Aging induces aberrant state transition kinetics in murine muscle stem cells

These discrete clusters correspond to the freshly isolated and 18 h activated cell time points, referred to as quiescent and activated cells.

“These discrete clusters correspond to the freshly isolated and 18 h activated cell time points, which we refer to as quiescent and activated cells, respectively (Fig. 2B).”

Aged and young cells did not occupy distinct regions in behavioral state space.

“Aged and young cells did not occupy distinct regions in behavioral state space (Fig. 1D).”

Biosynthetic processes activate more slowly in both aged MuSCs and nonLRCs.

“Identifying genetic pathways that are altered in both aged MuSCs and nonLRCs, we find that biosynthetic processes activate more slowly in both populations.”

Classifiers readily discriminate between MuSC ages and proliferative histories.

“Classifiers readily discriminate between MuSC ages and proliferative histories.”

Within the quiescent cell population, a subset of Pax7+ cells occupied the opposite end of quiescent state space from a set of Myod1+ cells.

“Within the quiescent cell population, a subset of Pax7+ cells occupied the opposite end of quiescent state space from a set of Myod1+ cells.”

Analysis of the mean expression in each cluster confirmed non-monotonic changes in Pax7 with activation.

“Analysis of the mean expression in each cluster confirmed non-monotonic changes in Pax7 with activation (Fig. S3A,B).”

From 21,555 MuSC transcriptomes, there are two discrete clusters of cells in transcriptional space.

“From 21,555 MuSC transcriptomes (Table S1), we found two discrete clusters of cells in transcriptional space.”

Aged cells are significantly less motile than young cells.

“Comparing individual cell behavior features between aged and young cells confirmed that aged cells are significantly less motile than young cells.”

The influence of age on cluster preference was significant after controlling for proliferative history.

“Motility state preferences were consistent across bootstrap samples (Fig. S1A,B) and the influence of age on cluster preference was significant after controlling for proliferative history (logistic regression, P<0.01).”

Measurements of cell heterogeneity are comparable between behavioral and transcriptional assays.

“We find that both behavioral and transcriptional-state spaces are continuous across MuSC activation and that measurements of cell heterogeneity are comparable between assays.”

Across all clusters, there were no distinct cell-cycle states observed.

“Across all clusters, there were no distinct cell-cycle states observed, consistent with reports of the first division occurring ≥24 hours after our 18 h time point (Gilbert et al., 2010; Rodgers et al., 2014) (Fig. S2D).”

• cites Gilbert et al., 2010 sourceNot checked yet

  Not checked in the free run.

• cites Rodgers et al., 2014 sourceNot checked yet

  Not checked in the free run.

Aged cells had lower state transition magnitudes.

“Aged cells had lower state transition magnitudes (Fig. 1G; P<0.05, t-test).”

Hierarchical clustering revealed three cell behavior states.

“Hierarchical clustering revealed three cell behavior states (colors in Fig. 1C).”

Two transcriptional clusters were identified within the quiescent and activated cell populations, respectively.

“Two transcriptional clusters were identified within the quiescent and activated cell populations, respectively (Fig. 2D).”

Pseudotiming revealed a distinct sequential ordering for these transcriptional clusters.

“Pseudotiming revealed a distinct sequential ordering for these transcriptional clusters (Fig. 2E).”

Aged cells show a significant preference for the less motile behavior cluster 1 relative to young cells.

“Quantification of the proportion of aged and young cells in each motility state revealed that aged cells show a significant preference for the less motile behavior cluster 1 relative to young cells (Fig. 1E, χ2, P<0.001).”

A subset of cells in the activated population expressed the quiescence marker Pax7+.

“Likewise, a subset of cells in the activated population expressed the quiescence marker Pax7+.”

In aged MuSCs, there are aberrant transition dynamics that lead to significantly delayed activation by both behavioral and transcriptional methods.

“In aged MuSCs, we find aberrant transition dynamics that lead to significantly delayed activation by both methods.”

Activation is the largest source of variation.

“Quantification of the proportion of variance explained by each experimental factor confirmed the qualitative observation that activation is the largest source of variation (Fig. S2A).”

