teach us about healing

    Journal of Musculoskeletal & Neuronal Interactions 10:35–45.

    • PubMed
    • Google Scholar
83. 1. Ulitsky I
    2. Maron-Katz A
    3. Shavit S
    4. Sagir D
    5. Linhart C
    6. Elkon R
    7. Tanay A
    8. Sharan R
    9. Shiloh Y
    10. Shamir R

    (2010) Expander: from expression microarrays to networks and functions

    Nature Protocols 5:303–322.

    https://doi.org/10.1038/nprot.2009.230

    • PubMed
    • Google Scholar
84. 1. Villarreal G
    2. Zhang Y
    3. Larman HB
    4. Gracia-Sancho J
    5. Koo A
    6. García-Cardeña G

    (2010) Defining the regulation of KLF4 expression and its downstream transcriptional targets in vascular endothelial cells

    Biochemical and Biophysical Research Communications 391:984–989.

    https://doi.org/10.1016/j.bbrc.2009.12.002

    • PubMed
    • Google Scholar
85. 1. Yue F
    2. Cheng Y
    3. Breschi A
    4. Vierstra J
    5. Wu W
    6. Ryba T
    7. Sandstrom R
    8. Ma Z
    9. Davis C
    10. Pope BD
    11. Shen Y
    12. Pervouchine DD
    13. Djebali S
    14. Thurman RE
    15. Kaul R
    16. Rynes E
    17. Kirilusha A
    18. Marinov GK
    19. Williams BA
    20. Trout D
    21. Amrhein H
    22. Fisher-Aylor K
    23. Antoshechkin I
    24. DeSalvo G
    25. See L-H
    26. Fastuca M
    27. Drenkow J
    28. Zaleski C
    29. Dobin A
    30. Prieto P
    31. Lagarde J
    32. Bussotti G
    33. Tanzer A
    34. Denas O
    35. Li K
    36. Bender MA
    37. Zhang M
    38. Byron R
    39. Groudine MT
    40. McCleary D
    41. Pham L
    42. Ye Z
    43. Kuan S
    44. Edsall L
    45. Wu Y-C
    46. Rasmussen MD
    47. Bansal MS
    48. Kellis M
    49. Keller CA
    50. Morrissey CS
    51. Mishra T
    52. Jain D
    53. Dogan N
    54. Harris RS
    55. Cayting P
    56. Kawli T
    57. Boyle AP
    58. Euskirchen G
    59. Kundaje A
    60. Lin S
    61. Lin Y
    62. Jansen C
    63. Malladi VS
    64. Cline MS
    65. Erickson DT
    66. Kirkup VM
    67. Learned K
    68. Sloan CA
    69. Rosenbloom KR
    70. Lacerda de Sousa B
    71. Beal K
    72. Pignatelli M
    73. Flicek P
    74. Lian J
    75. Kahveci T
    76. Lee D
    77. James Kent W
    78. Ramalho Santos M
    79. Herrero J
    80. Notredame C
    81. Johnson A
    82. Vong S
    83. Lee K
    84. Bates D
    85. Neri F
    86. Diegel M
    87. Canfield T
    88. Sabo PJ
    89. Wilken MS
    90. Reh TA
    91. Giste E
    92. Shafer A
    93. Kutyavin T
    94. Haugen E
    95. Dunn D
    96. Reynolds AP
    97. Neph S
    98. Humbert R
    99. Scott Hansen R
    100. De Bruijn M
    101. Selleri L
    102. Rudensky A
    103. Josefowicz S
    104. Samstein R
    105. Eichler EE
    106. Orkin SH
    107. Levasseur D
    108. Papayannopoulou T
    109. Chang K-H
    110. Skoultchi A
    111. Gosh S
    112. Disteche C
    113. Treuting P
    114. Wang Y
    115. Weiss MJ
    116. Blobel GA
    117. Cao X
    118. Zhong S
    119. Wang T
    120. Good PJ
    121. Lowdon RF
    122. Adams LB
    123. Zhou X-Q
    124. Pazin MJ
    125. Feingold EA
    126. Wold B
    127. Taylor J
    128. Mortazavi A
    129. Weissman SM
    130. Stamatoyannopoulos JA
    131. Snyder MP
    132. Guigo R
    133. Gingeras TR
    134. Gilbert DM
    135. Hardison RC
    136. Beer MA
    137. Ren B

    (2014) A comparative encyclopedia of DNA elements in the mouse genome

    Nature 515:355–364.

    https://doi.org/10.1038/nature13992

    • Google Scholar
86. 1. Zhang Y
    2. Liu T
    3. Meyer CA
    4. Eeckhoute J
    5. Johnson DS
    6. Bernstein BE
    7. Nussbaum C
    8. Myers RM
    9. Brown M
    10. Li W
    11. Liu XS

    (2008) Model-based analysis of ChIP-Seq (MACS)

    Genome Biology 9:R137.

