1. Computational and Systems Biology
  2. Stem Cells and Regenerative Medicine
Download icon

The biphasic and age-dependent impact of klotho on hallmarks of aging and skeletal muscle function

  1. Zachary Clemens
  2. Sruthi Sivakumar
  3. Abish Pius
  4. Amrita Sahu
  5. Sunita Shinde
  6. Hikaru Mamiya
  7. Nathaniel Luketich
  8. Jian Cui
  9. Purushottam Dixit
  10. Joerg D Hoeck
  11. Sebastian Kreuz
  12. Michael Franti
  13. Aaron Barchowsky
  14. Fabrisia Ambrosio  Is a corresponding author
  1. Department of Physical Medicine & Rehabilitation, University of Pittsburgh, United States
  2. Department of Environmental and Occupational Health, University of Pittsburgh, United States
  3. Department of Bioengineering, University of Pittsburgh, United States
  4. Department of Computational & Systems Biology, School of Medicine, University of Pittsburgh, United States
  5. Department of Physics, University of Florida, United States
  6. Department of Research Beyond Borders, Regenerative Medicine, Boehringer Ingelheim Pharmaceuticals, Inc, Germany
  7. McGowan Institute for Regenerative Medicine, University of Pittsburgh, United States
Research Article
  • Cited 0
  • Views 2,041
  • Annotations
Cite this article as: eLife 2021;10:e61138 doi: 10.7554/eLife.61138

Abstract

Aging is accompanied by disrupted information flow, resulting from accumulation of molecular mistakes. These mistakes ultimately give rise to debilitating disorders including skeletal muscle wasting, or sarcopenia. To derive a global metric of growing ‘disorderliness’ of aging muscle, we employed a statistical physics approach to estimate the state parameter, entropy, as a function of genes associated with hallmarks of aging. Escalating network entropy reached an inflection point at old age, while structural and functional alterations progressed into oldest-old age. To probe the potential for restoration of molecular ‘order’ and reversal of the sarcopenic phenotype, we systemically overexpressed the longevity protein, Klotho, via AAV. Klotho overexpression modulated genes representing all hallmarks of aging in old and oldest-old mice, but pathway enrichment revealed directions of changes were, for many genes, age-dependent. Functional improvements were also age-dependent. Klotho improved strength in old mice, but failed to induce benefits beyond the entropic tipping point.

Introduction

Aging is a universal process, and decades of research have gone into understanding the cellular mechanisms underlying aged tissue phenotypes. With the goal of conceptualizing common molecular and cellular mechanisms that underlie the effect of time’s arrow on mammalian tissue health, López-Otín et al. described nine ‘Hallmarks of Aging’ (López-Otín et al., 2013). These hallmarks are: epigenetic alterations, cellular senescence, altered intercellular communication, telomere attrition, nutrient sensing deregulation, mitochondrial dysfunction, stem cell exhaustion, loss of proteostasis, and genomic instability. A criterion in the identification of these hallmarks was that aggravation or attenuation of the biological process results in an accelerated aging or more youthful phenotype, respectively. These hallmarks, therefore, have the potential to pave the way toward the development of therapeutic approaches to counteract the effect of aging on organismal health.

Numerous longitudinal aging studies have positively associated skeletal muscle health with healthspan and longevity, and skeletal muscle strength has even been shown to be a reliable predictor of biological age and mortality (Bell et al., 2019; Voisin et al., 2020; Metter et al., 2002; Frontera et al., 2000). Furthermore, an age-related loss of skeletal muscle mass and function (i.e., sarcopenia) is associated with loss of mobility and increased fall risk (Kinney, 2004; Landi et al., 2012; Imagama et al., 2011; Goodpaster et al., 2006). Tissue-level features of sarcopenia have been described extensively, and include a decreased myofiber size, increased intramuscular fat accumulation, a preferential atrophy of type II (fast-twitch) muscle fibers, and compromised function (Doherty, 2003; Morley et al., 2014; Miljkovic et al., 2015). However, our understanding of the cellular mechanisms underlying sarcopenia is still lacking, a shortcoming that has hindered the development of targeted and specific interventions (Marzetti et al., 2017; Dhillon and Hasni, 2017; Woo, 2017; Wang et al., 2019). Currently, approaches to the treatment and prevention of sarcopenia largely focus on the prescription of physical activity and dietary modifications, strategies that have shown moderate success (Morley, 2018; Waters et al., 2010; Taaffe, 2006). Furthermore, proven pharmacological interventions for sarcopenia do not exist, although drugs such as vitamin D, insulin-like growth factors, testosterone, cetylpyridinium chloride (CPC2), as well as monoclonal antibody treatments have entered advanced clinical trials (Woo, 2017; Drescher et al., 2016; Rolland et al., 2011; Becker et al., 2015; Bauer et al., 2015; https://clinicaltrials.gov/ct2/show/NCT01963598, https://clinicaltrials.gov/ct2/show/NCT02333331, https://clinicaltrials.gov/ct2/show/NCT02575235). Interestingly, many of the drugs currently being tested also act on pathways associated with the longevity protein, α-Klotho (Klotho).

Klotho is best known for its role in delaying age-related pathologies in various organ systems, including skeletal muscle (Kurosu et al., 2005; Kuro-o et al., 1997). The pro-longevity effects of Klotho have been partially attributed to modulation of fibroblast growth factor 23 (FGF23), Wnt, and mTOR pathways (Hu et al., 2013; Liu et al., 2007; Zhao et al., 2015), several of which have also been a targets in sarcopenia research (Woo, 2017; Becker et al., 2015; Yoshida et al., 2019; D'Antona and Nisoli, 2010). With age, circulating Klotho levels gradually decline, and epidemiological studies have revealed that decreased circulating Klotho levels are associated with an accelerated loss of skeletal muscle mass and strength (Yamazaki et al., 2010; Semba et al., 2016). Similarly, mice deficient for Klotho display significant muscle wasting, which is hypothesized to be caused by increased mitochondrial reactive oxygen species (ROS) (Sahu et al., 2018; Phelps et al., 2013). These findings are consistent with the observation that mitochondrial accumulation of ROS leads to a decline in muscle mass and decreased regenerative potential (Sahu et al., 2018; Leduc-Gaudet et al., 2015). Taken together, these studies suggest an unexplored mechanistic link between sarcopenia and a decline of Klotho with increasing age.

In this study, we first thoroughly characterize and compare structural, functional, and transcriptomic changes in skeletal muscle across the lifespan in mice. In order to derive a global metric of the loss of molecular fidelity over time, we used an information-based calculation of network entropy from muscle RNA-seq data. A higher network entropy means a probabilistically greater degree of disorganization, or randomness, in the system. We found that network entropy increases from young (4–6 months) to old (21–24 months) mice, but then plateaus in the oldest-old mice (27–29 months). We then tested the ability of AAV-mediated Klotho delivery (AAV-Kl) to attenuate these changes and counteract sarcopenia in old and oldest-old mice. The impact of AAV-Kl was age-dependent, and whereas old mice displayed a significant improvement in function, AAV-Kl failed to induce a benefit in the oldest-old mice. Furthermore, AAV-Kl regulated a multitude of genes associated with all hallmarks of aging. Unexpectedly, however, the direction of change in response to treatment was, for many genes, dependent on the age of the host. Taken together, these results suggest Klotho enhances skeletal muscle structure and function at the early stages of sarcopenia, but the beneficial effects are lost after the entropic inflection point.

Results

Sarcopenic changes are subtle until mice reach an advanced age

To thoroughly describe the trajectory of sarcopenic alterations according to sex and over time, we characterized muscle structure and function in young (3–6 months), middle-aged (10–14 months), old (21–24 months), and oldest-old (27–29 months) male and female C57BL/6J mice. Age groups were selected to parallel stratifications commonly used in epidemiological studies and correspond, roughly, to individuals aged 20–30, 38–47, 56–69, and 78 + years (Dutta and Sengupta, 2016; Cruz-Jentoft et al., 2019; Sayer et al., 2013). As expected, a progressive decline in tibialis anterior (TA) wet weight (normalized to body weight) was observed over time for both males and females, though the percent decrease at the oldest-old age was significantly greater in male mice (34%) when compared to females (17%; Figure 1A). A decline in TA weight was concomitant with a decline in physiological cross-sectional area (CSA, Figure 1—figure supplement 1A). Bulk decreases in size and cross-sectional area were accompanied by a decrease in individual myofiber cross-sectional area for both sexes, though declines were greater in female mice (Figure 1B,C, Figure 1—figure supplement 1C). The total number of muscle fibers did not change with age (Figure 1—figure supplement 1B).

