Notch Signaling Regulates the Chondrogenic Potential of Both Articular Chondrocytes and Their Progenitors During Expansion

Articular cartilage has a limited capacity for self-repair and clinical approaches to cartilage regeneration are needed. The only such approach developed to date involves an expansion of primary autologous chondrocytes in culture, followed by their reimplantation into a cartilage defect. However, because of the formation of fibrocartilage instead of hyaline cartilage, the outcome is often not satisfactory. It happens due to the de-differentiation of chondrocytes during the expansion step. Indeed, articular chondrocytes are non-proliferative and require partial or complete dedifferentiation before actively proliferating. In recent years stem/progenitor cells in articular cartilage (artSPCs) have been described. These cells maintain their own population and renew articular cartilage in sexually mature mice. artSPCs can, theoretically, be superior to chondrocytes, for repairing damaged cartilage. Accordingly, here, we searched for conditions that allow rapid expansion of both artSPCs and chondrocytes with simultaneous preservation of their ability to form hyaline cartilage. Among the modulators of Wnt, Notch, and FGF signaling and of cell adhesion screened, only fibronectin and modulators of the Notch pathway promoted the rapid expansion of artSPCs. Surprisingly, both inhibition and activation of the pathway had this effect. However, only inhibition of Notch during expansion facilitated the chondrogenic potential of both artSPCs and primary chondrocytes, whereas activation of this pathway abrogated this potential entirely. This effect was the same for murine and human cells. Our present observations indicate that Notch signaling is the major regulator of the chondrogenic capacity of both artSPCs and chondrocytes during their expansion.


Introduction
Articular cartilage, an extensively hydrated, avascular, and aneural tissue that facilitates skeletal articulation, consists morphologically of a superficial, middle (or intermediate), and deep zone. The superficial zone contains small flatten cells expressing high levels of lubricin [encoded by proteoglycan 4 (Prg4)] and tenascin-C and low levels of aggrecan. 1,2 Currently, the only cell-based approach approved for clinical repair of focal cartilage defects, first proposed by Brittberg et al. (1994), 3 involves an expansion of autologous chondrocytes in culture followed by their re-implantation without (ACI) or with (MACI) matrix. Since this treatment results in a joint that functions properly for many years, thereby delaying the necessity for joint-replacement surgery, it is of particular value for young adults. 4 A problem often encountered with ACI/MACI is the formation of fibrocartilage rather than the more durable hyaline cartilage, 5 due to the progressive dedifferentiation of chondrocytes during their expansion in monolayer culture. [6][7][8] This dedifferentiation can be reversed to a certain extent, [8][9][10] but the outcome remains far from ideal.
On the basis of their capacity to form colonies in vitro, expand extensively and contribute to the development of various tissues in vivo, it has been proposed that progenitor cells in articular cartilage are mainly located in the superficial zone. 11 Definitive evidence for this was provided by the demonstration by genetic tracing that in Prg4-driven CreERt mice coupled with the R26R-LacZ reporter strain, Prg4-positive cells, located on the surface of the cartilage, can generate chondrocytes in articular cartilage. 12 In fact, the entire articular cartilage of adult mice is generated by the progeny of Prg4-positive cells, 13 which also contribute to the healing of focal cartilage defects. 14 It is noteworthy that these progenitor cells undergo renewal via symmetric division and express the stem cell markers CD73 (ecto-5ʹ-nucleotidase), 13 CD105 and CD34. 15 Collagen XXII is probably a constituent of the niche for these cells located at the surface of articular cartilage. 16 Accordingly, these features allow us to define the superficial chondroprogenitors as articular stem and progenitor cells (artSPCs) 17 and we hypothesize that artSPCs may provide a better cellular source for ACI/ MACI than chondrocytes.
In general, signaling pathways that play pivotal roles in development participate in the regulation of stem cells. 18,19 The pathways associated with joint formation include Wnt/b-catenin, hedgehog, FGF, BMP, and TGFb (reviewed in reference 20 ) and have been implicated in the maintenance of stem cells in various compartments of bone (reviewed in reference 17 ).
Wnt signaling initiates the formation of a joint 21 and overexpression of β-catenin soon after birth results in the thickening of the superficial zone. 22,23 FGF-2, -4, -10, -13, and -18 as well as the FGF receptor2 (FGFR2) and FGFR3 are expressed at high levels in the interzone, which later develops into the joint. 24 The outcome of Notch signaling, a highly conserved pathway involved in determining the fates of cells, including the self-renewal, proliferation, differentiation, and apoptosis of stem/progenitor cells, is dependent on the environment and past history of the cell in question. Relatively high levels of Notch1, Delta1, Jagged1, and Jagged2 are present in the superficial zone of immature bovine articular cartilage, 25 indicating that these factors are likely to help regulate the formation, maintenance, and homeostasis of artSPCs.
In addition to such morphogens, regulation of the artSPCs is likely to involve mechanical forces and the surrounding matrix. Indeed, movement during early development facilitates the appearance of the synovial cavity and the establishment of the articular joint. 26 Moreover, in direct response to sheer stress, artSPCs increase their production of lubricin through a process involving prostaglandin E2 and the transcription factor CREB. 27 Furthermore, in contrast to articular cartilage, the matrix of the superficial layer, which houses the artSPCs, contains collagen types I and XXII. In addition, artSPCs adhere faster to fibronectin 11 and propagate well on plastic coated with this protein. 28,29 As a consequence of this seminal work by Dowthwaite et al. (2004), plastic coated with fibronectin is now widely used to isolate and culture cells derived from the superficial layer of articular cartilage. 25,[30][31][32] In this current investigation, we attempted to optimize conditions for the expansion of artSPCs without compromising their chondrogenic potential. In this context we explored the potential usefulness of the Wnt, FGF, and Notch signaling pathways, utilizing fibronectin as a positive control for comparison.