    https://doi.org/10.1186/gb-2008-9-9-r137

    • Google Scholar
87. 1. Zhu J
    2. Adli M
    3. Zou JY
    4. Verstappen G
    5. Coyne M
    6. Zhang X
    7. Durham T
    8. Miri M
    9. Deshpande V
    10. De Jager PL
    11. Bennett DA
    12. Houmard JA
    13. Muoio DM
    14. Onder TT
    15. Camahort R
    16. Cowan CA
    17. Meissner A
    18. Epstein CB
    19. Shoresh N
    20. Bernstein BE

    (2013) Genome-wide chromatin state transitions associated with developmental and environmental cues

    Cell 152:642–654.

    https://doi.org/10.1016/j.cell.2012.12.033

    • PubMed
    • Google Scholar

Acceptance summary:

Your study describes bi-fated cells at the murine tendon-to-bone attachment, namely with activation of a mixture of chondrocyte and tenocyte transcriptomes, under the regulation of shared regulatory elements and KLF transcription factors, notably KLF2 and KLF4. The report provides novel insights into the yet unknown molecular and cellular architecture of the tendon-to-bone attachments, and hence is seen to be both novel and medically relevant.

Decision letter after peer review:

Thank you for submitting your article "Bi-fated tendon-to-bone attachment cells are regulated by shared enhancers and KLF transcription factors" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Clifford Rosen as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Senior Editor has drafted this decision to help you prepare a revised submission.

The manuscript from Kult and colleagues focuses on the transcriptional regulation of a unique set of cells located between bone and tendon known as the attachment unit. Using RNA-seq and ATAC-seq, the authors investigate shared and unique transcriptional programs and accessible genomic regions. The ATACseq and epigenome profiling reveals transcriptional enhancers, with overlapping intergenic areas between bifated and both fates. Transgenic enhancer reporters expose common enhancers for tenocytes and chondrocytes. Klf2 and Klf4 were identified as critically required for differentiation. These findings lead the authors to propose that bi-fated attachment cells that connect tendon to bone. Previous studies showed that these cells express cartilage and tendon markers so the present study would need to clearly highlight the advances made compared to previous work. As a general note for example, the co-option of existing enhancers does not rule out the existence of de novo ones. This needs to be addressed clearly in the study.

Essential revisions:

1) The test of enhancers by transgenesis is somewhat limited in scope (only 8 tested) and yields some surprising results that can be explored further. An important request would be to make sure of the reproducibility of experiments. It is said that 5 out of 8 selected elements drove expression in the forelimb, but not how many attempts were done for the negative ones. Moreover, as reported in the transparent report form, N=1 for Col1a1 element, N=3 for Klf2 element, N=2 for Sox9 element, N=3 for Mgp element, N=4 for Col11a1. The reviewers understand that a site-directed approach is likely to be more reproducible that random insertion, it is recommended to examine at least 3 instances per element, to be certain of some surprising results. For example, although Sox9 is not expressed in hypertrophic chondrocytes, the Sox9 element drives expression in hypertrophic chondrocytes. Moreover, Sox9 and Klf2 elements drive expression in hypertrophic chondrocytes but not in AU cells. If these results are confirmed, they may cast doubt on the conclusion that co-option of enhancers is the (only) mechanism that regulates expression in AU cells.

2) The authors have not systemically looked for AU enhancers that are not shared with tenocytes or chondrocytes. Combined ATAC-seq dataset and published ChIP-seq of histone marks can potentially identify new enhancers. The authors could speculate or assess if those enhancers were acquired de novo and exclusive of AU cells.

3) Figure 1A, the authors present a PCA biplot : Can they be more specific on how the data were transformed prior the dimension reduction (FPKM, VST, Log transformed, CPM…?). How many genes were taken into account in the PCA; All or as it is more commonly done on the 500 most variant ones ?

4) Figure 1 C – GO terms found are very generic, this information does not really seem to be useful. Can the authors can be more specific on the parameters they used in their GSEA analysis : test used and p-value correction (FDR q-value suggests a Benjamin and Hochberg correction, it that right ?

5) In general, one has to give the detail on the software version (including package version) and OS type used for the bioinformatic analysis, these informations are missing from the manuscript.

6) In the sentence: "This suggests that the attachment cell transcriptome is largely shared with both chondrocytes and tenocytes (Figure 1A, PC1 52.47%)", the word largely is misleading.

7) The authors do not discuss the variation both on the first and the second PC and of the attachment samples. This is a big issue because there are only 2 samples for this category of cells in which the intra-group variability is very high. This leads to a poor statistical parameter estimation giving rise to poor statistical test outcome.

8) Legend of Figure 1: The term MARS-Seq is slightly misleading as it is usually associated with single cell RNAseq analysis. For clarity, please write instead bulk-MARS-Seq.

9) "To further support our initial observation that the transcriptome of the attachment cells is a mixture of chondrocyte and tenocyte transcriptomes, we clustered the statistically significant differentially expressed genes between all samples into 5 clusters, using CLICK". Same remark as the use of contrasts in DESeq2.