Myofiber typing was quantified by staining for myosin heavy chain markers of the dominant fiber types in mouse TA muscles. A decrease in the percentage of the more oxidative Type IIA and IIX muscle fibers was observed in the oldest-old, but not old, age group for both sexes (Figure 1C,D, Figure 1—figure supplement 1C). These changes were accompanied by an increase in the percentage of the glycolytic Type IIB phenotype and a decrease in fiber area of all fiber types in oldest-old groups (Figure 1C, Figure 1—figure supplement 1C–G). These results are consistent with a loss of oxidative fiber types with age observed in previous murine studies (Augusto et al., 2017; Giacomello et al., 2020). We also observed increased intramuscular collagen deposition in the oldest-old age groups of both sexes (Figure 1C,E, Figure 1—figure supplement 1C). Unexpectedly, there was no age-related increase in intermuscular lipid accumulation in either sex (Figure 1C; Figure 1F). This contrasts with clinical reports of muscle quality in older individuals (Doherty, 2003; Morley et al., 2014). We did observe a small increase in intramuscular lipid only in old male mice (Figure 1G). Finally, given that denervation commonly accompanies a sarcopenic profile in clinical populations, we probed transcript levels of common markers of muscle denervation (Musk, Ncam1, Runx1) (Aare et al., 2016) in young, middle-aged, old, and oldest-old skeletal muscle RNA-seq data that is publicly available through the Tabula Muris Senis (TMS) database (Schaum et al., 2020). Musk and Runx1 displayed increased expression in oldest-old mice, although sex differences could not be determined due to inadequate power (Figure 1—figure supplement 2).

Figure 1 with 2 supplements see all
Declines in muscle structure are subtle until advanced age.

(A) Tibialis anterior (TA) muscle weight as a percentage of whole-body weight in young (3–6 months), middle-aged (10-14 months), old (21-24 months), and oldest-old (27-29 months) male (N = 57) and female mice (N = 78, one-way ANOVAs). (B) Average fiber cross-sectional area of uninjured male (N = 27) and female (N = 24) mouse TAs across age groups (one-way ANOVAs). (C) Representative images of TA sections stained for laminin (gray), type IIA (purple), type IIX (black/unstained), and type IIB (red) fibers (top, main scale bars = 500 μm, inset scale bars = 250 μm ); Masson’s trichrome staining (middle, 50 µm); and lipidtox staining (bottom, lipidtox = red, laminin = green, scale bars = 50 µm). (D) Percentage of IIA and IIX fibers in the whole TA cross-section of male (N = 27) and female (N = 24) mice (one-way ANOVAs). (E) Collagen area of TA sections across ages and sexes (male N = 17, female N = 19) quantified by Masson’s Trichrome staining (one-way ANOVAs). (F) Intermuscular lipid accumulation in the TA across ages and sexes (male N = 16, female N = 15) quantified by lipidtox staining (one-way ANOVA). (G) Intramuscular lipid accumulation in the TA across ages and sexes (male N = 16, female N = 15) quantified by lipidtox staining (one-way ANOVA). All data presented as mean ± SD (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).

Whole-body muscle strength, as measured by the four-limb hang test, gradually declined with increasing age for both male and female mice (Figure 2A). However, in vivo contractile testing of the TA muscle revealed a significant decline in muscle strength only in the oldest-old mice, regardless of sex (Figure 2B,C,D). Changes in temporal characteristics of the muscle contraction profile, including increased time to peak twitch contraction and half relaxation time, were also most prominent in the oldest-old female group (Figure 2E,F). These temporal changes are consistent with the decreased area of fast-twitch fiber phenotypes and coincide with clinical reports (Miljkovic et al., 2015; Larsson et al., 1978; Lexell and Taylor, 1991). Taken together, these data suggest that old mice display a mild sarcopenic profile and that common features associated with clinical metrics of sarcopenia only become evident in oldest-old mice.

Male and female mice display a progressive loss of muscle function over time.

(A) Whole-body endurance of young (3–6 months), middle-aged (10-14 months), old (21-24 months), and oldest-old (27-29 months) male (N = 67) and female (N = 77) mice measured by the four-limb hang test (one-way ANOVA). 'g'= gravity and 's'=seconds. (B) Peak specific tetanic force production in male (N = 25) and female (N = 24) TA muscles (one-way ANOVAs). (C) Force frequency curves for TA stimulation in male mice (N = 25). (D) Force frequency curves for TA stimulation for female (N = 24) mice. (E) Half relaxation time of the TA muscle following single twitch stimulation across ages and sexes (male N = 25, female N = 24, Kruskal-Wallis tests). (F) Time to maximum force following single twitch stimulation of the TA muscle (male N = 25, female N = 24, Kruskal-Wallis tests). All data presented as mean ± SD (*p<0.05, **p<0.01, ****p<0.0001).

Aging drives a progressive disruption in genes associated with hallmarks of aging

Structural and functional findings across age groups revealed exaggerated phenotypic changes between old and oldest-old female mice. To investigate corresponding molecular-level changes in the muscles of young, old, and oldest-old female hindlimb muscle, we performed RNA-seq analysis from gastrocnemius muscle lysates (Figure 3A). We excluded middle-aged muscles from the analyses since this age group was phenotypically similar to old mice. Principal Component Analysis (PCA) revealed distinct gene expression profiles according to age, with young muscle segregating from the old and oldest-old muscle profiles, the latter of which slightly overlapped (Figure 3B). These findings are consistent with previous transcriptomic studies of human skeletal muscle in which the bulk of gene expression changes occur between young and old age, after which time they plateau (Cellerino and Ori, 2017; Byrne et al., 2019).

Figure 3 with 4 supplements see all
Network entropy increases from young to old mice, after which time it plateaus.

(A) RNA-seq analysis workflow. (B) Principle component analysis (PCA) showing overall gene expression patterns in young (3–6 months), old (21-24 months), and oldest-old (27-29 months) mice (N = 12). (C) Heatmap showing genes associated with aging progression derived from the 100 genes with the highest PC2 loadings. The two highlighted sections show 15/100 genes that have an increasing trend and 23/100 genes that have a decreasing trend. (D) Schematic description of network entropy computation and interpretation. PPI networks were generated based on RNA-seq data. We capture the degree sequence and the edge weights from the network obtained from experimental data in the form of constraints. The ensemble of networks that follow these constraints have similar network features. If the probability distribution is skewed, then it has a low network entropy and, if not, then it is has a high network entropy. (E) Protein-protein interaction (PPI) network entropy computed from transcriptomic data indicates an increase in molecular disorder of hallmarks of aging genes. A non-parametric Kruskal Wallis test (p=0.0741) and Dunn’s post-hoc test were performed. Entropy of young to old changed with p=0.07. Blue shaded portion indicates standard deviation (n = 4). (F) Venn Diagram showing the number of differentially expressed (DE) genes between groups old vs. young, and oldest-old vs. young mice. (G) Network plot denotes the total number of DE genes per hallmark and the corresponding proportion of DE genes that are unique to old and oldest-old, when compared to young counterparts for each hallmark of aging. Edge weights denote the number of genes that are common between the two hallmarks the edge connects. The node sizes are proportional to the number of genes that fall into each hallmark.

Given that distinction across age clusters was most evident for PC2, we probed for the top 100 genes in PC2 that varied according to age. Of these, 59 out of 100 genes were associated with hallmarks of aging and, most prominently, genes associated with altered intercellular communication and nutrient-sensing deregulation. The heatmap shows some PC2 genes displayed a steady trend with aging progression (Figure 3C). Gene Ontology enrichment confirmed that the top 100 genes for PC2 are associated with cellular homeostasis, cellular structure, and intercellular communication (Figure 3—figure supplement 1A,B).

To date, most RNA-seq studies in skeletal muscle have used a bottom-up approach, focusing on differentially expressed (DE) genes and their corresponding pathways. Whereas this approach is valuable for identifying specific signaling pathways that may contribute to sarcopenia, it does not give an integrative view of underlying pathogenic mechanisms. Therefore, we employed a top-down approach, whereby we constructed interactomes according to muscle transcriptional profiles over time. It has been proposed that ‘the common denominator that underlies all modern theories of aging is change in molecular structure and, hence, function.’ (Hayflick, 2007) Although the entropy theory of aging has existed for centuries, entropy has generally been considered in the context of thermodynamics (i.e., ‘dissipation’ of energy at the biomolecular level). While theoretically intriguing, this form of entropy is challenging to quantify. Given that the essence of entropy is a loss of information in a system, we estimated network entropy as described by Menichetti et al., 2015; Menichetti and Remondini, 2014. To do this, protein-protein interaction (PPI) networks were generated from RNA-seq data across young, old, and oldest-old age groups. Given our focus on age-related alterations, only those genes associated with the hallmarks of aging were included in the analysis (López-Otín et al., 2013). In the network, nodes represent the corresponding proteins, edges represent the physical interaction of proteins, and edge weights represent the differential gene expression (Figure 3D). Genes associated with ‘stem-cell exhaustion’ largely overlapped with all other hallmarks, and, hence, were excluded from our analysis.