artSPCs Cells Form Clones and Undergo Multi-Lineage Differentiation
Previously, we demonstrated that murine artSPCs, which generate chondrocytes in vivo, express the mesenchymal stem cell marker CD73. 13 In agreement with this earlier finding, most CD73 high cells detected here in the superficial zone of articular cartilage, with a few located at the surface of the meniscus ( Fig.  1A-1C). Thereafter, articular cartilage was isolated surgically and digested briefly with collagenase, followed by isolation of CD73 high cells by magnetic-activated cell sorting (MACS) and subsequent analysis of their purity on the basis of expression of both CD73 and Sox9, as determined by FACS ( Fig. 1D-1F). We found that 95.5% of cells isolated were CD73 high /Sox9+, 0.66% CD73 high /Sox9−, and only 3.79% have low CD73 level, including 3.41% CD73 low /Sox9+ and 0.38% CD73 low / Sox9−. Thus, the specificity of the MACS procedure and the purity of the cells obtained were confirmed.
To verify the stem cell properties of the cells obtained, colony formation from diluted solutions of cells was assayed. After 12 days of culture distinct colonies containing >60 cells each were observed (Fig. 1G). To determine whether these colonies originated from individual cells, CD73 high cells were isolated from Prg4-CreERt2:R26R-Confetti mice pulsed with tamoxifen 3-4 days earlier. Mono-clonal colonies were detected (Fig. 1H, 1I), thereby confirming the ability of individual cells to form colonies.
Finally, to explore the capacity of the superficial cells obtained to differentiate into different mesenchymal lineages, CD73 high cells were placed immediately after MACS into either adipogenic or osteogenic culture medium for 21 days, following which numerous adipocytes or osteoblasts, respectively, could be identified (Fig. 1J, 1K).
Altogether, these experiments are in good agreement with previous observations [11][12][13] and confirm that the CD73 high cells obtained are artSPCs.

The Notch Pathway is Among the Most Potent Stimulators of artSPCs Proliferation
We next explored the influence of Notch, FGF, and Wnt signaling on the expansion of artSPCs during 3 passages (as shown in Supplementary Fig. S1A), utilizing fibronectincovered plastic as a positive control. Cell growth was promoted most extensively by fibronectin (a 76-fold increase from the starting number, 54.8% greater than control cells), as well as by both an activator (Jagged-1, 121-fold increase from the starting number, 98.6% greater than control) and inhibitor (DAPT, 71-fold increase, 95.4% greater than control) of Notch signaling (Fig. 1L, 1M). FGF2 promoted cell expansion transiently (Fig. 1N); while activation of the Wnt/bcatenin pathway by LiCl had no effect and inhibition of this same pathway by XAV attenuated proliferation (Fig. 1M). Thus, fibronectin, DAPT, and Jagged1 resulted in the most extensive expansion, with both activation and inhibition of the Notch pathway exerting similar effects.
At the same time, Jagged-1 upregulated the osteoblast marker collagen I ( Supplementary Fig. S1F) and downregulated both markers of chondrocytes, collagen II and Sox9 ( Supplementary Fig. S1G, S1H). In contrast, inhibition of Notch signaling with DAPT decreased the expression of collagen I (Supplementary Fig. S1F) and elevated the levels of both collagen II and Sox9, albeit the latter not to a significant extent ( Supplementary Fig. S1G, S1H). Cells expanded on fibronectin influenced the levels of collagens I and II and Sox9 to an extent similar to DAPT (Supplementary Fig. S1F-S1H). Analysis of the expression of Notch transmembrane receptors (Notch1, Notch2, Notch3, and Notch4), as well as of the Hes1 and Hey1 target genes confirmed the presence of the necessary receptors and the expected cellular response to Jagged 1 ( Supplementary Fig. S1I-S1N).
These observations indicate that among the agents examined, DAPT and fibronectin preserve chondrogenic potential most effectively during cell expansion.