10) In the sentence : "From these two clusters, 374 genes, 320 of them tenogenic and 54 chondrogenic, were also found to be expressed by attachment cells." It is unclear what "so found to be expressed by attachment cells" mean? For instance, for the tenogenic markers, does this mean that in the attachment vs. chondrocytes comparison these genes are up-regulated in the attachment cells? In that case, how are the tenogenic markers defined, using the tenocytes vs chondrocytes comparison? Would it be possible to have a Venn diagram to help follow the process to define the different marker identifications? Has a simpler method using contrasts in DESeq2 been tested? If yes, do the results converge with the ones presented here? Are these "statistically significant differentially expressed genes between all samples" coming from a pairwise wald-test or a likelihood ratio test? the Materials and methods suggest that the wald-test was used. Please clarify.

11) In a previous study the authors investigated the emergence of the attachment unit (AU) with focus on bone eminence progenitors (co-expressing Sox9 and SCX up to E12.5 and expressed Col2 after E12.5 according to Col2a1CreERT2 lineage tracing). Here, they focus on the transcriptome of E14.5 attachment cells from the deltoid tuberosity, however these appear to be different from tuberosity progenitors (adjacent chondrocytes) as described in Figure 1—figure supplement 1. A better definition of what is defined as attachment unit in this paper vs previous papers and/or the AU subcompartments would help clarify the populations that are being examined.

12) Moreover, as the constitutive Col2a1Cre did not label the AU, but in the previous study did label the AU/bone eminence progenitors, it is unclear what the exact definition of AU is.

13) When using the Col2a1-Cre, R26R-tdTomato and Scx-GFP, the authors mention : Unexpectedly, the two reporters failed to label the attachment cells that were located in between these two populations. This failure might be due to a missing regulatory element in one of the constructs that was used to produce each transgenic reporter. However, in Figure 1—figure supplement 1, subpopulation 5 seems to have SCX+ cells. Is this an error of labelling? What is the orientation of this section?

14) For FACS and ATACseq analysis the authors use Sox9CreERT2;tdTomato;SCX-GFP and Col2CreERT2;tdTomato;SCX-GFP. It is not clear why for FACS the Col2CREERT2 line is used while for LCM the constitutive one is used. Moreover, as previously reported, they isolate attachment cells as double positive SOX9/SCX cells. Here again, do the cells taken for analysis include those of the tuberosity itself? Col2CreERT2 with Tamox at E12.5 should be labeling the tuberosity too. It is unclear which cells from which Cre/reporter combination have been used for the ATACseq experiments of Figure 3.

15) For Figure 2, a scheme showing where exactly in the bone we are located and how it has been sectioned would be helpful. Also, it would be nice to perform single molecule FISH on top of Col2Cre:R26TOM:SCX lineage tracing to show the specificity of the colocalization in the "double reporter-negative" area. Also, including the KLF2/4 FISH at this point would help visualize distinctions between genes belonging to cluster 5 (unique to AU) vs genes referred as mixed transcriptome (Wwp2, Bgn).

16) The authors propose a role of Klf2/4 in attachment differentiation. What is the temporality in expression of Klf2/4 vs the putative downstream factors such as Gli1, Col5a1?

17) In ISH figures, some cells in the cartilage compartment also seem to coexpress tenocyte/cartilage markers. Can the authors comment on that?

18) How did the authors adapt MARS-Seq (a single cell RNA seq pipeline taking advantage of cell sorting) to a bulk analysis? More specifically, it isn't clear how laser capture technique was combined with the MARS-seq protocol.

19) The resolution on the single molecule FISH does not allow to really appreciate a large coexpression of the presented markers in the area.

20) Figure 4: Could the authors indicate more clearly the demarcation between cartilage and connective tissue where double labeling is found?

21) It would be interesting to know if the loss of the gene expression in Figure 5 results in a morphological abnormal attachment at later postnatal stages. If the authors have looked, it would be helpful to comment in the Discussion or include the data. How much of the intermediate gene expression program in the attachment site is dependent upon Klf regulation?

22) Is this transcriptional state-sharing permanent or transitional? Their work could be nicely contrasted and compared with some studies examining transcriptional heterogeneity/the co-expression of multiple cell fates as a mechanism cells used to transition from (multipotent) progenitor states to committed fates. Enthesis tissue would be an interesting and unique situation where possibly this intermediate shared transcriptional state is maintained to generate a new cell type. Possible references for transcriptional heterogeneity in progenitors include: Soldatov et al., 2019 and Johnson et al., 2015.

23) The section on the AEG/esophagus-stomach boundary should be better integrated with their own data or removed from the Discussion. It was not clearly stated how these two tissues are similar other than being border tissues. It is recommended to expand this section to include more specific examples how these regions (enthesis and esophagus) are related. Perhaps this esophageal boundary has also been shown to have a shared transcriptional/epigenetic state with neighboring tissues?