There is heterogeneity within both the quiescent and activated populations.

“However, single cell analysis of these markers in our large sample of MuSCs, revealed heterogeneity within both the quiescent and activated populations.”

Classification of the activated young and aged MuSC pools (each sampled to model physiological LRC:nonLRC ratios) identified a set of 99 genes that provide ∼95% classification accuracy.

“Classification of the activated young and aged MuSC pools (each sampled to model physiological LRC:nonLRC ratios) identified a set of 99 genes that provide ∼95% classification accuracy (Fig. 4H; Table S7).”

Among the genes differentially expressed between aged and young MuSCs, stress responses are downregulated with aging.

“GO analysis of these genes indicated that they represent biosynthetic processes and stress responses, with protein translation processes upregulated with aging and stress responses downregulated (Fig. 4C).”

SVM classification accuracy versus regularization (L1) strength identifies a subset of genes for age discrimination.

“SVM classification accuracy versus regularization (L1) strength identifies a subset of genes for age discrimination.”

No single gene serves as a good biomarker of aging, as assessed using the area under the receiver operating characteristic (AUROC).

“No single gene serves as a good biomarker, as assessed using the area under the receiver operating characteristic (AUROC) (Fig. 4D).”

• AssumptionRests on an assumption: The field of aging biology seeks biomarkers of aging that can be used as an assay of the aged phenotype.

Aged cell gene expression is more variable in the quiescent state.

“We found that aged cell gene expression is more variable in the quiescent state (P<0.05, Wilcoxon Rank Sums).”

A set of candidate 'chrono-variant' genes that change with age was identified, focusing on activated MuSCs where differences are more pronounced.

“We identified a set of candidate ‘chrono-variant’ genes that change with age, focusing on the case of activated MuSCs where differences are more pronounced (Fig. 4F,G).”

GO analysis suggested that the genes differentially expressed between aged and young MuSCs in quiescence are related to protein folding and cellular responses to environmental stresses.

“GO analysis suggested that these genes are related to protein folding and cellular responses to environmental stresses.”

The majority of myogenic markers are not differentially expressed between LRCs and nonLRCs in this assay.

“We found that the majority of myogenic markers are not differentially expressed between LRCs and nonLRCs in this assay (Fig. 5B).”

There is high correlation between the gene expression profiles of young and aged MuSCs.

“Furthermore, we found high correlation between the gene expression profiles of young and aged MuSCs (Fig. 4E).”

Approximately 60% of activated LRCs were in the most activated transcriptional cluster 4, compared with only 40% of the activated nonLRCs.

“Approximately 60% of activated LRCs were in the most activated transcriptional cluster 4, compared with only 40% of the activated nonLRCs (Fig. 5F).”

The differentially expressed genes between LRCs and nonLRCs include known stress response genes, such as heat shock proteins and inflammation signatures, largely enriched within the nonLRC population.

“The differentially expressed genes include known stress response genes, such as heat shock proteins and inflammation signatures, largely enriched within the nonLRC population.”

For nonLRC classification, 104 genes yield a predictive accuracy of ∼96%.

“For nonLRC classification, we identified 104 genes that yield a predictive accuracy of ∼96% (Fig. 4H; Table S6; Fig. S9).”

In the quiescent state, LRCs and nonLRCs did not display significant differences in distribution between the two quiescent transcriptional clusters.

“In the quiescent state, LRCs and nonLRCs did not display significant differences in distribution between the two quiescent transcriptional clusters.”

Both quantification schemes show a non-monotonic relationship between behavioral activation state and Pax7 intensity.

“Both quantification schemes show a non-monotonic relationship between behavioral activation state and Pax7 intensity (95% confidence intervals from 1000 bootstrap samples).”

Module 3 contained genes related to cell cycle regulation and developmental processes.

“Module 3 contained genes related to cell cycle regulation and developmental processes (Fig. S4B).”

The overall difference between young and aged cells did not change significantly across conditions with physiologically observed LRC:nonLRC ratios or equal LRC:nonLRC ratios.

“The overall difference between young and aged cells did not change significantly across these conditions (Fig. S10A,B).”

In quiescence, 200 genes are differentially expressed between aged and young MuSCs.