During Expansion, the Chondrogenic Potential Is Controlled by the Notch Pathway
Thereafter, we set to explore the ability of cells expanded as above (see Supplementary Fig. S1A) to form hyaline cartilage. Expansion conditions were found to exert a clear impact on the ability of the expanded artSPCs to form hyaline cartilage. Cells, expanded in the presence of DAPT or on fibronectincoated dishes formed the largest cartilage pellets (Fig. 1O, 1P) whereas expansion in the presence of Jagged1 or LiCl resulted in significantly smaller pellets (Fig. 1O, 1P).
Staining with SafraninO/Fast Green revealed that most proteoglycans were observed in the matrix deposited by artSPCs expanded in the presence of DAPT, DMSO (the vehicle for DAPT), or the Wnt-pathway inhibitor XAV ( Fig.  2A-2H). Expansion on fibronectin-coated plastic resulted in small patches that stained positive for proteoglycans in subsequent pellet culture (Fig. 2G). FGF2 resulted in the typical chondrogenic pellets with only proteoglycans deposited in the center (Fig. 2H) while those expanded in the presence of Jagged 1 or LiCl did not deposit any visible proteoglycans in their produced matrix in subsequent pellet culture (Fig. 2B, 2F).
For further characterization of the obtained pellets, we used Sox9 to mark chondrocytes and chondroprogenitors, collagen II to mark mature chondrocytes, collagen I, which marks osteoblasts as well as fibrocartilage, and Runx2, a master transcription factor for osteolineage differentiation but also marks hypertrophic chondrocytes. The highest levels of Sox9 and Collagen II, and the lowest levels of collagen I and Runx2 were observed in the case of cells expanded in the presence of DAPT (Fig. 2Dʹ-2D″″). Moreover, this same group displayed the most homogenous distribution of collagen II (Fig. 2Dʹ-D″ʹ), in contrast to several other groups, including the controls and those expanded in the presence of FGF2 and fibronectin, where this distribution was uneven. In these latter cases, surrounding cells expressed collagen I, suggesting that only a portion of the artSPCs retained their chondrogenic potential during an expansion ( Fig. 2Aʹ-2A″″, 2Gʹ-2G″ʹ, 2Hʹ-2H″ʹ). Expansion of artSPCs in the presence of either FGF2 or fibronectin gave rise to a large number of Mef2c-positive cells, indicative of chondrocyte hypertrophy (Fig. 2G″″, 2H″″). Interestingly, artSPCs expanded in the presence of Jagged1 did not differentiate into chondrocytes at all (Fig. 2Bʹ-2B″″) and the associated pellets expressed collagen I abundantly ( Fig. 2B″) but no Runx2 or Mef2c ( Fig. 2B″ʹ and B″″), suggesting that they contained fibroblast-like cells.
These findings demonstrate clearly that the conditions used for expansion exert an impact on the subsequent chondrogenic potential of artSPCs, with Notch signaling playing the predominant role in this connection.