Essential revisions

1) The test of enhancers by transgenesis is somewhat limited in scope (only 8 tested) and yields some surprising results that can be explored further. An important request would be to make sure of the reproducibility of experiments. It is said that 5 out of 8 selected elements drove expression in the forelimb, but not how many attempts were done for the negative ones. Moreover, as reported in the transparent report form, N=1 for Col1a1 element, N=3 for Klf2 element, N=2 for Sox9 element, N=3 for Mgp element, N=4 for Col11a1. The reviewers understand that a site-directed approach is likely to be more reproducible that random insertion, it is recommended to examine at least 3 instances per element, to be certain of some surprising results. For example, although Sox9 is not expressed in hypertrophic chondrocytes, the Sox9 element drives expression in hypertrophic chondrocytes. Moreover, Sox9 and Klf2 elements drive expression in hypertrophic chondrocytes but not in AU cells. If these results are confirmed, they may cast doubt on the conclusion that co-option of enhancers is the (only) mechanism that regulates expression in AU cells.

We appreciate the reviewer’s comments about the results of the site-directed LacZ enhancer-reporter assay. The paper that describes the new site-directed transgenic analysis method that we used (termed “enSERT”, for “enhancer INSERTION”) has been published very recently (Kvon et al., 2020). This publication informs about the superior specificity and reproducibility of this new approach, as verified by extensive analyses in mouse embryos. This method, which uses CRISPR/Cas9 for site-directed integration (knock-in) of LacZ reporter transgenes at the H11 safe-harbor locus, allows in vivo validation of enhancer activity with higher accuracy and specificity than previous, mostly random integration-based methods. Specifically, enSERT allows to achieve “an average transgenic rate of 50% for transgenes as large as 11 kb (based on >1,200 transgenic mice resulting from the injection of >150 independent transgenic constructs), compared to a 12% transgenic rate observed in conventional random transgenesis (based on >3,000 independent transgenic constructs)” (Kvon et al., 2020).

In Kvon et al. (Figure S2), the reliable reproducibility of the method was confirmed at E11.5 in a wide range of embryonic tissues, e.g. limbs, brain and heart. Typically, site-specific multi-copy transgenic insertion events resulted in the most sensitive (and reproducible) patterns of enhancer activity (see Materials and methods subsection “Tandem integration at H11” in Kvon et al.).

In our study, we leveraged the reliable nature of enSERT to validate in vivo multiple enhancer regions at E14.5 that had been predicted by our ATAC-seq experiments. The majority of tested regions indeed revealed in vivo enhancer activities with superior reproducibility in multiple tissues. For example, while mm1988 (Klf2 element) displayed reproducible patterns in craniofacial tissues and limbs, mm1990 (Mgp element) had reproducible patterns specifically in limbs (n=3/3 for both).

Regarding the number of examined embryos, we updated and increased the number of repetitions. We examined at least 3 embryos per enhancer and, in some cases, between 6 and 11 repetitions were made. As the reviewers suggested, we have also added the transgenic reproducibility numbers (X/Y) to Figure 4. As is the standard in the field, we compare the number of embryos with sub-regionally reproducible LacZ staining (X) to the total number of transgenic embryos obtained per tested genomic element.

As shown in Author response table 1, while some of the elements, such as the Klf2 element mm1988, displayed a high and extremely convincing reproducibly rate (N=3/3), limb-specific reproducible staining for the Col1 element mm1995 was observed in only 2 out of 9 embryos. Nevertheless, because we confirmed the limb-subregional reproducibility (on sections) in two independent biological replicates and because this element was selected based on our limb-specific ATAC-seq and ENCODE datasets of histone modification marks associated with enhancers and promoters (H3K27Ac and H3K4me1 ChIP-seq), we have reasons to believe that this result is real and represents the activity of a Col1 enhancer in attachment and tendon cells.

One explanation for the reduced reproducibility of Col1 element may be the temporal in vivo changes in enhancer usage during development. Previous studies have shown that enhancers exhibited tightly temporally restricted predicted activity windows (Nord et al., Cell 2013). It is therefore possible that mm1995 element activity at E14.5 exemplifies the start or end of this element activity window. We therefore agree that in order to establish mm1995 as an enhancer, additional repetitions are required. Unfortunately, due to the COVID-19 situation, we do not know when this will be possible. Therefore, if the reviewers approve, we would like to keep this result in our manuscript. Obviously, we clearly state its limitation.

We fully agree with the reviewers that de novo enhancers may play a role in the regulation of attachment cells in addition to co-option of enhancers (i.e. sharing of elements with tendon or cartilage cells). This issue is discussed in detail in our next answer.

2) The authors have not systemically looked for AU enhancers that are not shared with tenocytes or chondrocytes. Combined ATAC-seq dataset and published ChIP-seq of histone marks can potentially identify new enhancers. The authors could speculate or assess if those enhancers were acquired de novo and exclusive of AU cells.