Consistent with PCA findings, but contrary to phenotypic changes, we observed the greatest increase in network entropy between young and old groups, with little difference between the old and the oldest-old (Figure 3E). This suggests that gene expression changes likely precede the observed functional deficits in oldest-old mice (Figure 2). The same entropy trend emerged even when all genes were considered, suggesting that the subset of genes associated with the hallmarks of aging captures the overall tissue transcriptomic profile without loss of information (Figure 3—figure supplement 2). Furthermore, we found that the increase in entropy between young and old muscles was driven by a relatively small pool of genes that displayed a large magnitude of change (Figure 3—figure supplement 3). In contrast, there were many more DE genes when comparing old and oldest-old muscles, although the magnitudes of change were generally low (Figure 3—figure supplement 3). This suggests a loss of specific changes and a broadening of gene expression alterations across the genome in the oldest-old age group. We cross-checked entropy trends in skeletal muscle of young, middle-aged, old, and oldest-old male mice using RNA-seq data available through the TMS database (Schaum et al., 2020). Unlike females, we saw that entropy continued to increase from old to oldest-old timepoints (Figure 3—figure supplement 4A), which we found interesting considering that male mice showed less severe sarcopenic declines at the oldest-old age (Figures 1 and 2). Also interesting, entropy was lowest at the middle-aged timepoint, then gradually increased.

To probe the translational potential of the network entropy approach, we used open-source RNA-seq data from aging male rats and humans (SRA: PRJNA516151 and GSE164471) (Shavlakadze et al., 2019; Tumasian et al., 2021). We did not present female data for either species due to a lack of sufficient sample sizes. In rats, entropy decreased from young to middle-age, after which time it increased into old age, consistent with the aforementioned trend observed in male mice (Figure 3—figure supplement 4B). Oldest-old muscles were not available in rats. Entropy in humans similarly decreased from young adulthood into the fourth decade of life, after which time it increased, reaching a tipping point in the sixth decade of life, then decreasing slightly (Figure 3—figure supplement 4C). It is interesting to note that the entropic nadir in humans coincides with the commonly observed onset of sarcopenic declines in the fourth decade of life (Walston, 2012; Yazar and Olgun Yazar, 2019). With the goal of providing the broader research community an easily accessible tool for estimating network entropy from transcriptomic and proteomic datasets, we created an open-source Python streamlit web-app (https://network-entropy-calculator.herokuapp.com/). Preprocessing steps are detailed on the Github link (https://github.com/sruthi-hub/sarcopenia-network-entropy).

We next probed for ‘weakest links’ among the hallmarks of aging in which the effect of aging may become manifest prior to the others. Unexpectedly, we found that all of the hallmarks were similarly affected when considering changes between young and old mice and young and the oldest-old mice, although roughly twice as many genes changed in the oldest-old mice (Figure 3F,G). This further implies a dysregulation of the transcriptomic profile in the oldest-old mice, a finding that is consistent with previous reports (Cellerino and Ori, 2017).

AAV-Klotho administration improves muscle regeneration after severe, but not exercise-induced, injury in old mice

Given that a criterion in the identification of aging hallmarks is a susceptibility to interventions that may aggravate or attenuate age-related declines, we next tested whether the declines in muscle function and structure observed in aged mice could be attenuated through Klotho supplementation. As a first step, we confirmed that circulating Klotho gradually declines with age in mice, consistent with human findings (Yamazaki et al., 2010Figure 4A). In both sexes, we also observed a stepwise increase in circulating fibroblast growth factor-23 (FGF23), a known co-factor for Klotho signaling and a key hormone in the regulation of mineral ions and vitamin D (Chen et al., 2018Figure 4B). It was interesting to note that circulating FGF23 levels in the oldest-old females clustered into two different groups, with approximately 40% of the population displaying very high levels compared to the more linearly increasing levels of the remaining mice. This variation may be a result of varied hormonal status across the cohort. Although hormone levels were not measured here, it has previously been shown that 20–40% of female mice spontaneously regenerate follicles after menopause (Diaz Brinton, 2012), which is relevant given that estrogen levels have been shown to regulate circulating FGF23 in both rodents and humans (Ix et al., 2011; Saki et al., 2020).

Figure 4 with 1 supplement see all
Development and validation of an AAV approach for systemic delivery of Klotho.

(A) Changes in circulating Klotho levels measured via ELISA in young (3–6 months), middle-aged (10-14 months), old (21-24 months), and oldest-old (27-29 months) male (N = 41), and female (N = 47) mice (one-way ANOVAs). (B) Changes in circulating FGF23 levels in male (N = 20) and female (N = 39) mice. Red symbols represent undetectable levels and were set to zero (Kruskal-Wallis tests, KO values were excluded from statistical analysis). (C) Schematic of the AAV-Klotho plasmid design. (D) Liver expression of AAV vector genomes quantified via qPCR (N = 33, Kruskal-Wallis test). (E) Circulating Klotho levels measured via MSD-ELISA in young female (N = 33) mice injected with AAV-Klotho at varying doses (Kruskal-Wallis test). (F) Gene count normalized to library size for Klotho in the gastrocnemius muscle of female mice treated with GFP and AAV-Kl (N = 20, one-way ANOVA). (G,H,I) Serum concentration levels for insulin (N = 29), cholesterol (N = 35), and glucose (N = 20) in GFP- and Kl-treated female mice (one-way ANOVA). All data presented as mean ± SD (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).

Next, an AAV vector carrying a liver-specific promoter and mouse Klotho full-length cDNA (AAV-Kl) was constructed and administered via tail vein injection (Figure 4C). Three weeks after administration, qPCR confirmed that the vector was inserted into the genome in the liver and replicated (Figure 4D). Meso Scale Discovery-Enzyme-linked Immune Sorbent Assay (MSD-ELISA) conducted on serum further confirmed a dose-dependent increase in circulating Klotho levels (Figure 4E). Klotho transcript levels in the gastrocnemius muscles of old and oldest-old mice were unaffected by AAV-Kl administration, consistent with use of a liver-specific promoter (Figure 4F). Lastly, the AAV expression of Klotho had no significant effects on non-fasting serum insulin, glucose, cholesterol, or other circulating lipids (Figure 4G,H,I; Figure 4—figure supplement 1).

As a means to confirm the physiological effect of AAV-Kl administration, we tested whether AAV-Kl administration could replicate data from a previous report demonstrating improved regeneration following supplementation with recombinant Klotho protein (Sahu et al., 2018). AAV-Kl (1 × 1010 vector genomes/animal) or an equal titer AAV-GFP control was administered via tail vein injections to old male mice 5 days before bilateral injuries to the tibialis anterior (TA) muscles with cardiotoxin. Histological analysis and in vivo contractile testing was performed 14 days after injury (19 days after AAV injection) (Figure 5A). There was no difference in myofiber cross-sectional area in the injured TAs of animals treated with AAV-Kl when compared to control counterparts (Figure 5B). However, fibrosis was significantly decreased in the TAs of AAV-Kl-treated mice, as evidenced by decreased collagen deposition (Figure 5C,D,E). This decrease in fibrosis was concomitant with a decreased number of fibroadipogenic progenitor cells (FAPs) with AAV-Kl treatment, as quantified by flow cytometry and following a previously published protocol (Yi and Rossi, 2011; Liu et al., 2015; Figure 5F). In the light of previous work demonstrating that Klotho supplementation enhances mitochondrial ultrastructure in aged myogenic cells (Sahu et al., 2018), we also performed TEM and compared mitochondrial integrity in muscles across experimental groups. Indeed, AAV-Kl administration significantly increased the percentage of healthy mitochondria in aged regenerating muscle (Figure 5G,H).

Figure 5 with 1 supplement see all
Gene delivery of Klotho enhances functional muscle regeneration following an acute injury.

(A) Experimental design using old (21–24 months) male mice. (B) Quantification of TA average myofiber cross-sectional area (N = 13). (C) Collagen IV expression in the TA muscle of GFP- versus KL-treated mice (one-tailed Mann-Whitney test, N = 13). (D) Top: Representative images of injured TA muscles stained for collagen IV (green) and DAPI (blue, scale bars = 50 µm). Bottom: Masson’s Trichrome staining of the TA (scale bars = 50 µm). (E) Collagen area percentage in the TA quantified from Masson’s Trichrome staining (one-tailed student’s t-test, N = 16). (F) FAPs to MuSCs ratio in injured TA muscles, as determined by flow cytometry (N = 6, one-tailed student’s t-test). (G) Representative TEM images showing mitochondria in the TA muscle fibers of AAV-GFP vs. KL-treated mice. Aberrant and empty mitochondria show abnormal shape and high proportion of white space respectively (scale bars = 1μm). (H) Quantification of the quality of mitochondria (two-way ANOVA, N = 13). (I) TA specific twitch force produced 14 days post-injury (dpi) (one-tailed Student’s t-test, N = 20). (J) TA maximum specific tetanic force 14 dpi (one-tailed Student’s t-test, N = 20). (K) Change in force production of the TA over time as mice underwent a fatigue protocol consisting of repeated TA stimulation for a total of 7 min, followed by recovery over two 5-min intervals (two-way ANOVA, N = 19). (L) Fold change in whole body endurance compared to one day post injury hang impulse score (Mixed-effects analysis, N = 16). (M) TA peak tetanic specific force for mice 7 days after an eccentric injury treadmill protocol (N = 19). All data presented as mean ± SD (*p<0.05, **p<0.01).