Modulation of Notch Activity Influences Chondrocytes and artSPCs in a Similar Manner
Analysis by FACS did not reveal 2 distinct populations of cells containing different levels of CD73, but rather a continuum of cells containing low to high levels (Supplementary Fig. S2A-S2D), which theoretically may reflect a gradual loss of CD73 expression during differentiation of artSPCs toward chondrocytes. When we defined CD73 high and CD73 low populations by gating as shown in Supplementary  Fig. S2C and S2D, the morphology of these sub-populations appeared to differ, as reflected in their scattering properties ( Supplementary Fig. S2A).
As expected, artSPCs (CD73 high ) formed numerous colonies, whereas chondrocytes (CD73 low ) did not (Fig. 3A, 3B). Surprisingly, both populations of cells differentiated toward the adipogenic and osteogenic lineages to a highly similar extent ( Supplementary Fig, S3A-S3D). To directly prove that chondrocytes can differentiate into osteoprogenitors and adipocytes, chondrocytes expressing collagen type II were labeled genetically by pulsing Col2CreERT;R26-Confetti mice on postnatal day 2 with tamoxifen and subsequently isolating Confetti-positive cells from surgically dissected cartilage by FACS. Upon appropriate differentiation, both osteoprogenitors (Osx-positive) and adipocytes (perilipin1positive) cells were detected among the genetically labeled cells ( Supplementary Fig. S3E-S3P). This observation aligns well with the extensive plasticity of hypertrophic chondrocytes in vivo observed recently. 33 Expansion of either artSPCs or chondrocytes in the presence of the Notch inhibitor preserved more chondrogenic potential in comparison with the control cells or fibronectin-expanded cells (Fig. 3E-3Lʹ). In contrast, Notch activation by Jagged1 during expansion totally eliminated the chondrogenic potential of both cell types (Fig. 3I-3Jʹ). Collagen type I, a marker of fibroblasts/osteoblasts, was not detected in cultures expanded in the presence of DAPT (Fig. 3H, Hʹ), but was abundant after culturing in the presence of Jagged1 (Fig. 3J, Jʹ).
Quantification of these data ( Fig. 3M-3O), along with markers of chondrocyte hypertrophy (Mef2c) and osteogenic differentiation (Osterix, Osx) (Supplementary Fig.  S2N-S2Oʹ), confirmed that expansion in the presence of the Notch inhibitor preserved the chondrogenic potential of both artSPCs and chondrocytes most effectively, without any Fast green was used to visualize cartilage and proteoglycan content. Immunodetection of (Aʹ-Hʹ) collagen type II (Col II, red) and the SOX9 transcription factor (green), (A″-H″) Col II and collagen type I (green, Col I), (A″ʹ-H″ʹ) Col II and the Runx2 transcription factor (green), and (A″″-H″″) the Mef2c transcription factor in combination with nuclei being visualized by DAPI (blue). All images were captured using the same setting and processed with Imaris software. Representative images of 3 independent experiments are shown and cells for each experiment were collected from 5 to 6 mice. adverse effect on cell hypertrophy or trans-differentiation toward the osteo-lineage. Interestingly, no Osx was detected after expansion in the presence of Jagged1 (Supplementary Fig. S2M and S2Mʹ), which, coupled with the abundant expression of collagen type I in this case, indicated dedifferentiation of artSPCs and chondrocytes to fibroblast-like cells. The size of the pellets obtained from both artSPCs and chondrocytes was enhanced by DAPT only, although not to a significant extent (Supplementary Fig. S2G). We noted that isolation by FACS damaged artSPCs somewhat, with these cells appearing morphologically unhealthy immediately after sorting and during the first passage (not shown), which may explain their moderate rate of expansion ( Supplementary Fig.  S2E, S2F) in comparison to MACS-sorted cells (Fig. 1M).
Thus, the Notch pathway appears to be a major regulator of the chondrogenic potential of both chondrocytes and artSPCs. Inhibition of this pathway during expansion preserves the ability of both these types of cells to form hyaline cartilage, whereas activation of this pathway causes a complete loss of their chondrogenic potential.

Inhibition of Notch Signaling Preserves the Chondrogenic Potential of Human Articular Cartilage Cells
To explore the potential clinical relevance of the observations above, we investigated the effect of inhibiting Notch signaling on the rate of expansion and chondrogenic potential of human articular cartilage cells. CD73 turned out not to be a specific marker for human superficial cells (not shown). Accordingly, we have tested a panel of various surface antigens (CD73, CD29, CD44, CD90, CD105, and CD146) to identify the one, which can selectively label the superficial cells in human articular cartilage. This search revealed that CD29 is a specific marker for human superficial cells, flat in their morphology, and does not label underlying round chondrocytes (Fig. 4A). Sorting by FACS resulted in 2 distinct populations of cells, expressing CD29 at high levels (CD29 high ) and essentially lacking CD29 (CD29 low ) ( Supplementary Fig. S4A-S4D), with the former producing significantly more colonies of larger size (Fig. 4B-4E). Both populations underwent osteogenic ( Supplementary Fig. S4E-S4F) and adipogenic differentiation ( Supplementary Fig. S4G-S4H), in a manner similar to murine chondrocytes and CD73 high murine art SPCs.
Inhibition of Notch signaling during cell expansion elevated the chondrogenic potential, ie, increased the number of Sox9-positive cells in the pellets obtained from either CD29 low or CD29 high cells (Fig. 4Fʹ-4Iʹ, 4J) while the levels of collagen type II were increased only in artSPCs ( Fig. 4F″-4I″, 4K). Importantly, expansion in the presence of DAPT reduced the amount of collagen type I in both types of pellets ( Fig. 4F″ʹ-4I″ʹ, 4L), suggesting that de-differentiation during the culture period was inhibited. The scarcity of Mef2c in these pellets ( Fig. 4F″″-4I″″, 4M) indicated a lack of hypertrophic differentiation. As in the case of murine chondrocytes and artSPCs, the presence of DAPT resulted in an expansion of human chondrocytes and artSPCs that was almost 2.5-fold as great as with control cultures (Fig. 4N).
Thus, inhibition of Notch signaling during the expansion of human articular chondrocytes and especially chondroprogenitors preserves their chondrogenic ability to form cartilage-like pellets afterward and can potentially lead to a better cartilage formation upon reimplantation of expanded cells in such therapies as ACI and MACI.