We thank the reviewer for this comment. It is important to emphasize that we do not exclude the possibility that attachment cells express a unique set of genes. However, our main purpose in this manuscript is to describe the dual fate of attachment cells. Yet, to accommodate the reviewer’s concern, we have included a list of genes that we identified to be expressed specifically in the AU. In addition, we also identified a group of intergenic elements that were accessible specifically in attachment cells. These elements may act as enhancers that drive gene expression in the attachment cells, thereby providing another level of specificity to the regulation of the development of this unique tissue. We have added a table listing these intergenic regions (newly added Supplementary file 1). Among the genes that are associated with AU-specific elements, we identified several transcription factors such as Klf4, Runx3 and Nfia, in addition to ECM-associated genes such as Col11a1 and Col1a1. As the reviewers suggested, we have added text to the Results and Discussion sections describing these findings.

3) Figure 1A, the authors present a PCA biplot : Can they be more specific on how the data were transformed prior the dimension reduction (FPKM, VST, Log transformed, CPM…?). How many genes were taken into account in the PCA; All or as it is more commonly done on the 500 most variant ones ?

The PCA shows log transformed normalized data of the 500 most variant genes. The analysis was done using a modified plotPCA function of DESeq2 (rlogTransformation with parameter blind=TRUE).

4) Figure 1 C – GO terms found are very generic, this information does not really seem to be useful. Can the authors can be more specific on the parameters they used in their GSEA analysis : test used and p-value correction (FDR q-value suggests a Benjamin and Hochberg correction, it that right ?

As the reviewer suggested, we excluded the GO analysis from the revised version of the paper. We used GOrilla to find GO annotations by loading the list of 374 genes on the background of all genes and received the GO processes that were shown in Figure 1C (please find the report with all results attached as GOPROCESS_374.txt).

5) In general, one has to give the detail on the software version (including package version) and OS type used for the bioinformatic analysis, these informations are missing from the manuscript.

As requested by the reviewer, the following information has been added to the Materials and methods section:

“The data were analyzed using the Pipeline Pilot-designed pipeline for transSeq (by INCPM, https://incpmpm.atlassian.net/wiki/spaces/PUB/pages/36405284/tranSeq+on+Pipeline-Pilot). Briefly, the analysis included adapter trimming, mapping to the mm9 genome, collapsing of reads with the same unique molecular identifiers (UMI) of 4 bases (R2) and counting of the number of reads per gene with HTseq-count (Anders et al., 2015), using the most 3’ 1000 bp of each RefSeq’s transcript. DESeq2 (version 1.4.5, Love, Huber and Anders, 2014) was used for normalization and differential expression analysis with betaPrior set to true, cooksCutoff=FALSE, independentFiltering=FALSE. Benjamini-Hochberg method was used to adjust the raw p-values for multiple testing. Genes with adjusted p-value ≤ 0.05 and fold change ≥ 2 between every two conditions were considered as differential. PCA analysis was done using log transformed normalized data (DESeq2 function rlogTransformation with parameter blind=TRUE) was done with a modified plotPCA function of DESeq2.”

6) In the sentence: "This suggests that the attachment cell transcriptome is largely shared with both chondrocytes and tenocytes (Figure 1A, PC1 52.47%)", the word largely is misleading.

The word “largely” was removed from the PCA description.

7) The authors do not discuss the variation both on the first and the second PC and of the attachment samples. This is a big issue because there are only 2 samples for this category of cells in which the intra-group variability is very high. This leads to a poor statistical parameter estimation giving rise to poor statistical test outcome.

We thank the reviewer for this comment. Indeed, we fully agree that more samples would have increased our statistical power and possibly identify more AU-specific genes. However, obtaining this tissue was very challenging technically. Nevertheless, applying adjusted p-values to these samples provided us with a list of statistically significant genes that are differentially expressed in attachment cells. Importantly, the PCA results were further validated by an extensive set of experiments, such as in situ hybridization and scRNA-Seq.

8) Legend of Figure 1: The term MARS-Seq is slightly misleading as it is usually associated with single cell RNAseq analysis. For clarity, please write instead bulk-MARS-Seq.

We have changed the term MARS-Seq to bulk-MARS-Seq in the manuscript.

9) "To further support our initial observation that the transcriptome of the attachment cells is a mixture of chondrocyte and tenocyte transcriptomes, we clustered the statistically significant differentially expressed genes between all samples into 5 clusters, using CLICK". Same remark as the use of contrasts in DESeq2.

As requested by the reviewer, this text have been added to the Materials and methods section: “Clustering of the log normalized read count of differentially expressed genes was done using click algorithm (Expander package version 7.1, Ulitsky et al., 2010), followed by visualization by R (R Core Team, 2013).”