“In quiescence, we found 200 differentially expressed genes between aged and young MuSCs (Fig. S7A; Table S5).”

The mean gene expression levels between LRCs and nonLRCs have a near perfect correlation in quiescent cells (r=0.99).

“Likewise, the mean gene expression levels between the two states have a near perfect correlation in quiescent cells (r=0.99; Fig. 5C).”

Differential expression analysis revealed that 97 genes were differentially expressed between the LRC and nonLRC pools.

“Differential expression analysis revealed that 97 genes were differentially expressed between the LRC and nonLRC pools (Fig. S11; Table S8).”

During activation, 275 genes are differentially expressed between aged and young MuSCs.

“During activation, we identified 275 differentially expressed genes, suggesting that activation accentuates age-related transcriptional differences (Fig. S7C; Table S6).”

Among the differentially expressed genes between aged and young MuSCs, young cells displayed elevated levels of the stress response heat-shock proteins Hspb1 and Hspa5.

“Among the differentially expressed genes, young cells displayed elevated levels of the stress response heat-shock proteins Hspb1 and Hspa5, suggesting that aged MuSCs are less able to mount appropriate stress responses.”

After activation, LRCs were significantly enriched in the most activated transcriptional cluster 4 (χ2 test, P<0.05).

“However, after activation, LRCs were significantly enriched in the most activated transcriptional cluster 4 (χ2 test, P<0.05).”

Aged cells have significantly decreased behavior state transition magnitude.

“Aged cells have significantly decreased behavior state transition magnitude (two-tailed unpaired Student's t-test, *P<0.05), suggesting delayed activation.”

The majority of transcriptional differences between aged and young cells are independent of proliferative history.

“Although we found that the majority of transcriptional differences between aged and young cells are independent of proliferative history, we also investigated whether LRCs and nonLRCs were transcriptionally distinct from one another.”

Many genes in modules 1 and 3 displayed maximum expression at points in between the most quiescent and most activated states.

“For example, many genes in modules 1 and 3 displayed maximum expression at points in between the most quiescent and most activated states.”

Classifying activated LRC age, a set of 54 genes yields a prediction accuracy of ∼85% (holdout validation).

“Classifying activated LRC age, we identified a set of 54 genes that yield a prediction accuracy of ∼85% (holdout validation; Table S6) (Fig. 4H).”

Gene-wise AUROC analysis demonstrates that no single gene is predictive of MuSC age state.

“Gene-wise AUROC analysis demonstrates that no single gene is predictive of MuSC age state.”

Module 1 contained genes related to mRNA processing and splicing.

“Module 1 contained genes related to mRNA processing and splicing, as determined by gene ontology analysis.”

LRCs and nonLRCs appear to share a transcriptional state space.

“Similar to aged and young cells, LRCs and nonLRCs appear to share a transcriptional state space (Fig. 5A).”

Aged cells display a significant preference for less-motile behavior clusters.

“Aged cells display a significant preference for less-motile behavior clusters (χ2 test, age × behavior cluster contingency table, P<0.001).”

Differential expression analysis identified 174 differentially expressed genes between aged and young MuSCs when comparing both quiescent and activated cells.

“We used differential expression analysis to identify 174 differentially expressed genes between aged and young cells when comparing both quiescent and activated cells (Fig. 4B; Table S4).”

Young cells are significantly more motile than aged cells.

“Young cells are significantly more motile than aged cells, suggesting that aged cells are delayed in activation.”

The most motile and most activated cell behavior states are enriched for Pax7 protein.

“We found that the most motile and most activated cell behavior states are enriched for Pax7 protein (Fig. 3).”

GO analysis of the 174 genes differentially expressed between aged and young MuSCs indicated that they represent biosynthetic processes and stress responses.

“GO analysis of these genes indicated that they represent biosynthetic processes and stress responses, with protein translation processes upregulated with aging and stress responses downregulated (Fig. 4C).”

Differential expression analysis revealed 3864 genes altered by activation.

“Differential expression analysis revealed 3864 genes altered by activation.”

Only 40 of the genes identified for LRC and nonLRC classification are common to both LRCs and nonLRCs.