Discussion
In the current investigation we show that articular stem cells (artSPCs) from both mice and humans can be successfully expanded in vitro and, in addition, identified conditions that facilitate both the expansion of these cells and simultaneously preserve their chondrogenic potential. Our findings reveal that Notch signaling is a master regulator of chondrogenic potential during the expansion of artSPCs, with activation of this pathway resulting in complete loss of their ability to make hyaline cartilage and inhibition preserving this ability. We also demonstrate that the Notch pathway regulates the chondrogenic potential of cultured primary articular chondrocytes in a similar manner.
At present, treatment with primary autologous articular chondrocytes (eg, involving MACI) is the only therapy approved by the FDA for cartilage regeneration. However, these cells expand slowly and undergo dedifferentiation during this expansion. To overcome these problems, we explored here the expansion and chondrogenic potential of artSPCs for potential alternative use in cartilage regeneration.
These cells generate the entire articular cartilage during postnatal life and express several markers of MSCs, including CD73. 13,34 Murine artSPCs, isolated on the basis of their high-level expression of CD73, form colonies and, like artSPCs isolated by adhesion to fibronectin, are multi-potent, capable of differentiating into 3 mesenchymal lineages, osteoblasts, adipocytes, and chondrocytes. 11,32,35,36 Interestingly, CD73, identified in our laboratory as a surface marker for epiphyseal stem cells 13 in both mice and rats (not shown), proved not to be a specific marker for human artSPCs. Instead, we found that human artSPCs express specifically the surface marker CD29. It is noteworthy that human CD29 high cells form colonies and differentiate into osteoblasts, adipocytes, and chondrocytes in a manner similar to murine CD73 high cells. The lack of evolutionary conservation of surface markers for other stem and progenitor cells, such as MSCs and HSCs, is well documented [37][38][39] and, therefore, not very surprising. CD29 is a β1 subunit of α5β1 integrin, a classical receptor for fibronectin, and all human superficial cells selected by high adherence to fibronectin were reported to be positive for CD29. 30 CD73 is a common name for 5ʹ-nucleotidase (gene NT5E), a common marker for MSCs. CD73 converts mononucleotides to nucleosides, preferentially adenosine monophosphate (AMP) to adenosine, a ligand for adenosine receptors. Both CD73 and adenosine receptors participate in cartilage homeostasis and OA development. 40,41 Mice and humans differ substantially concerning OA development, which is attributed to differences in posture, weight, and thickness of the articular cartilage. 42 Inflammation plays a very important role in OA 43 and mice differ substantially in their inflammatory response from humans. 44 Altogether, it results in low clinical predictability of mouse models in relation to drug development for OA therapy. 45,46 Whether the expression of different markers and disease-associated genes by artSPCs contributes to this low predictability remains to be addressed.
Nevertheless, artSPCs can be isolated from both humans and mice and may theoretically allow more efficient MACI. However, their rates of and optimal conditions for expansion, as well as their ability to retain chondrogenic potential during expansion were previously unclear and, accordingly, have been characterized here. Since expansion ex vivo is an important step in the MACI procedure, 47 we first screened for and found 4 conditions that facilitate the expansion of artSPCs, ie, both activation and inhibition of Notch signaling and expansion in the presence of FGF2 or on plastic covered with fibronectin.
However, further analysis revealed that chondrogenic potential is preserved only when these cells are expanded in the presence of the Notch inhibitor, DAPT, whereas activation of this pathway completely abrogates the ability to make hyaline cartilage. Moreover, neither FGF2 nor fibronectin exerts a significant impact on chondrogenic potential, and a certain degree of de-and/or trans-differentiation, similar to that observed in control cultures, is observed under these conditions. Partial loss of chondrogenic potential is probably the main cause of the general problem with the MACI approach, ie, frequent formation of fibrocartilage (containing collagen type I and poorly sustainable clinically) instead of hyaline cartilage. 48 This problem is thought to be caused by the partial dedifferentiation of chondrocytes during their expansion. 7,49,50 Thus, inhibition of Notch signaling during expansion may reduce the production of fibrocartilage in connection with MACI.
Importantly, we observed that inhibition of Notch signaling is beneficial not only for the expansion of artSPCs, but also for articular chondrocytes. Since articular chondrocytes are physiologically non-proliferative, they probably have to partially dedifferentiate before being able to expand on plastic 51 and blocking of Notch signaling might potentially direct this de-differentiation toward chondroprogenitors. The role of Notch signaling in the de-differentiation of artSPCs and chondrocytes is exemplified further by the observation that activation of Notch by Jagged 1 during the expansion phase results in the total loss of the ability of these cells to form cartilage.