10) In the sentence : "From these two clusters, 374 genes, 320 of them tenogenic and 54 chondrogenic, were also found to be expressed by attachment cells." It is unclear what "so found to be expressed by attachment cells" mean? For instance, for the tenogenic markers, does this mean that in the attachment vs. chondrocytes comparison these genes are up-regulated in the attachment cells? In that case, how are the tenogenic markers defined, using the tenocytes vs chondrocytes comparison? Would it be possible to have a Venn diagram to help follow the process to define the different marker identifications? Has a simpler method using contrasts in DESeq2 been tested? If yes, do the results converge with the ones presented here? Are these "statistically significant differentially expressed genes between all samples" coming from a pairwise wald-test or a likelihood ratio test? the Materials and methods suggest that the wald-test was used. Please clarify.

Genes were selected as follows: First, DESeq2 yielded 865 differential genes (i.e., between all samples, Figure 1—figure supplement 2), based on criteria that are described in Materials and methods (“Genes with adjusted p-value ≤ 0.05 and fold change ≥ 2 between every two conditions were considered as differential”). Second, to demonstrate how attachment cells function as an intermediate tissue, from the 865 differential genes we selected the genes in cluster 1 (up-regulated in tenocytes vs. chondrocytes samples, hence “tenogenic”) and cluster 2 (up-regulated in chondrocytes vs. tenocytes samples, hence “chondrogenic”). Third, we averaged the normalized number of reads in attachment samples of each gene on this list and selected genes with 30 normalized reads or more. Thus, the 374 genes that are shown in Figure 1 represent the activation of both tenogenic and chondrogenic gene expression in E14.5 attachment cells. We have added this description to the Materials and methods section.

11) In a previous study the authors investigated the emergence of the attachment unit (AU) with focus on bone eminence progenitors (co-expressing Sox9 and SCX up to E12.5 and expressed Col2 after E12.5 according to Col2a1CreERT2 lineage tracing). Here, they focus on the transcriptome of E14.5 attachment cells from the deltoid tuberosity, however these appear to be different from tuberosity progenitors (adjacent chondrocytes) as described in Figure 1—figure supplement 1. A better definition of what is defined as attachment unit in this paper vs previous papers and/or the AU subcompartments would help clarify the populations that are being examined.

Indeed, the reviewer accurately cites our previous paper and we understand how the use of AU in both papers can lead to some confusion regarding its definition. We have added to the Introduction a paragraph that explains the development of the AU, serving as the enthesis primordium, whose cells co-express Sox9 and Scx. As development proceeds, this field is divided into sub compartments, namely the cartilaginous bone eminence (Col2-positive) on one side, the tendon (Scx-positive) on the other side and, connecting between them, the attachment cells, which will eventually give rise to the enthesis. The objective of this paper is to address the understudied differentiation sequence of this third population.

12) Moreover, as the constitutive Col2a1Cre did not label the AU, but in the previous study did label the AU/bone eminence progenitors, it is unclear what the exact definition of AU is.

In the previous study, as in this work, the constitutive Col2a1Cre labeled the bone side of the AU, namely the bone eminence. In this work, the term AU refers to the tissue that forms between tendon and cartilage, as explained in the previous response.

13) When using the Col2a1-Cre, R26R-tdTomato and Scx-GFP, the authors mention : Unexpectedly, the two reporters failed to label the attachment cells that were located in between these two populations. This failure might be due to a missing regulatory element in one of the constructs that was used to produce each transgenic reporter. However, in Figure 1—figure supplement 1, subpopulation 5 seems to have SCX+ cells. Is this an error of labelling? What is the orientation of this section?

The orientation of this section is transverse. When observing a series of transverse sections from E14.5 forelimbs, we identified in each section Col2+ labeled cells, Scx-GFP+ labeled cells and, between them, cells that were unlabeled. Figure 1—figure supplement 1 is one of many collected sections in which subpopulation 5 is of unlabeled cells. Regarding GFP-labelled cells at the margins of the laser-captured area, to maximize the size of extracted tissue, we stretched the borders of the AU all the way to the flanking Scx-GFP+ cells. When the laser cuts the piece of tissue it burns the boundaries, so the RNA in cells located at the boundaries is degraded.

14) For FACS and ATACseq analysis the authors use Sox9CreERT2;tdTomato;SCX-GFP and Col2CreERT2;tdTomato;SCX-GFP. It is not clear why for FACS the Col2CREERT2 line is used while for LCM the constitutive one is used. Moreover, as previously reported, they isolate attachment cells as double positive SOX9/SCX cells. Here again, do the cells taken for analysis include those of the tuberosity itself? Col2CreERT2 with Tamox at E12.5 should be labeling the tuberosity too. It is unclear which cells from which Cre/reporter combination have been used for the ATACseq experiments of Figure 3.

When we initiated the LCM experiment, we assumed that Col2 would label the attachment cells. However, to our surprise, attachment cells were Col2-negative. Nevertheless, as the Scx and Col2 reporters clearly defined the boundaries between cell types, we could still use this line for LCM. This part is explained in detail in the Materials and methods section.