“Of these genes, only 40 are common to both LRCs and nonLRCs, suggesting that transcriptional aging manifests differently in LRC and nonLRC populations.”

There is no apparent preference in aged cells among the activated transcriptional clusters.

“There is no apparent preference in aged cells among the activated transcriptional clusters (χ2 test, P>0.9).”

In each case (LRC, nonLRC and pooled), classification of activated cells is more effective than classification of quiescent cells.

“In each case (LRC, nonLRC and pooled), we found that classification of activated cells is more effective than classification of quiescent cells, further suggesting that activation reveals transcriptional aging phenotypes.”

GO analysis identified that genes differentially expressed between aged and young MuSCs during activation are associated with catabolic processes downregulated in aging.

“GO analysis likewise identified that these genes are associated with catabolic processes (downregulated in aging), and stress responses (upregulated in aging) (Fig. S7D).”

Among the genes differentially expressed between aged and young MuSCs, protein translation processes are upregulated with aging.

“GO analysis of these genes indicated that they represent biosynthetic processes and stress responses, with protein translation processes upregulated with aging and stress responses downregulated (Fig. 4C).”

There is a non-monotonic regulation for Pax7 across cell behavior states.

“This analysis also indicated a non-monotonic regulation for Pax7 across cell behavior states, as inferred from the ordering of transcriptional clusters.”

The differentially expressed marker genes largely reflect biosynthetic and metabolic pathways.

“Gene ontology (GO) analysis of the differentially expressed marker genes suggests that these genes largely reflect biosynthetic and metabolic pathways, consistent with the notion that myogenic activation corresponds to a dramatic metabolic and geometric rearrangement of cellular state (Fig. S2C).”

Hierarchical clustering identifies four patterns of pseudotemporal gene expression during MuSC activation.

“Hierarchical clustering identifies four patterns of pseudotemporal gene expression during MuSC activation.”

GO analysis identified that genes differentially expressed between aged and young MuSCs during activation are associated with stress responses upregulated in aging.

“GO analysis likewise identified that these genes are associated with catabolic processes (downregulated in aging), and stress responses (upregulated in aging) (Fig. S7D).”

Pseudotime analysis of MuSC activation correctly recapitulated the sequence of ground truth time points.

“Pseudotime analysis of MuSC activation, correctly recapitulating the sequence of ground truth time points.”

WGCNA elucidated two gene modules during activation.

“This interpretation was further reinforced by weighted gene correlation network analysis (WGCNA) (Langfelder and Horvath, 2008), which elucidated two gene modules during activation (Fig. S6A).”

• cites Langfelder and Horvath, 2008 sourceNot checked yet

  Not checked in the free run.

Young cell gene expression is more variable than activated aged cells at the activated time point.

“By contrast, young cell gene expression was more variable than activated aged cells at the activated time point (P<0.0001) (Fig. S8).”

Visualizing myogenic regulatory gene expression across pseudotime reveals that Pax7 does not decrease monotonically.

“Visualizing myogenic regulatory gene expression across pseudotime reveals that Pax7 does not decrease monotonically.”

Aged MuSCs do not segregate discretely in transcriptional space.

“Similar to our cell behavior analysis, we found that aged MuSCs do not segregate discretely in transcriptional space (Fig. 4A).”

The computation of numerical derivatives for pseudotime coordinates, Δpseudotime, showed that young phase points appeared to progress more rapidly from the earliest time steps.

“The computation of numerical derivatives for pseudotime coordinates, Δpseudotime, showed that young phase points appeared to progress more rapidly from the earliest time steps (Fig. 6E).”

We surprisingly found minor transcriptional differences between aged and young MuSCs at a given point in the activation process.

“We surprisingly found minor transcriptional differences between aged and young MuSCs at a given point in the activation process, consistent with previous reports (Cosgrove et al., 2014; Keyes et al., 2016; Sousa-Victor et al., 2014).”

• cites Cosgrove et al., 2014 sourceNot checked yet

  Not checked in the free run.

• cites Keyes et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Sousa-Victor et al., 2014 sourceNot checked yet

  Not checked in the free run.