Structural improvements were consistent with functional outcomes, and mice receiving AAV-Kl also displayed increased specific twitch and maximum specific tetanic force production (Figure 5I,J). Resistance to a fatiguing protocol was not affected by treatment, although force recovery following completion of the fatiguing protocol was significantly enhanced in mice treated with AAV-Kl (Figure 5K). Additionally, whole-body strength improved almost twofold over 14 days for mice treated with AAV-Kl; control mice showed no improvement (Figure 5L). These results confirmed that systemically elevated Klotho levels enhance skeletal muscle regeneration, similar to a previous report using intraperitoneal injection of recombinant Klotho (Sahu et al., 2018).

To assess the efficacy of Klotho overexpression in a second, less severe, injury model, we evaluated the ability of AAV-Kl administration to enhance recovery following an exercise-induced injury. To this end, we conducted an eccentric exercise protocol adopted from Armand et al., 2003Figure 5—figure supplement 1A,B. This downhill running model has been shown to induce eccentric muscle injury in tibialis anterior and soleus muscles (Armand et al., 2003; Parise et al., 2008; Breen and Phillips, 2011). Briefly, old male mice that received either AAV-Kl or GFP, as above, were acclimated to treadmill running for 4 days. On the fifth day, mice were placed on a treadmill to run for 60 min at 10 m/min on a 15-degree decline. We assessed exercise adherence according to a previous report in order to confirm that animals in both groups were similarly able to complete the protocol (Ríos-Kristjánsson et al., 2019Figure 5—figure supplement 1C). Mice with an adherence score of less than 0.30 were excluded from the analysis (n = 2 per group; Figure 5—figure supplement 1C). We also confirmed that the protocol induced TA muscle injury using histological analysis (Figure 5—figure supplement 1D,E). In vivo contractile testing performed 7 days after injury revealed that animals treated with AAV-Kl displayed no significant increase in muscle performance (Figure 5M).

Intravenous AAV-Klotho administration enhances muscle structure and function in old, but not in the oldest-old, mice

We next asked whether AAV-Kl administration could reverse age-related declines in muscle quality and function even in the absence of injury. Female mice received either AAV-Kl or AAV-GFP at a titer of 3 × 108 vg/mouse via tail vein injection. This dose was chosen because it provided the greatest improvement in muscle strength, as determined over the course of a preliminary dose-testing study performed in old mice (Figure 6—figure supplement 1A). Fourteen days after injection, mice were euthanized and endpoints collected (Figure 6A). A challenge of studying sarcopenia in mouse models is the high rate of mortality and the onset of confounding pathologies in very old animals. Of the 51 old and oldest-old mice included, only 30 mice were included in final analyses due to incidental morbidity and mortality, with most of these losses occurring prior to randomization (Figure 6B).

Figure 6 with 1 supplement see all
AAV-Klotho enhances muscle function in old, but not oldest-old, mice.

(A) Experimental design and timeline using old (21–24 months) and oldest-old (27-29 months) female mice. (B) Animal inclusion flow chart. Mortality describes mice that died over the course of the experiment. Morbidity describes mice in whom pathology was found at the time of euthanasia . These mice were subsequently excluded from analyses. (C) Representative images showing TA myofiber area (Laminin; green), lipid (red), and DAPI (blue) of the TA 14 dpi in old and oldest-old female mice treated with GFP or Kl. Scale bars = 50 µm. (D) TA wet weight (as a percent of total body weight) of old female mice treated with AAV-GFP or AAV-Kl (N = 10). (E) Quantification of TA muscle-fiber cross-sectional area (CSA) for old female mice (N = 10). (F) Percentage of type IIA and IIX muscle fibers in whole TA cross-sections of old female mice (N = 10). (G) Inter- and intramuscular lipid intensity in TA cross-sections of old female TAs (two-way ANOVA, N = 10). (H) Old female TA maximum specific tetanic force production (one-tailed Student’s t-test, N = 12). (I) Hang-test performance 14 days after injection of AAV-Kl or AAV-GFP, calculated relative to baseline performance (one-tailed Student’s t-test, N = 15). (J) TA wet weight (as a percent of total body weight) of oldest-old female mice treated with AAV-GFP or AAV-Kl (N = 16). (K) Quantification of TA muscle-fiber CSA for oldest-old female mice (N = 12). (L) Percentage of type IIA and IIX muscle fibers in whole TA cross-sections of oldest-old female mice (N = 12). (M) Inter- and intramuscular lipid intensity in TA cross-sections of oldest-old female TAs (two-way ANOVA, N = 8). (N) Oldest-old female TA maximum specific tetanic force production (N = 15). (I) Hang-test performance 14 days after injection, calculated relative to baseline performance (N = 14). All data presented as mean ± SD (*p<0.05, **p<0.01).

Histological analysis of old muscles revealed no improvement in the muscle weight, myofiber cross-sectional area, fiber number, or percentage of type IIA/X fibers with AAV-Kl treatment (Figure 6C,D,E,F, Figure 6—figure supplement 1B). There was also no change in expression of the previously mentioned denervation genes with Klotho treatment (Figure 6—figure supplement 1C). However, intramuscular lipid accumulation was significantly decreased following treatment (Figure 6C,G). AAV-Kl enhanced muscle function in old female mice. Specifically, maximum specific force production improved by 17% (Figure 6H), and whole-body endurance measured via four-limb hang test was 60% greater when compared to control counterparts (Figure 6I). Encouraged by the results in old mice, we then administered AAV-Kl or AAV-GFP to a cohort of oldest-old female mice, which displayed a more severe sarcopenic phenotype (Figures 1 and 2). Unlike old mice, there was no improvement in any of the metrics evaluated (Figure 6C,J–O, Figure 6—figure supplement 1B,D). In fact, AAV-Kl administration significantly increased overall lipid accumulation (Figure 6C,M). These findings suggest that systemic upregulation of Klotho induces therapeutic benefits in old mice, but not in mice of more advanced age.

Klotho affects hallmark of aging genes across age groups, but oldest-old mice exhibit a dysregulated response

To understand the benefit of Klotho in the old group and lack of benefit in the oldest-old group, we revisited the transcriptomic profiles across experimental groups according to the hallmarks of aging. When comparing age-matched control and AAV-Kl treatment groups, we observed approximately three times more DE genes in the oldest-old group treated with Klotho as compared to the old group (Figure 7A). Existing literature suggests that Klotho attenuates hallmarks associated with mitochondrial dysfunction, telomere attrition, and cellular senescence (Sahu et al., 2018; Kuro-o, 2008; Ullah and Sun, 2019; Frenk and Houseley, 2018). However, we found that AAV-Kl affected all the hallmarks similarly (Figure 7B).

Figure 7 with 2 supplements see all
The effect of AAV-Kl administration on genes associated with hallmarks of aging is age-dependent.

(A) Venn Diagram showing the number of differentially expressed (DE) genes between groups treated with AAV-Kl (n = 4) vs AAV-GFP (n = 4) mice. (B) Network plot with each node as a pie chart that denotes the total number of DE genes in that hallmark, and the wedges denote the proportion of DE genes between groups treated with AAV-Kl vs AAV-GFP for each hallmark of aging. The edge weights denote the number of genes that are common between the two connected hallmarks. The node sizes are proportional to the number of genes that fall into each hallmark. (C) Barplots showing GO terms associated with old vs old klotho (green), and oldest-old vs oldest-old klotho (yellow) DE genes. (D) Bar plot showing the top 20 KEGG pathways that change oppositely between old and oldest-old groups after AAV-Kl treatment ranked by largest absolute difference in total accumulation. Total accumulation is a measure of gene perturbation.

To improve resolution between the hallmarks, we probed for pathways associated with the 473 and 1419 DE genes in Klotho-treated old and oldest-old mice relative to their age-matched controls. DE genes responsive to Klotho therapy in old mice were associated with pathways related to inhibition of FGFR4/calcineurin/NFAT signaling. These findings are consistent with previous reports demonstrating that overexpression of Klotho reverts pathogenic FGF23 signaling to more normal tissue homeostasis (Xiao et al., 2019; Ho and Bergwitz, 2021). Intriguingly, however, FGFR4/calcineurin/NFAT signaling did not rise to the top of gene expression changes in response to Klotho treatment in the oldest-old mice (Figure 7C). Of note, neither circulating FGF23 protein levels, nor FGF23 transcript levels, nor the transcript levels of FGF23 and Klotho primary interactors were affected by Klotho treatment in either age group (Figure 7—figure supplement 1). The data suggest that the gain in Klotho may revert pathogenic FGF23 signaling through FGFR4 in old, but not oldest-old, mice (Ho and Bergwitz, 2021).