De-differentiation of chondrocytes or, more generally, their plasticity is rather remarkable. Approximately 40%-60% of the hypertrophic chondrocytes in the growth plate transdifferentiate into osteoblasts, 52 with a recent report suggesting that this process involves de-differentiation followed by subsequent differentiation toward osteogenic, adipogenic, and other stromal lineages. 33 Indeed, the transition from hypertrophic chondrocytes to other lineages includes a cell stage characterized by the expression of numerous markers specific to skeletal stem cells. 33 Our unexpected observation that articular chondrocytes, even those obtained from elderly patients, are capable of differentiating into osteogenic and adipogenic lineages in culture is consistent with the extensive plasticity of chondrocytes and also indicates that cultured chondrocytes undergo dedifferentiation. Interestingly, such an ability of aged human deep zone articular chondrocytes to differentiate toward osteo-and adipolineages in vitro has been reported but was interpreted as potential contamination by multipotent progenitors. 25 Our current observations including those with lineage tracing suggest rather an intrinsic plasticity of articular chondrocytes. Of note, at least fetal/neonatal articular chondrocytes give rise to bone cells in vivo, 13 again exemplifying their remarkable plasticity. Thus, although the mechanism(s) underlying de-differentiation of chondrocytes in vivo are currently poorly understood, the plasticity of these cells ex vivo appears to be tightly controlled by Notch signaling, which may have important implications for cartilage regeneration.
The role of the Notch pathway in determining the fate of embryonic cells is well known. 53 For example, Notch participates actively in skeletal development by controlling progression from less toward more differentiated cells of both the osteogenic and chondrogenic lineages in vivo. [54][55][56][57] In cultures of limb bud pellets inhibition of the Notch pathway promotes chondrogenic differentiation. 58 Activation of this same pathway in chondrogenic ATDC5 cells inhibits their differentiation toward chondrocytes. 59,60 Thus, the ability of Notch signaling to modulate chondrogenic potential during the expansion of artSPCs observed here is consistent with the general Notch-dependent regulation of cell fate. At the same time, the role of Notch in development may not reflect its role in the pathophysiology of adult articular cartilage.
Members of the Notch signaling pathway are expressed differentially in the different zones of articular cartilage, with Notch1 being expressed exclusively in the superficial layer, where the artSPCs are located, and Notch2 being expressed more uniformly throughout the articular cartilage. 11,31,61 The Notch agonists Jagged1 and Delta4 have been detected in human articular cartilage, predominantly in the superficial zone. 31 Furthermore, the level of expression of both Notch receptors and ligands increases dramatically in connection with the development of osteoarthritis, 31,61,62 suggesting a role for this pathway in pathological processes involving cartilage. Any such role is likely to be negative since genetic activation of the Notch pathway in cartilage for a prolonged period causes degeneration of this tissue in a manner dependent on IL6. 63 At the same time, ablation of the key downstream component of the Notch pathway, RBP-Jk, protects the joint from surgery-induced osteoarthritis 62 but results in age-dependent osteoarthritis. 64 It appears plausible that surgical induction of osteoarthritis activates Notch signaling throughout the cartilage, 62 but in mice lacking RBP-Jk, this elevation of Notch activity is abrogated. Probably a certain low level of Notch activity is required for maintenance of the physiological homeostasis of cartilage and ablation of Notch may disrupt this homeostasis and cause the degeneration of cartilage with age. 64 In vitro observations concerning the role of the Notch pathway in the regulation of chondrocytes and chondroprogenitors are rather confusing. Although blocking Notch signaling with DAPT has been reported to impair the proliferation of chondrocytes, 61 as well as the formation of colonies by artSPCs that were isolated via adherence to fibronectin, 11 such blockage has also been reported to improve the proliferation of chondrocytes. 65 The latter observation agrees with our present findings.
One can only speculate on the potential reasons for such discrepancies, which may involve differences in the dose and duration of treatment, preparation of cells (exemplified here by the difference in the extent of expansion of artSPCs isolated by MACS or FACS upon DAPT treatment), epigenetic status of the cells obtained, and many other factors as well. However, we would like to emphasize that the experiments documented here were reproduced in two different laboratories by 3 independent researchers (L.L., A.D.K., and A.P.U). Thus, minor variations in conditions and handling can probably be excluded.
In summary, we conclude that the Notch pathway is a primary regulator of chondrogenic potential during culturing of articular chondrocytes or artSPCs. Inhibition of this pathway allows rapid expansion of these cells with simultaneous preservation of their ability to form hyaline cartilage.