Comparing inferred coordinates, young phase points progressed more rapidly through the activation process than aged phase points.

“Comparing these inferred coordinates, young phase points progressed more rapidly through the activation process than aged phase points (Fig. 6D).”

When comparing LRCs with nonLRCs, LRCs regress less frequently than nonLRCs in young and aged animals (χ2 test, P<0.0001).

“When comparing LRCs with nonLRCs, LRCs regress less frequently than nonLRCs in young and aged animals (Fig. S13B,C; χ2 test, P<0.0001).”

After many time steps, both young and aged phase points reached similar inferred pseudotime locations as they neared the edge of the observed pseudotime trajectory.

“After many time steps, both young and aged phase points reached similar inferred pseudotime locations as they neared the edge of our observed pseudotime trajectory (Fig. 6D).”

This regression behavior appears to be robust to age-related changes.

“This regression behavior appears to be robust to age-related changes (Fig. 7C; Fig. S13A).”

Ccnd1 was one of the most upregulated genes in LRCs relative to nonLRCs.

“Consistent with earlier cell cycle entry, we observed that Ccnd1 was one of the most upregulated genes in LRCs relative to nonLRCs (Fig. S11).”

Regression is more frequent in activated (∼20%) than quiescent (∼15%) MuSCs.

“Regression is more frequent in activated (∼20%) than quiescent (∼15%) MuSCs.”

By measuring stem cell state transitions directly, we observed that aged MuSCs have dampened state transition rates.

“By measuring stem cell state transitions directly, we observed that aged MuSCs have dampened state transition rates.”

Quantification of the magnitude of RNA velocity across pseudotime in MuSCs revealed that RNA velocity follows a concave curve.

“Quantification of the magnitude of RNA velocity across pseudotime in MuSCs revealed that RNA velocity follows a concave curve (Fig. 6C).”

LRCs incorporated EdU at a higher rate (21%) than nonLRCs (17%) in 20,000 cells across three animals (P<0.001).

“LRCs incorporated EdU at a higher rate (21%) than nonLRCs (17%) in 20,000 cells across three animals (Fig. 5G; P<0.001, Wald test, logistic regression).”

We do not observe bistable transcriptional states between the most-quiescent and most-activated cells in our experiments.

“For instance, although we do not observe bistable transcriptional states between the most-quiescent and most-activated cells in our experiments, stable intermediary activation states might exist in vivo (Rodgers et al., 2014).”

Our data reveal heterogeneity in the stem cell activation process.

“Our data also reveal heterogeneity in the stem cell activation process.”

Quantification of the frequency of lineage regression across pseudotime for cells from the later time point (18 h in vitro) revealed that cells regress more frequently in the later stages of activation observed.

“Quantification of the frequency of ‘lineage regression’ across pseudotime for cells from our later time point (18 h in vitro) revealed that cells regress more frequently in the later stages of activation we observed (Fig. 7B).”

Phase point analysis of RNA velocity vectors suggests that aged cells transition more slowly through transcriptional states during the earliest phases of activation than young counterparts.

“Similarly, phase point analysis of RNA velocity vectors suggests that aged cells transition more slowly through transcriptional states during the earliest phases of activation than young counterparts.”

Highly regenerative Pax3+ MuSCs activate more rapidly than Pax3− counterparts (P<0.001).

“Similar to our experiment in LRCs/nonLRCs, we found that highly regenerative Pax3+ MuSCs also activate more rapidly than Pax3− counterparts (Fig. 5H; P<0.001, Wald test, logistic regression), consistent with a report by De Morree et al. (2019).”

• cites De Morree et al., 2019 sourceNot checked yet

  Not checked in the free run.

Visualization of the density of phase points in transcriptional space showed there were qualitatively few differences in the shape of trajectories between young and aged cells.

“Visualization of the density of phase points in transcriptional space as the number of phase points to pass through a region, showed there were qualitatively few differences in the shape of trajectories between young and aged cells (Fig. 6F; Movie 5).”

This analysis reveals that ∼16% of young MuSCs are regressing in pseudotime during the period of myogenic activation measured.

“This analysis reveals that ∼16% of young MuSCs are regressing in pseudotime during the period of myogenic activation we measure (Fig. 7A).”