To better understand the differing responses to Klotho treatment between the old and the oldest-old groups, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Regulation of actin cytoskeleton, cGMP-PKG signaling pathway, and sphingolipid signaling pathway were among the top pathways in which the gene response to Klotho was opposed in the old versus oldest-old mice (Figure 7D, Figure 7—figure supplement 2, Table 1). A full list of affected KEGG pathways is included as Supplementary file 1. Significantly, the reversal of sphingosine/ceramide signaling by Klotho may have facilitated the reduction of intramyocellular lipid accumulation observed in old animals (Figure 6C; Hamrick et al., 2016). This reversal was absent in the Klotho-treated oldest-old mice, suggesting that a dysfunctional metabolism prevails with extreme aging. Further scrutiny of the top three pathways for mechanisms of Klotho’s action revealed cytoskeletal regulation and calcium ion transport as sensitive to Klotho regulation in old mice (Figure 7—figure supplement 2). These are all tightly linked to pathogenic, non-canonical FGF23 signaling (Xiao et al., 2019; Ho and Bergwitz, 2021). In contrast, Klotho overexpression in the oldest-old mice disrupted pathways regulating cell structure, cell membrane integrity, and intercellular communication, which is consistent with the observed decreased contractile capacity (Figure 6N).

Table 1
The top 25 GO terms associated with DE genes from old vs. old AAV-KL treated mice.
TermP-valueAdjusted P-valueOdds RatioCombined Score
regulation of endothelial cell chemotaxis to fibroblast growth factor (GO:2000544)8.17E-101.51332E-08266.6318185579.371
sodium ion export from cell (GO:0036376)7.41E-152.92305E-13149.4724774863.183
sodium ion export (GO:0071436)7.41E-152.92305E-13149.4724774863.183
cGMP biosynthetic process (GO:0006182)1.79E-211.98668E-1990.91162794343.159
cGMP metabolic process (GO:0046068)1.41E-232.15989E-2182.19439254324.53
vascular endothelial growth factor receptor signaling pathway (GO:0048010)2.24E-392.73359E-3647.49148424226.557
cyclic nucleotide biosynthetic process (GO:0009190)6.81E-184.38219E-1697.81524253866.443
adenylate cyclase-inhibiting G-protein coupled acetylcholine receptor signaling pathway (GO:0007197)3.18E-084.17609E-07221.6893423827.416
phospholipase C-activating G-protein coupled acetylcholine receptor signaling pathway (GO:0007207)3.18E-084.17609E-07221.6893423827.416
calcineurin-NFAT signaling cascade (GO:0033173)8.94E-122.35047E-10119.0319633028.225
cellular potassium ion homeostasis (GO:0030007)8.32E-132.71081E-11100.6578952799.85
cellular sodium ion homeostasis (GO:0006883)7.48E-142.64978E-1289.67431192710.338
cellular monovalent inorganic cation homeostasis (GO:0030004)6.58E-152.68262E-1382.38620692690.261
cellular response to forskolin (GO:1904322)3.21E-095.4068E-08133.3090912607.271
response to forskolin (GO:1904321)3.21E-095.4068E-08133.3090912607.271
transmembrane receptor protein tyrosine kinase signaling pathway (GO:0007169)1.92E-674.69774E-6416.45666072528.075
ceramide metabolic process (GO:0006672)4.51E-247.3577E-2245.962470.574
regulation of cardiac conduction (GO:1903779)1.84E-304.49925E-2835.71652152445.428
peptidyl-serine dephosphorylation (GO:0070262)2.96E-106.14116E-09103.9157182279.845
cyclic purine nucleotide metabolic process (GO:0052652)3.13E-192.38927E-1751.93355482212.838
calcineurin-mediated signaling (GO:0097720)2.63E-116.62923E-1089.26940642174.732
cAMP biosynthetic process (GO:0006171)2.28E-126.89002E-1180.52219682158.455
positive regulation of protein kinase B signaling (GO:0051897)1.97E-361.20529E-3322.9166811884.074
regulation of myosin-light-chain-phosphatase activity (GO:0035507)1.09E-071.28925E-06110.8390021776.808
cyclic nucleotide metabolic process (GO:0009187)1.16E-155.76736E-1448.89260971681.61

Discussion

In this study, we evaluated phenotypic and gene expression trajectories across the lifespan of mice. Although declines in structure and function were observed in old mice and more prominently in oldest-old mice, alterations in the expression of genes linked to the hallmarks of aging displayed a non-linear trajectory, with the most significant changes in gene expression profiles occurring between young and old mice. We also evaluated the ability of AAV-Kl to attenuate a sarcopenic phenotype in both old and oldest-old mice. These studies were performed only in female mice, which we found to display age-associated declines in muscle quality and function that are more comparable to clinical indices of sarcopenia when compared to male counterparts. Interestingly, we found that Klotho overexpression enhanced muscle function in the old mice. Notably, Klotho overexpression in old mice reduced pathogenic expression of genes involved with cell structure, ion homeostasis, and signaling for muscle adiposity. Unfortunately, the beneficial effects of Klotho were lost in the animals of most advanced age (Figure 8).

Graphical abstract.

Entropy increases with increasing age before plateauing at old age (21–24 months) in mice (top panel). This is concomitant with a decline in skeletal muscle function, which continues to progress into oldest-old age (second panel). We show that AAV-Kl administration can rescue muscle functional declines when administered to old mice (third panel), but this effect is lost when AAV-Klotho is delivered to oldest-old (27–29 months) mice (bottom panel).

The onset of sarcopenia in mice over time and according to sex has not been well-documented (Romanick et al., 2013; Argilés et al., 2015), as studies of sarcopenia in animal models often exclude the oldest-old cohort (Romanick et al., 2013; Uchitomi et al., 2019; van Dijk et al., 2017). We found that decreased muscle mass and function – two classical markers of sarcopenia – become prominent in the oldest-old group, whereas muscle declines in old mice are relatively modest. Additionally, the onset of sarcopenia in mice appears to be sexually dimorphic, with females displaying slightly accelerated functional declines. These findings are in agreement with one previous clinical report suggesting that sarcopenia progresses more rapidly in females (Masanes et al., 2012), although other studies demonstrated minimal differences between males and females (Morley et al., 2014; Tay et al., 2015). Future studies are needed to elucidate potential mechanisms by which the progression and severity of sarcopenia may be mediated by sex.

Clinical studies performed in older adults have suggested that loss of muscle function is predominantly a result of a loss of muscle quality, as opposed to a loss of muscle mass (Goodpaster et al., 2006). Likewise, in our study, age-related functional declines were not accompanied by a significant loss of muscle mass. Instead, muscle mass and myofiber area were, for the most part, preserved over time. Although muscle mass decreased by only approximately 10% between the old and oldest-old age groups, the accompanying decrease in peak force production was over two-fold greater.

Several transcriptional studies using functional enrichment of differentially expressed genes have reported that age-related changes play a key role in the progression of sarcopenia (Shavlakadze et al., 2019; Barns et al., 2014; Zahn et al., 2006). In the current study, we built on the previous body of work by using an information-based approach to describe integrative changes in gene expression over time, as defined by network entropy. Although an aging system is not necessarily closed (a now overturned criteria for an entropic system) (Hayflick, 2007; Samaras, 1974), evaluation of the interconnectedness of the PPI interactome allows for a single, integrative metric to estimate the ‘disorderliness’ of the system (Menichetti et al., 2015; Menichetti and Remondini, 2014). We chose to use PPI network given that the network edges are based on experimentally validated interactions. As such, the PPI network allows visualization of the functional relationship between genes, an approach that has been used in several studies that explore the mechanism of complex diseases (Zahn et al., 2006; Samaras, 1974; Shannon, 1948; Teschendorff and Enver, 2017). In contrast, gene co-expression networks contain edges that are based on correlation coefficients between gene expression values. Despite the advantages, limitations in the PPI network include the fact that mapping of genes to proteins is not necessarily one-to-one in the PPI network. Moreover, there are a number of proteins for which PPI information has not been documented.