Animals
All animal experiments were pre-approved by the bioethics committee of Sechenov University (Moscow, Russian Federation) and the Ethical Committee on Animal Experiments (Stockholm North Committee/Norra Djurförsöksetiska Nämd, Stockholm, Sweden). Mice were used on postnatal day (P) 3-7. For experiments not involving transgenic strains, as well as for the collection of cartilage, C57Bl6 mice were used. Col2-CreER mice were obtained from Susanne Mackem 66 ; the Rosa26-Confetti animals from Snippert et al. 67 ; and the Prg4CreER strain from Andrew Lassar. 12 Prior to tissue collection mice were euthanized by an overdose of isoflurane.

Human Samples
Following the provision of written informed consent, human articular cartilage was obtained from patients in connection with knee surgery. The surgeries were conducted due to osteoarthritis of the joint leading to total knee replacement. Cartilage from 5 female patients with one-sided (either lateral or medial) osteoarthritis was used for experiments. Patients were 46, 66, 66, 69, and 71 years of age at the time of cartilage collection. Only intact cartilage was collected for the experiments.

Cell Isolation
Cartilage from the knees of mice 5 ± 2 days old was dissected out into ice-cold medium (DMEM/F-12-HEPES containing 0.1 mg/mL gentamycin) and as much of the ligaments, tendons, and other tissue surrounding the cartilage as possible gently removed under a stereomicroscope. The cartilage was subsequently incubated with 0.15% collagenase type II for 2-5 h at 37 °C on a roller. For every experiment, cells were isolated from one litter (5-6 mice).
Human samples were washed in sterile PBS containing 0.1 mg/mL gentamycin and, thereafter, intact cartilage was dissected out, washed in dissection medium, cut into 1-2 mm fragments, and incubated in 0.1% collagenase type II overnight at 37 °C on a roller.

Magnetic-Activated Cell Sorting (MACS)
Isolated murine cells were passed through a filter (40 um) into MACS buffer (phosphate-buffered saline (PBS), 2% FBS, 1 mM EDTA) and then subjected to MACS as described by Gagliardi and colleagues (2018). In brief, the cells were incubated with magnetic beads covered with an antibody against CD73 conjugated with PE (diluted 1:10, clone TY/11.8, Biolegend) at 4 °C for 15 minutes. After subsequent washing with MACS buffer, MicroBeads (Miltenyi Biotec) were mixed with the cell suspension for 15 minutes at 4 °C. Following 3 washes, CD73 high cells were then collected on a magnetic platform.

Flow Cytometry and Fluorescence-Activated Cell Sorting (FACS)
To test their purity, CD73 high cells obtained by MACS were fixed and permeabilized in accordance with the manufacturer's instructions (BD Bioscience). Intracellular staining for Sox9 (HPA001758 Sigma) was also performed as described by BD Bioscience and the data acquired employing LSRII software (BD Bioscience).
For the performance of FACS, cells were stained with the same CD73 antibody (diluted 1:100 and conjugated with PE or APC). Five hundred thousand human cells were stained with CD29 antibodies conjugated with PE (Miltenyi Biotec, clone TS2/16, 1:100, 100 uL per aliquot). Murine cells were sorted on a FACSAria III column (BD Bioscience) and human cells on a Sony SH800 device. Flow cytometric data were analyzed with the FlowJo software (version 10, TreeStar).

Analysis of Colony Formation and Size
CD73 high cells obtained from Prg4-CreERT2 mice 12 crossed with R26R-Confetti mice 67 as described above were seeded at low density (250 cells/cm 2 ) and cultured for 12 days. Thereafter, colony number and size were evaluated by staining with H&E as well as directly by fluorescence microscopy. Mice were pulsed with tamoxifen on day 3 ± 1 and cells were collected 2-3 days later.