RNA velocity estimation showed that each state of MuSC transcription gave rise to a neighboring state in the sequence inferred by pseudotiming.

“RNA velocity estimation showed that each state of MuSC transcription gave rise to a neighboring state in the sequence inferred by pseudotiming (Fig. 6A).”

Murine muscle stem cells (MuSCs) experience a transition from quiescence to activation that is required for regeneration.

“Murine muscle stem cells (MuSCs) experience a transition from quiescence to activation that is required for regeneration, but it remains unknown if the trajectory and dynamics of activation change with age.”

• cites Megeney et al., 1996 sourceNot checked yet

  Not checked in the free run.

• cites Zou et al., 2018 sourceNot checked yet

  Not checked in the free run.

In vitro MuSC activation recapitulates many aspects of in vivo activation.

“However, our in vitro assay allows for a homogeneous activation stimulus across all cells, unlike in vivo injury models and previous work demonstrating that in vitro MuSC activation recapitulates many aspects of in vivo activation (Blau et al., 2015; Cornelison and Wold, 1997; Cosgrove et al., 2014; Gilbert et al., 2010).”

• cites Blau et al., 2015 sourceNot checked yet

  Not checked in the free run.

• cites Cornelison and Wold, 1997Not checked yet

  Not checked in the free run.

• cites Cosgrove et al., 2014 sourceNot checked yet

  Not checked in the free run.

The result of aged-cell preference for less motile states was replicated in a second independent experiment.

“This result was replicated in a second independent experiment (Fig. S1F).”

A defect in activation dynamics may contribute to impaired regeneration.

“However, aged MuSCs have been reported to show impaired activation in multiple studies, suggesting that a defect in activation dynamics may also contribute to impaired regeneration (Brack et al., 2007; Cosgrove et al., 2014; Egerman et al., 2015; Gilbert et al., 2010).”

These results support the different rates model of MuSC aging, in which aberrant dynamics between states contribute to the functional defects observed in aged MuSCs.

“We interpret these results as support for the different rates model of MuSC aging noted above, in which aberrant dynamics between states contribute to the functional defects observed in aged MuSCs.”

This approach allows for the direct observation of cell-state transitions during MuSC activation.

“This approach allows for the direct observation of cell-state transitions during MuSC activation.”

The temporal window of 48 h after plating captures the early stages of MuSC activation, including the switch from a quiescent Pax7+/Myod1− state to a Myod1+ state.

“This temporal window captures the early stages of MuSC activation, including the switch from a quiescent Pax7+/Myod1− state to a Myod1+ state (Cornelison and Wold, 1997).”

• cites Cornelison and Wold, 1997Not checked yet

  Not checked in the free run.

Quantitative measurements of cell motility behavior from time-lapse imaging data are sufficient to resolve states of MuSC activation and state transitions.

“Previously, we demonstrated that quantitative measurements of cell motility behavior from time-lapse imaging data are sufficient to resolve states of MuSC activation and state transitions (Kimmel et al., 2018).”

• cites Kimmel et al., 2018 sourceNot checked yet

  Not checked in the free run.

FACS-isolated cells used here express some markers of early MuSC activation, even immediately after isolation.

“Recent studies have reported that the FACS-isolated cells used here express some markers of early MuSC activation, even immediately after isolation. Therefore, our ‘quiescent’ cell populations probably experience some early activation stimulus during isolation (Machado et al., 2017; Velthoven et al., 2017).”

• cites Machado et al., 2017 sourceNot checked yet

  Not checked in the free run.

• cites Velthoven et al., 2017 sourceNot checked yet

  Not checked in the free run.

Motility state preferences were consistent across bootstrap samples.

“Motility state preferences were consistent across bootstrap samples (Fig. S1A,B) and the influence of age on cluster preference was significant after controlling for proliferative history (logistic regression, P<0.01).”

The discontinuity between the freshly isolated and activated time points suggests that there are intermediary activation states that are not captured by this method.

“The discontinuity between the freshly isolated and activated time points suggests that there are intermediary activation states that are not captured by this method.”

Our 'quiescent' cell populations probably experience some early activation stimulus during isolation.