In our study, ‘disorderliness’ of the gene expression profiles over time is represented by the probability of a network connection, mathematically analogous to Shannon, 1948 entropy. Similar to a previous report using microarray gene data from human circulating T-lymphocytes, we observed that network entropy increased from young to old mice (Menichetti et al., 2015). Comparable to entropy change estimates in the context of stem cell differentiation, cancer, and aging (Menichetti et al., 2015; Teschendorff and Enver, 2017; Kannan et al., 2020; Park et al., 2016; Conforte et al., 2019), the absolute magnitude of entropy change in our study is small. Mathematically, this is not surprising given that the spatial network entropy represents the total number of possible configurations (i.e. the flexibility of the system), which has a magnitude on the order of 106. The entropy of the network is dependent on the network architecture, which includes the number of nodes, the number of edges, and the edge weights (i.e. the difference in gene expression). Since fundamental features of network architecture remain relatively stable over time, the change in the magnitude of network entropy represents only a small fraction of all possible configurations (1.1%). However, this small percentage captures the change in expression of a large number of genes (~7600 genes), each of which has the potential to elicit a cascade of downstream phenotypic changes. Indeed, from a physiological perspective, there is growing evidence that very small-scale mRNA changes can lead to extreme changes at the level of protein expression and cellular phenotype (Ruzycki et al., 2015; Cheng et al., 2016). In the context of cancer, for example, a single base mutation led to extensive cell remodeling that extended well beyond the predicted downstream responses, a so-called ‘butterfly effect’ (Hart et al., 2015). Others have similarly shown that cells display a sensitive dependence to initial conditions (in this case, gene expression), and that very small changes in transcriptional regulation and equilibrium may exert potent – perhaps even chaotic – impacts on disease progression (Desi and Tay, 2019; Dorn, 2013).

Whereas T-lymphocytes displayed a sharp decline in entropy in the oldest-old group, potentially suggestive of a ‘survival effect’ (Menichetti et al., 2015), we observed a plateau in entropy after old age (Hayflick, 2007; Menichetti et al., 2015; Samaras, 1974). The data suggest that this peak then plateau may result from a small subset of genes that display highly specific changes from young to old age, followed by a large number of small, non-specific changes between old and oldest-old ages. Our observation that network entropy in mice peaks at old age but that phenotypic changes are not prominent until later in life supports the hypothesis that phenotypic alterations are a lagging indicator of gene expression changes (Frenk and Houseley, 2018). Evaluation of entropy at additional time points across the lifespan would be valuable to better identify an optimal therapeutic window for sarcopenia. A therapeutic challenge lies in how to promote healthier aging and best address the root of these gene expression changes before a sarcopenic phenotype develops. Moreover, although here we focused on quantification of entropy at the gene-level, given the progressive loss of histological and functional muscle integrity, it would be interesting to determine whether entropy at the proteomic level (e.g. using mass spectrometry) is able to capture entropic changes even into very old age.

Given the aforementioned observations, we tested whether modulation of genes associated with the hallmarks of aging is capable of restoring a more youthful skeletal muscle profile. For this, we chose systemic overexpression of Klotho via AAV. AAV allows for consistent modulation of Klotho levels without the need for frequent re-administration and without the challenges of recombinant Klotho protein administration (Cheikhi et al., 2019). Advanced clinical trials using a similar capsid construct are underway (George, 2017; Kishnani et al., 2019), demonstrating the potential translatability of this method. We found that AAV-Kl administration modulated all hallmarks of aging in old and oldest-old animals, although the magnitude of change was manifold times greater in the oldest-old mice. This enhanced response in oldest-old mice was not, however, accompanied by a greater therapeutic benefit. Instead, though we observed a significant increase in muscle function in old mice receiving AAV-Kl, intervention failed to produce a benefit in the oldest-old mice.

Observations of treatment resistance have been reported in advanced age and critically ill populations in other pharmaceutical treatment studies (Rennie, 2009; Burd et al., 2013). In humans, dieting and exercise intervention studies reveal blunted muscle protein synthesis in an older population (greater than 60 years old) (Breen and Phillips, 2011; Durham et al., 2010). Interestingly, this coincides with the same stage in which we observed a tipping point in network entropy (Figure 3—figure supplement 4C). The lack of beneficial response to treatment with increased age is hypothesized to stem from deficiencies in downstream anabolic pathways, that is ‘anabolic resistance’ (Burd et al., 2013; Glover et al., 2008; Burd et al., 2012). In order to identify potentially compromised signaling pathways in oldest-old mice that may preclude a therapeutic benefit of AAV-Kl administration, we compared KEGG pathway activation differences across treatment groups. Old and oldest-old mice displayed opposing responses to AAV-Kl in pathways associated with cellular structure, lipid membrane integrity, and intercellular communication. When we further probed age-dependent pathways that were responsive to Klotho treatment, we found that cytoskeletal signaling, calcium ion transport (cGMP-PKG signaling), and signaling for muscle adiposity (sphingolipid signaling) displayed biphasic effects (Figure 7D). Overwhelmingly, the top GO terms for genes that were decreased by Klotho therapy in the old mice are potentially affected by aberrant FGF signaling, including genes associated with calcineurin (Figure 7C, Table 2). Soluble Klotho binds FGF23 to increase its affinity for FGFR1 and prevents its interaction with FGFR4 to reduce expression of the pathogenic calcineurin/NFAT transcriptome (Ho and Bergwitz, 2021; Han et al., 2020). Indeed, FGF23/FGFR4 signaling is associated with muscle pathogenesis in the heart as well as renal fibrosis (Han et al., 2020; Hao et al., 2021), processes that, if active in skeletal muscle, could explain the decline in muscle quality over time. Our findings support the hypothesis that restoration of normal calcium signaling and calcineurin/NFAT interactome may be important in the development of therapeutics for sarcopenia. Although Klotho is well-known to regulate these pathways by promoting interactions with FGF23 and FGFR1 (70, 71), the age-dependency of Klotho treatment has not been shown.

Table 2
The top 25 GO terms associated with DE genes from old vs. old AAV-Kl treated mice.
TermP-valueAdjusted P-valueOdds RatioCombined Score
negative regulation of transcription from RNA polymerase II promoter in response to stress (GO:0097201)0.0003310.0942424113.1472045105.3348
regulation of protein modification by small protein conjugation or removal (GO:1903320)0.000230.076587599.865793282.66108
response to unfolded protein (GO:0006986)5.3E-060.008962275.7218505369.50717
response to chemokine (GO:1990868)0.0060440.30782999613.130035367.07703
response to granulocyte macrophage colony-stimulating factor (GO:0097012)0.0060440.30782999613.130035367.07703
regulation of histone deacetylase activity (GO:1901725)0.0060440.30782999613.130035367.07703
negative regulation of vasculature development (GO:1901343)0.0060440.30782999613.130035367.07703
regulation of respiratory system process (GO:0044065)0.0060440.30782999613.130035367.07703
negative regulation of myeloid cell apoptotic process (GO:0033033)0.0060440.30782999613.130035367.07703
cellular response to granulocyte macrophage colony-stimulating factor stimulus (GO:0097011)0.0060440.30782999613.130035367.07703
regulation of microtubule nucleation (GO:0010968)0.0060440.30782999613.130035367.07703
cellular response to chemokine (GO:1990869)0.0060440.30782999613.130035367.07703
RNA stabilization (GO:0043489)0.0007870.1697492447.1739634351.27482
chaperone-mediated protein complex assembly (GO:0051131)0.0007870.1697492447.1739634351.27482
mitochondrial translation (GO:0032543)1.36E-060.0049707033.6307731749.05655
3'-UTR-mediated mRNA destabilization (GO:0061158)0.0037350.291113278.7581329648.95746
primary miRNA processing (GO:0031053)0.0037350.291113278.7581329648.95746
cellular response to leucine starvation (GO:1990253)0.0037350.291113278.7581329648.95746
endothelial tube morphogenesis (GO:0061154)0.0037350.291113278.7581329648.95746
negative regulation of inclusion body assembly (GO:0090084)0.0037350.291113278.7581329648.95746
mitochondrial gene expression (GO:0140053)4.92E-050.0257763034.7973619647.58752
regulation of transcription from RNA polymerase II promoter involved in heart development (GO:1901213)0.0100220.412099559.8469964745.32527
mitochondrial electron transport, ubiquinol to cytochrome c (GO:0006122)0.0020720.2813627297.3024298245.12396
mitochondrial translational elongation (GO:0070125)8.63E-060.008962273.697662243.11427
mitochondrial translational termination (GO:0070126)1.22E-050.008962273.5916266740.63

We also found that the sphingolipid signaling pathway was inhibited in old, but not in the oldest-old, mice following Klotho treatment. Previous aging studies in C. elegans and D. melanogaster, suggest that inhibition of sphingolipid synthesis may be beneficial for maintaining proper lipid homeostasis and even extending lifespan (Hla and Dannenberg, 2012; Johnson and Stolzing, 2019). Alterations in the sphingolipid signaling pathway could explain our observation of decreased lipid accumulation in old muscle following Klotho treatment, but a reversed trend in oldest-old mice. Further investigation to more deeply probe this possible mechanism is warranted. It is also possible that the increased lipid accumulation with Klotho treatment in oldest-old mice may be due to a dysregulation in mitochondrial metabolism (Johannsen et al., 2012). Indeed, many of the DE genes in oldest-old Klotho-treated mice were related to mitochondrial translation and electron-transport pathways (Table 3). Metabolic dysregulation would also be consistent with the trend of intramuscular lipid accumulation, as disrupted carbohydrate metabolism could cause an intracellular buildup of stored lipids (Johannsen et al., 2012). Additional mechanistic studies that link FGFR4 activity to sphingolipid/ceramide promoted muscle adiposity (Hamrick et al., 2016) and increased expression of calcium and potassium channels are warranted, as genes associated with these pathways were also the most greatly suppressed by AAV-Kl in an age-dependent manner.