Because for the ATAC-seq experiment we needed to isolate the cells using FACS, we had to change our labeling strategy. Therefore, we used a combination of Sox9-CreER and Scx-GFP mice. This indeed worked well for tendon and attachment cells, but the number of isolated chondrocytes was insufficient. Therefore, we used the Col2-CreER^T line to label chondrocytes and isolate them. To clarify this issue, we have added text to the Results section.

As the reviewer noted, this labeling should also label the chondrocytes that compose the tuberosity. However, due to the large difference in cell numbers between the tuberosity and the rest of the cartilage that was collected, we think that this had a limited (if any) effect on our results.

15) For Figure 2, a scheme showing where exactly in the bone we are located and how it has been sectioned would be helpful. Also, it would be nice to perform single molecule FISH on top of Col2Cre:R26TOM:SCX lineage tracing to show the specificity of the colocalization in the "double reporter-negative" area. Also, including the KLF2/4 FISH at this point would help visualize distinctions between genes belonging to cluster 5 (unique to AU) vs genes referred as mixed transcriptome (Wwp2, Bgn).

We have added to Figure 1 a scheme showing how we sectioned the deltoid and great tuberosities for FISH. Using E14.5 ScxGFP-Col2Cre-tdTomato transverse sections for Bgn and Wwp2 double smFISH (conjugated to Quasar 670 and CAL Fluor Red 610, respectively) is problematic, since tdTomato (540nm-581nm) is a very strong reporter that may mask the signal of the Wwp2 probe coupled to Quazar 610 (590nm-610nm), due to a "bleed through" of the tdTomato into the Quazar610 filter.

As an alternative, to address the reviewer’s comment we performed a 10x genomics experiment on Sox9+Scx+ cells, where we could search for the transcripts of Bgn and Wwp2 in addition to expression of Klf2/Klf4 in Sox9+Scx+ cells. Figure 2A,B summarizes this analysis. In short, this experiment exemplifies that Sox9+Scx+ cells indeed express both chondrogenic and tenogenic markers, in addition to KLF’s (Figure 5B).

16) The authors propose a role of Klf2/4 in attachment differentiation. What is the temporality in expression of Klf2/4 vs the putative downstream factors such as Gli1, Col5a1?

To address the temporality in gene expression, we performed transcriptome analysis (bulk MARS-Seq) of FACS-sorted attachment cells E13.5, when it is first possible anatomically to identify the attachment site (Sox9+Scx+ cells, n=3 for each gene; see Author response table 2). Results showed that at that time point, both Klf2 and Klf4, as well as Col5a1 and Gli1, are already expressed. Since our analysis of the Klf2/4 double mutants clearly showed a reduction (Figure 5) in expression of Gli1 and Col5a1, bona fide markers of the differentiating enthesis (Felsenthal et al., 2018; Schwartz et al., 2015), we concluded that Klf2 and Klf4 are necessary for attachment cell differentiation.

17) In ISH figures, some cells in the cartilage compartment also seem to coexpress tenocyte/cartilage markers. Can the authors comment on that?

We agree with the reviewer that in some cases, at the borders the expression patterns of tendon and cartilage markers are not sharply segregated. This is mainly because during development, the borders between the different tissues that compose the attachment site are not sharp and clear, as the process of compartmentalization is still underway. Nevertheless, we believe that in general, expression of ISH markers followed the expected spatial patterns. The scRNA-seq results further support this conclusion.

18) How did the authors adapt MARS-Seq (a single cell RNA seq pipeline taking advantage of cell sorting) to a bulk analysis? More specifically, it isn't clear how laser capture technique was combined with the MARS-seq protocol.

MARS-Seq was indeed originally developed as a FACS-based method for single-cell RNA sequencing [69] (Jaitin et al., 2014; Keren-Shaul, 2019). Due to the low RNA input in scRNA-seq, the authors further developed a bulk MARS-seq protocol, which is an adaptation of the single-cell MARS-seq. The input for bulk MARS-seq is clean RNA and hence it can be used for expression profiling by purifying RNA from LCM-isolated samples, FACS-sorted cells and more.

As explained in detail in the Materials and methods subsection “Laser capture microdissection”, following LCM of E14.5 cryo-sections of the attachment site, we purified the RNA using an RNA purification kit (RNeasy FFPE Kit, Qiagen). The resulting RNA was the input for the bulk MARS-seq protocol. Briefly, RNA from each sample was barcoded during reverse transcription and pooled. Following Agencourct Ampure XP beads cleanup (Beckman Coulter), the pooled samples underwent second strand synthesis and were linearly amplified by T7 in vitro transcription. The resulting RNA was fragmented and converted into a sequencing-ready library by tagging the samples with Illumina sequences during ligation, RT, and PCR. Libraries were quantified by Qubit and TapeStation as well as by qPCR for actb gene as previously described (Jaitin et al., 2014; Keren-Shaul et al., 2019). Sequencing was done on a Hiseq 2500 SR50 cycles kit (Illumina).