“Recent studies have reported that the FACS-isolated cells used here express some markers of early MuSC activation, even immediately after isolation. Therefore, our ‘quiescent’ cell populations probably experience some early activation stimulus during isolation (Machado et al., 2017; Velthoven et al., 2017).”

The preference for less motile states among aged cells suggests they exhibit a slower phenotypic change from quiescence to activation.

“The preference for less motile states among aged cells suggests they exhibit a slower phenotypic change from quiescence to activation.”

Aberrant transition dynamics might be related more generally to impaired regenerative potential, as reflected in a comparison of LRCs to less regenerative nonLRCs.

“These findings are reflected in a comparison of LRCs to less regenerative nonLRCs, suggesting aberrant transition dynamics might be related more generally to impaired regenerative potential.”

LRCs might possess a biosynthetic advantage over nonLRCs.

“These expression differences suggest that LRCs might possess a biosynthetic advantage over nonLRCs.”

Activation accentuates age-related transcriptional differences.

“During activation, we identified 275 differentially expressed genes, suggesting that activation accentuates age-related transcriptional differences (Fig. S7C; Table S6).”

LRCs might activate more quickly than nonLRCs.

“The higher proportion of LRCs in the most-activated region of transcriptional space suggests that LRCs might activate more quickly than nonLRCs.”

Activation induces different transcriptional responses in LRCs and nonLRCs.

“Therefore, activation induces different transcriptional responses in LRCs and nonLRCs.”

LRCs express higher levels of Pax7.

“LRCs express higher levels of Pax7 (Chakkalakal et al., 2012), such that rapid activation of LRCs might give rise to a population structure that present non-monotonic changes in Pax7 when averaging across all cells at a given activation state.”

• cites Chakkalakal et al., 2012 sourceNot checked yet

  Not checked in the free run.

This classification result represents the first effective assay to discriminate the age of individual muscle stem cells.

“This classification result represents the first effective assay to discriminate the age of individual muscle stem cells.”

• AssumptionRests on an assumption: The field of aging biology seeks biomarkers of aging that can be used as an assay of the aged phenotype.

Activation reveals transcriptional aging phenotypes.

“In each case (LRC, nonLRC and pooled), we found that classification of activated cells is more effective than classification of quiescent cells, further suggesting that activation reveals transcriptional aging phenotypes.”

LRC-to-nonLRC conversion does not dramatically alter the overall 'magnitude' of age-related transcriptional change.

“The similarity in classification accuracy and divergence magnitude, in the face of changes to the LRC:nonLRC ratio, suggests that LRC-to-nonLRC conversion does not dramatically alter the overall ‘magnitude’ of age-related transcriptional change.”

Transcriptional programs during myogenic activation exhibit a variety of temporal behaviors, including non-monotonic and nonlinear temporal regulation.

“These results support the notion that transcriptional programs during myogenic activation exhibit a variety of temporal behaviors, including non-monotonic and nonlinear temporal regulation.”

Aged MuSCs have significantly impaired regenerative capacity.

“Aged MuSCs have significantly impaired regenerative capacity.”

The levels of gene expression variance are altered by aging in a context-dependent manner.

“This result suggests that the levels of gene expression variance are altered by aging in a context-dependent manner.”

Myogenic activation corresponds to a dramatic metabolic and geometric rearrangement of cellular state.

“Gene ontology (GO) analysis of the differentially expressed marker genes suggests that these genes largely reflect biosynthetic and metabolic pathways, consistent with the notion that myogenic activation corresponds to a dramatic metabolic and geometric rearrangement of cellular state (Fig. S2C).”

Aged MuSCs are less able to mount appropriate stress responses.

“Among the differentially expressed genes, young cells displayed elevated levels of the stress response heat-shock proteins Hspb1 and Hspa5, suggesting that aged MuSCs are less able to mount appropriate stress responses.”

The enrichment of LRCs in transcriptional cluster 4 may also underlie the non-monotonic behavior of Pax7 across pseudotime.

“The enrichment of LRCs in transcriptional cluster 4 may also underlie the non-monotonic behavior of Pax7 across pseudotime (Fig. 2G).”