Table 3
Top 25 GO terms associated with DE genes from oldest-old vs. oldest-old AAV-Kl-treated mice.
Klotho mediated Bi-phasic KEGG Pathways Gene Expression
Old + log2FCOldest-Old + log2FC
Regulation of actin cytoskeletonF2-1.92627Bcarl0.40772Rras0.30578Rhoa0.08657
Kng2-1.09411Arpc30.22899Pik3rl-0.94428Itga50.30051
Diaph20.25901Brkl0.17587Rafl-0.19865Hras0.22869
Itga90.39371Arhgap35-0.24801Pfn10.19127
Pdgfrb-0.35413Pip4k2c-0.16982Pfn2-0.20196
ltgax1.6131Apc2-0.52877
Arpc20.1243Cdc420.11714
cGMP-PKG signalingF2-1.9263Dgke-0.3491Pik3rl-0.94428Cyth10.13784
Lpar31.82228Pdgfrb-0.35413Rafl-0.19865Rhoa0.08657
Adcy9-0.20643Dnm3-0.54357Hras0.22869
Rras0.30578Pla224e-0.25837
Sphingolipid signaling pathwayKng2-1.0941Sgppl-0.1274Pik3rl-0.94428Hras0.22869
Map3k50.27371Rafl-0.19865
Ctsd0.17911Rhoa0.08657

Taken together, the data presented here suggest that intervention with AAV-Kl may be more effective in slowing the progression of sarcopenia at an earlier timepoint, rather than rescuing advanced pathology, at which time the transcriptomic response to intervention appears to be more stochastic. An interesting area of future investigation includes the determination of whether network entropy and PPI network architecture may be predictive of the efficacy of therapies designed to counteract the effect of time on skeletal muscle health and function. As an extension of this work, it would also be interesting in future studies to determine whether upregulation of Klotho at a younger age could attenuate functional declines into old, and possibly even oldest-old, age.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Mus musculus)klotho (Kl)MGIMGI:1101771
Strain, strain background (mouse)C57BL/6JNIARRID:IMSR_JAX:000664
Cell line (Homo sapiens)HEK-293HThermo FisherCat#: 11631017
RRID:CVCL_6643
For vector construction
AntibodyAnti-Laminin
(Rabbit polyclonal)
AbcamCat#: ab11575,
RRID:AB_298179
IF(1:500)
AntibodyAnti-Type IIA Fibers (mouse monoclonal)DSHBCat#: SC-71,
RRID:AB_2147165
IF(1:100)
AntibodyAnti-Type IIB Fibers (mouse monoclonal)DSHBCat#: BF-F3,
RRID:AB_2266724
IF(1:100)
AntibodyAnti-Collagen IV (rabbit polyclonal)AbcamCat#: ab6586,
RRID:AB_305584
IF(1:500)
AntibodyAnti-Klotho capture antibody
(goat polyclonal)
R and D SystemsCat#: AF1819,
RRID:AB_2296612
MSD-ELISA(4 µg/mL)
AntibodyAnti-Klotho detection antibody
(goat polyclonal)
R and D SystemsCat#: BAF1819,
RRID:AB_2131927
MSD-ELISA(1 µg/mL)
AntibodyAnti-CD31 (rat monoclonal)Thermo FisherCat#:1 11–0311081,
RRID:AB_465011
FACS(1:500)
AntibodyAnti-CD45 (mouse monoclonal)Thermo FisherCat#: 11-0451-81,
RRID:AB_465049
FACS(1:500)
AntibodyAnti-Sca1 (rat monoclonal)Thermo FisherCat#: 25-5981-82,
RRID:AB_469669
FACS(1:33)
AntibodyAnti-α−7 (rat monoclonal)Thermo FisherCat#: MA5-23555,
RRID:AB_2607368
FACS(1:200)
Peptide, recombinant proteinKlothoR and D SystemsCat#: 1819 KL-050MSD-ELISA
OtherLipidtox stainThermo FisherCat#: H34476IF(1:500)
OtherDAPI stainInvitrogenCat#: D1306
RRID:AB_2629482
IF(1:1000)
OtherTrichrome Stain SolutionSigmaCat#: HT10516
OtherWeigarts’s HaemotoxylinPoly Scientific R and DCat#: S216B
OtherBouin’s solutionSigmaCat#: HT101128
Commercial assay or kitMouse Klotho ELISACloud-Clone Corp.Cat#: SEH757MuLot: L180223640
Commercial assay or kitFGF23 ELISAAbcamCat#: ab213863Lot: GR3326863
Commercial assay or kitAllPrep DNA/RNA 96 kitQiagenCat#: 80311
Chemical compound, drugPoly/Bed 812PolysciencesCat#: 08792–1
Chemical compound, drugCardiotoxinSigmaCat#: 217503
Recombinant DNA reagentAAV-GFPStrobel et al., 2019AAV8-LP1-eGFP
Recombinant DNA reagentAAV-KlothoThis paperAAV8-LP1-mKlotho
Mouse Klotho version of AAV vector
Sequence-based reagentLP-1 promoter (forward)This paperPCR primersGACCCCCTAAAATGGGCAAA
Sequence-based reagentLP-1 promoter (reverse)This paperPCR primersTGCCCCAGCTCCAAGGT
Biological sample (M. musculus)Mouse gastrocnemius muscleNIAFreshly dissected from C57BL/6J mice
Biological sample (M. musculus)Mouse serumNIAFreshly dissected from C57BL/6J mice
Biological sample (M. musculus)Mouse tibialis anterior muscleNIACollected fresh from C57BL/6J via cardiac puncture
Software, algorithmSTAR_2.7.0aDobin et al., 2013RRID:SCR_004463
Software, algorithmRR Project for Statistical ComputingRRID:SCR_001905
Software, algorithmggplot2R Project for Statistical ComputingRRID:SCR_014601R package
Software, algorithmclusterProfilerBioconductorRRID:SCR_016884R package
Software, algorithmVennDiagramR Project for Statistical ComputingRRID:SCR_002414R package
Software, algorithmigraphR Project for Statistical ComputingRRID:SCR_019225R package
Software, algorithmBioMartBioMart ProjectRRID:SCR_002987R package
Software, algorithmFiji-ImageJNIHRRID:SCR_002285
Software, algorithmMuscleJ MacroMayeuf-Louchart et al., 2018RRID:SCR_020995
Software, algorithmGraphPad Prism v9.0GraphPadRRID:SCR_002798
Software, algorithmR ShinyApp resourceThis paperhttps://sruthisivakumar.shinyapps.io/HallmarksAgingGenes/Gene classification based on hallmarks of aging. Building details are provided in ‘Hallmarks of aging genes classification' (Materials and methods).
Software, algorithmPython StreamlitThis paperhttps://network-entropy-calculator.herokuapp.com/Network entropy calculatorapp hosted on Heroku
Software, algorithmGitHubThis paperhttps://github.com/sruthi-hub/sarcopenia-network-entropyCode for RNA -seq network entropy provided with explanation
OtherSTRINGSTRING ConsortiumRRID:SCR_005223Protein interaction database

Animals and ethics

Request a detailed protocol

All animal experiments were approved by the University of Pittsburgh’s Institutional Animal Care and Use Committee. These experiments were conducted in accordance with protocol 17080802 (University of Pittsburgh ARO: IS00017744). Experiments were performed using young (3–6 months), middle-aged (10–14 months), old (21–24 months), and oldest-old (27–29 months) male and female C57 BL/6J mice. Mice were obtained from the NIA Rodent Colony, Jackson Laboratories, and Charles River Laboratories.

Study rigor

Request a detailed protocol

All mice were housed in pathogen-free conditions under a 12 hr light and 12 hr dark cycle. Animals in the AAV portion of the study were housed and experimented on under stricter biosafety regulations (Biosafety level II). Animals with obvious health problems (i.e, tumors, malocclusion, etc.) were eliminated prior to inclusion into the study when possible and were excluded from endpoint analyses if pathologies were noted during experimentation. Given the large amount of variability in physical endurance capacity, animals were randomized into cohorts only if they met the criteria of falling into the 25th-75th percentile for the four-limb hang test (described below). This allowed for elimination of outliers that displayed baseline physical performance falling in the first and fourth quartiles, which may confound results. All animals meeting criteria for inclusion were then sorted by restricted randomization, ensuring comparable baseline physical performance across treatment groups. For in vivo experiments, each experiment was repeated across a minimum of two cohorts, and experimenters performing endpoint analysis were blinded to the experimental groups. Power analyses were performed a priori when possible, assuming a two-sided alternative hypothesis with alpha = 0.05% and 80% power (G*Power 3.1.9.2). Based on the prior findings from the laboratory on peak tetanic force in young and aged animals and after adjusting for unforeseen circumstances estimated at 20%, the final sample size was calculated to be 6–10 animals/group. Each endpoint presents data for biological replicates (individual animals). Some datapoints represent mean values calculated from technical replicates – multiple images captured from one animal sample. This is mentioned in the methods subsections below. The raw data for all experiments reported are included in the accompanying Source Data file.