19) The resolution on the single molecule FISH does not allow to really appreciate a large coexpression of the presented markers in the area.

Figure 2C, showing the smFISH of biglycan and Wwp2 in E14.5 attachment cells, was increased in resolution to easily appreciate the coexpression of the markers.

20) Figure 4: Could the authors indicate more clearly the demarcation between cartilage and connective tissue where double labeling is found?

We have added clearer demarcation between cartilage and connective tissue.

21) It would be interesting to know if the loss of the gene expression in Figure 5 results in a morphological abnormal attachment at later postnatal stages. If the authors have looked, it would be helpful to comment in the Discussion or include the data. How much of the intermediate gene expression program in the attachment site is dependent upon Klf regulation?

We agree with the reviewers. However, due to postnatal lethality, our analysis of E18.5 is the latest time point that can be studied in this mouse line.

22) Is this transcriptional state-sharing permanent or transitional? Their work could be nicely contrasted and compared with some studies examining transcriptional heterogeneity/the co-expression of multiple cell fates as a mechanism cells used to transition from (multipotent) progenitor states to committed fates. Enthesis tissue would be an interesting and unique situation where possibly this intermediate shared transcriptional state is maintained to generate a new cell type. Possible references for transcriptional heterogeneity in progenitors include: Soldatov et al., 2019 and Johnson et al., 2015.

As suggested, we have added a paragraph to the Discussion regarding this issue.

23) The section on the AEG/esophagus-stomach boundary should be better integrated with their own data or removed from the Discussion. It was not clearly stated how these two tissues are similar other than being border tissues. It is recommended to expand this section to include more specific examples how these regions (enthesis and esophagus) are related. Perhaps this esophageal boundary has also been shown to have a shared transcriptional/epigenetic state with neighboring tissues?

This part of the Discussion was edited to fit with the reviewer requirements.

Author details

1. Shiri Kult

   Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Conceptualization, Validation, Investigation, Visualization, Methodology, Writing - original draft

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-8216-2908

2. Tsviya Olender

   Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Data curation, Software, Formal analysis, Validation, Visualization, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

3. Marco Osterwalder

   1. Environmental Genomics and Systems Biology Division, Lawrence Berkeley National, Berkeley, United States
   2. Department for BioMedical Research (DBMR), University of Bern, Bern, Switzerland

   Contribution

   Validation, Investigation, Visualization, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1969-2313

4. Svetalana Markman

   Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Visualization, Writing - review and editing

   Competing interests

   No competing interests declared

5. Dena Leshkowitz

   Life Sciences Core Facilities, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Data curation, Software, Formal analysis, Validation, Visualization, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

6. Sharon Krief

   Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Validation, Investigation, Visualization, Methodology

   Competing interests

   No competing interests declared

7. Ronnie Blecher-Gonen

   Department of Immunology, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Investigation, Methodology

   Competing interests

   No competing interests declared

8. Shani Ben-Moshe

   Department of Molecular Cell Biology, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Investigation, Visualization, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

9. Lydia Farack

   Department of Molecular Cell Biology, Weizmann Institute of Science, Rehovot, Israel

   Contribution

   Software, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9597-5078

10. Hadas Keren-Shaul

    Life Sciences Core Facilities, Weizmann Institute of Science, Rehovot, Israel

    Contribution

    Investigation, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

11. Tomer-Meir Salame

    Life Sciences Core Facilities, Weizmann Institute of Science, Rehovot, Israel

    Contribution

    Data curation, Formal analysis, Validation, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

12. Terence D Capellini

    Department of Human Evolutionary Biology, Harvard University, Department of Human Evolutionary Biology, United States; Broad Institute of Harvard and MIT, Cambridge, United States

    Contribution

    Resources, Writing - review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-3842-8478

13. Shalev Itzkovitz

    Department of Molecular Cell Biology, Weizmann Institute of Science, Rehovot, Israel

    Contribution

    Resources, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-0685-2522

14. Ido Amit

    Department of Immunology, Weizmann Institute of Science, Rehovot, Israel

    Contribution

    Resources, Methodology

    Competing interests

    No competing interests declared

15. Axel Visel

    1. Environmental Genomics and Systems Biology Division, Lawrence Berkeley National, Berkeley, United States
    2. U.S. Department of Energy Joint Genome Institute, Lawrence Berkeley National Laboratory, Berkeley, United States
    3. School of Natural Sciences, University of California, Merced, Merced, United States

    Contribution

    Resources, Funding acquisition, Methodology, Writing - review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-4130-7784

16. Elazar Zelzer

    Department of Molecular Genetics, Weizmann Institute of Science, Rehovot, Israel

    Contribution

    Conceptualization, Supervision, Funding acquisition, Methodology, Writing - original draft, Project administration

    For correspondence

    eli.zelzer@weizmann.ac.il

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-1584-6602

• 2,549

  Page views

• 310

  Downloads

• 24

  Citations

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