There are intermediary transcriptional states of myogenic activation that are not simple interpolations of the initial and final transcriptional states.

“Expression peaks and valleys in the non-monotonically-regulated gene modules provide evidence that there are intermediary transcriptional states of myogenic activation that are not simple interpolations of the initial and final transcriptional states.”

The distribution of LRCs and nonLRCs among clusters of activation confirmed a differential activation response.

“The distribution of LRCs and nonLRCs among clusters of activation confirmed a differential activation response.”

Transcriptional aging manifests differently in LRC and nonLRC populations.

“Of these genes, only 40 are common to both LRCs and nonLRCs, suggesting that transcriptional aging manifests differently in LRC and nonLRC populations.”

This decreased frequency of lineage regression may contribute to the more rapid activation of LRCs observed based on transcriptional profiles and EdU incorporation.

“This decreased frequency of lineage regression may contribute to the more rapid activation of LRCs we observe based on transcriptional profiles (Fig. 5) and EdU incorporation (Fig. 5G).”

These observations indicate that phenotypic change at the cellular level might involve considerably more noise.

“Although the macroscopic processes of muscle development and regeneration proceed without these apparent reversals, these observations indicate that phenotypic change at the cellular level might involve considerably more noise.”

Aged cells might progress more slowly than young cells through the activation process in a similar manner to the phase point simulations.

“This result suggests that aged cells might progress more slowly than young cells through the activation process in a similar manner to the phase point simulations.”

Aged cells appear to activate at a slower rate than young counterparts.

“However, aged cells appear to activate at a slower rate than young counterparts.”

These results do not necessarily suggest a subset of cells that permanently fails to activate.

“RNA velocity is a measure of instantaneous change in the cell state, such that these results do not necessarily suggest a subset of cells that permanently fails to activate.”

Consistency in state transition measurements between RNA velocity and cell behavior phenotyping suggests that cell behavior state transitions reflect the underlying transcription state kinetics.

“Consistency in state transition measurements between RNA velocity and cell behavior phenotyping suggests that cell behavior state transitions reflect the underlying transcription state kinetics.”

Activation reveals differences between cell populations that are masked in quiescence.

“As in our age classification experiments, this result suggests that activation reveals differences between cell populations that are masked in quiescence.”

Rapid activation of LRCs might give rise to a population structure that presents non-monotonic changes in Pax7 when averaging across all cells at a given activation state.

“LRCs express higher levels of Pax7 (Chakkalakal et al., 2012), such that rapid activation of LRCs might give rise to a population structure that present non-monotonic changes in Pax7 when averaging across all cells at a given activation state.”

Concave transition rates are consistent with a model in which the peak transition rate represents a 'switch' between two states.

“Concave transition rates are consistent with a model in which the peak transition rate represents a ‘switch’ between two states, corroborating our earlier observations made by cell behavior analysis (Kimmel et al., 2018).”

• cites Kimmel et al., 2018 sourceNot checked yet

  Not checked in the free run.

The set of intermediate transcriptional states a MuSC visits in the course of activation is largely similar between young and aged cells.

“Collectively, results of these phase point simulations suggest that the set of intermediate transcriptional states a MuSC visits in the course of activation is largely similar between young and aged cells.”

These data support a conceptual model in which aging MuSCs exhibit 'different rates' of activation, even if they follow the same trajectory.

“These data support a conceptual model in which aging MuSCs exhibit ‘different rates’ of activation, even if they follow the same trajectory.”

These points support the different rates model of aging pathology.

“Each of these points supports the different rates model of aging pathology outlined above.”

These differences in activation rate suggest that some transcriptional differences between young and aged MuSCs observed at the ensemble level may be the result of differences in the distribution of cells across the activation trajectory.

“These differences in activation rate suggest that some transcriptional differences between young and aged MuSCs observed at the ensemble level may be the result of differences in the distribution of cells across the activation trajectory.”

The heterogeneity of MuSC positions along the activation trajectory arises as an accumulation of differences in the rate of cell state transitions.

“This suggests that the heterogeneity of MuSC positions along the activation trajectory arises as an accumulation of differences in the rate of cell state transitions.”