Determination of the Cells' Differentiation Potential
Adipogenic and osteogenic differentiation were assessed as described recently. 68 In brief, when isolated cells at first passage had reached 80% confluence, either adipogenic or osteogenic differentiation medium was added for 21 days, following which the cells were stained with Oil Red O or alizarin red, respectively.

Cell Culture
Following MACS, CD73 high cells were seeded onto plates (5000 cells/cm 2 ) and expanded in presence of either LiCl, which stimulates Wnt/b-catenin (10 mM, Sigma-Aldrich), the Wnt/b-catenin inhibitor XAV-939 (10 uM, Selleckchem), the Jagged1, which is a Notch ligand (10 ug/mL, R&D Systems), the Notch inhibitor DAPT (10 uM, Sigma-Aldrich), or FGF2 (10 ng/mL) for 3 passages. For analysis of the role of adhesion molecules, Petri-dishes were covered with fibronectin (Fi) (3 ug/cm 2 ), while the control dishes received only PBS. The medium was replaced with fresh medium supplemented with activators/inhibitors once every 3 days and in connection with each passage, 10 000 cells were collected for analysis of gene expression.
The procedures for delivery of active Jagged1 to cells in culture have been described previously. 69 In brief, 20 ug/mL Fc-specific human IgG (Sigma-Aldrich) diluted in PBS was incubated in the culture dishes at room temperature for 2 h. Thereafter, 10 ug/mL of either recombinant rat Jagged1 fused to human Fc (R&D Systems) or human Fc control protein (R&D Systems) dissolved in PBS was added to the dishes pre-coated with Fc-specific human IgG and incubated at 4 °C overnight. This medium was removed before seeding the cells.
After FACS, the cells were seeded (28 000 cells/cm 2 ) and subsequently treated with DAPT, Jagged1, and fibronectin as described above. The higher density, as compared with MASC, was used to facilitate cell survival upon the FACS isolation.

Analysis of Gene Expression
Total RNA was extracted from cells in the 3rd passage with the RNeasy Micro Kit (QIAGEN) and reverse-transcribed into cDNA utilizing the PrimeScript RT Reagent Kit (TAKARA). The SYBR GREEN QuantiTect RT-PCR Kit (Thermo Fisher) was used for real-time PCR analysis, employing the primers listed in Table 1. The levels of mRNA encoding Prg4, GDF5, DKK3, CD73, Col1a1, Col2a1, and SOX9 were normalized to the housekeeping gene Hprt employing the 2-ΔCt formula. The levels of Notch1, Noch2, Notch3, Notch4, Hes1, and Hey1 expression were normalized to the corresponding levels on P0, and the calculation formula 2-ΔΔCt used.
Samples of human articular cartilage were fixed with 10% neutral buffered formalin for 6 h. In the case of staining for CD29 (1:100) antigen retrieval was achieved with boiling DAKO buffer followed by 0.5% pepsin in 5 mM HCl for 20 minutes at 37 °C.
For immunocytochemical staining, antibodies against Osx (1:400 PA5-40411 Invitrogen) or Perilipin-1 (1:400 9349 Cell Signaling Technology) were used. The cells were fixed with 10% neutral buffered formalin for 30 min. Primary antibodies were applied for 3 h and secondary antibodies for 30 min, in both cases at room temperature.
Imaging was performed with a confocal Zeiss LSM880 fluorescence microscope. Positive cells were counted, and the surface was reconstructed with the Imaris 7.4.2 or ImageJ software.
The protocol for staining with safranin O & fast green has been described elsewhere. 70,71 Statistical Analysis All values presented are means ± SD. The normal distribution of all data was confirmed utilizing the Shapiro-Wilk test. The unpaired Student's t test was used to compare 2 groups and one-way ANOVA with Tukey's multiple-comparison test for the comparison of multiple groups. For clarity, the figures illustrate data relative to the appropriate controls; however, for statistical analysis, the control and experimental values were compared.
The standard deviation for relative change (Δ%) was calculated using the formula where X is the experimental values, Y is the control values, and n is the number of data points. The standard deviation for absolute difference (%-%, Fig.  3, P-R) were calculated using the formula where SD x and SD y are the standard deviations for the experimental and control groups, respectively.

Acknowledgments
We would like to thank Karolinska Institutet, Sechenov University, and Gothenburg University for providing the premises, infrastructure, and financial support necessary for the performance of this study. The experiments with MACS and transgenic data were performed with external financial support from the Swedish Research Council to A.S.C (2020-02298

Hey1
Hes related family BHLH transcription factor with YRPW motif 1