Cryopreservation and preparation of skeletal muscle tissue

Request a detailed protocol

Post euthanasia , the tibialis anterior (TA) muscles of mice were carefully extracted and wet weight was measured using a standard balance. Immediately after weighing, TAs were frozen in liquid nitrogen-cooled 2-methyl butane for one minute, and stored at −80°C. Slides for histological analyses were prepared using a Thermo Fisher CryoStarNX50 cryostat. Tissue was cut at 10 μm with the cryostat set at −20°C. Each slide captured equidistant tissue sections throughout the entire length of the TA. Slides were stored in a −80°C freezer until use.

Immunofluorescence staining and imaging

Request a detailed protocol

One slide from each experimental replicate was thawed and immediately fixed by covering sections with 2% paraformaldehyde using a pipette for 10 min. After fixing, slides were washed with 1X-Phosphate Buffered Saline (PBS) three times for two minutes each. A hydrophobic barrier was then drawn around the sections, and they were permeabilized with 0.1% triton-X (Fluka 93420) in PBS for 15 min, followed by a 1-hr blocking step using 0.1% triton-X plus 3% Bovine Serum Albumin (BSA, Sigma A7906) in PBS. Primary antibodies were diluted (Table 4) in a solution of 0.1% triton-X, 3% BSA, and 5% normal goat serum in PBS then added onto the sections and incubated overnight at 4°C. One negative control slide per staining set was generated by deleting the primary antibody in the antibody solution. The secondary antibodies used were goat anti-rabbit and goat anti-rat IgG cross-adsorbed antibodies (Invitrogen) and were diluted in the same base solution as the primary antibodies. After washing three times for 2 min in PBS, the secondary antibodies were added to the sections for 1 hr and incubated in the dark at room temperature for one hour. The slides were washed with PBS three times and DAPI stain (Invitrogen D1306, 1:500 dilution in PBS) was added onto the slides for 2 min. Next, slides were washed twice with PBS, dried, and mounted with coverslips using Gelvatol mounting medium (Source: Center of Biologic Imaging, University of Pittsburgh). Slides were allowed to dry for at least 24 hr at 4°C prior to imaging.

Table 4
Antibody/stain list.

Primary antibodies and corresponding dilutions used for immunofluorescence imaging.

Antibody/stainSourceDilution
Rabbit anti-LamininAbcam ab115751:500
Lipidtox Red StainInvitrogen H344761:200
Collagen IVAbcam ab65861:500
Type IIA Muscle FibersDSHB SC-711:100
Type IIB Muscle FibersDSHB BF-F31:100
DAPI StainInvitrogen D13061:1000

Slides were imaged using a Zeiss Observer Z1 semi-confocal microscope. For uninjured slides, regions of interest were randomly selected by navigating to an imaging site of normal nuclear density based on the DAPI channel and capturing an image. At least three images (technical replicates) were captured per slide (each slide contained tissues from one animal), and images were collected at ×20 magnification. Negative control slides were used to threshold for the signal intensity and to set the exposure time for individual channels. The exposure time was kept consistent for each set of stains based on the negative control. For injured tissues, the injury site was located by finding the localized areas of highest nuclear infiltration. Images were then collected at the injury site and at least three images were collected per slide. Intensity was analyzed for Collagen IV using the Zen software and was normalized to image area.

Immunofluorescence staining, imaging, and analysis for muscle fiber types

Request a detailed protocol

Fiber-type staining method was adapted from a previously published protocol (Rao and Mohanty, 2019). Briefly, frozen muscle sections were allowed to dry completely at room temperature, then a blocking solution of 0.5% triton-x, 1% BSA, and 10% goat serum in PBS was added and slides were incubated for 1 hr at room temperature. Antibodies for laminin, type IIA fibers, and type IIB fibers were prepared in 0.1% Tween-20, 1% BSA, and 5% goat serum according to the dilutions in Table 4. This antibody mixture was added onto the sections and incubated overnight at room temperature. Next, slides were washed three times in PBS for five minutes each, then fixed in 10% neutral buffered formalin for 10 min, followed by three more washes. Corresponding secondary antibodies (goat anti-mouse IgM and IgG1, goat anti-rabbit, Invitrogen) were prepared at 1:500 dilution in the same solution used for primary antibodies. Sections were incubated for 1 hr and washed three more times. Then, nuclei were stained with DAPI as above followed by washing, airdrying, and coverslipping using Gelvatol mounting medium.

Whole muscle section images were captured at 20X using a Zeiss Axio Observer with tile stitching settings. One image was captured per animal. These images were analyzed using the Fiji-ImageJ MuscleJ macro developed by Danckaert and Mayeuf-Louchart (Mayeuf-Louchart et al., 2018) and downloaded from Github, which gave outputs for fiber number, area, and type.

Intramuscular vs. intermuscular adipose tissue quantification

Request a detailed protocol

Inter- and intra-muscular lipid deposition was quantified using Fiji-ImageJ software. The laminin rings were converted to a binary scale with positive particles indicating intra-muscular regions. Regions of Interest (ROI) were set and converted to a mask as selected by the positive particles. Total lipid deposition was calculated by measuring the total lipid intensity and then normalizing it to the image area. Intra-muscular lipid deposition was calculated by overlaying the ROI mask onto the lipidtox image where lipid intensity was measured specific to the locations marked by the ROI prior to normalization. Inter-muscular lipid deposition was determined using the resulting difference between total lipid deposition and intra-muscular lipid deposition. Mean values were calculated from three images captured for each sample.

Masson’s trichrome staining and analysis

Request a detailed protocol

Staining was conducted using glass Coplin jars. Slides containing frozen tissue sections were fixed in acetone for 10 min followed by two 5-min PBS washes. The slides were then dipped in 95% ethanol and air-dried for 1 hr. Refixing was done in Bouin’s solution (Sigma HT101128) for 15 min at 56 °C, then the slides were rinsed under running tapwater for 1–5 min until excess yellow coloring dissipated. Next, Wiegert’s Hematoxylin working solution (Poly Scientific S216B) was added for 5 min followed by another tapwater rinse. Then, the slides were stained with Trichome Stain Solution (Sigma HT10516) for 5 min and rinsed with tapwater. Finally, the slides were dehydrated quickly through 95% ethyl alcohol, 100% alcohol (twice), cleared with Xylenes, and mounted with Clearvue Mountant (ThermoScientific 4211). Images were captured with brightfield at 10x using a Nikon Eclipse 50i.

Fiji-ImageJ was used for analysis. First, a threshold was set to remove blank background. The collagen signal was isolated by using color deconvolution and thresholding. The area of the collagen signal was then compared to the net area of the muscle section in the image. Mean collagen coverage was calculated for each sample from three separate images.

Transmission electron microscopy

Request a detailed protocol

Tibialis anterior muscles from the experimental groups were fixed in 2.5% Glutaraldehyde for 24 hr in 4°C. The samples were then washed with PBS three times and cut longitudinally into small pieces. The tissue pieces were submerged in post-fixation aqueous solution comprising of 1% osmium tetroxide, 1% Fe6CN3 for 1 hr, following which they were washed in PBS three times. The tissues were then treated with aqueous 1% osmium tetroxide, 1% Fe6CN3 for 1 hr. Samples were then washed with PBS three times and then dehydrated with a series of 30–100% ethanol. Following this, the samples were embedded in by inverting Poly/Bed 812 (Polysciences, Warrington, PA) embedding resin-filled BEEM capsules on top of the samples. The resin blocked were then cured twice—once overnight at 37°C, and next at 65°C for 2 days. After the curing procedure, the samples were peeled from the coverslip and cross-sectioned ultrathin on a Reichart Ultracut E microtome. Samples were submerged in uranyl acetate for 10 min and 1% lead citrate for 7 min. The investigators performing imaging were blinded to the hypotheses. Data analysis on mitochondrial ultrastructure was performed by a blinded investigator as well following a previously established classification method (Owen et al., 2019).

Four-limb hang test

Request a detailed protocol

Whole body endurance was measured using a modified version of a previously established protocol (Carlson, 2011). The four-limb hang test was used to measure muscular endurance non-invasively and, thus, characterize performance before and/or after treatment. Mice were suspended upside down from a steel mesh grid (1 cm x 1 cm squares) above a custom-built chamber (~30 cm high) with appropriate padding to prevent harm. Four limb strength was evaluated using a Hang Impulse (HI) score (bodyweight in grams*time spent hanging). The time spent hanging was recorded from the time of inversion until all paws are released from the grid. Prior to recorded trials, mice were acclimatized to the grid and the chamber for 5 min. Mice were subjected to five trials in one session with a rest of at least 5 min between trials. Sessions were conducted at roughly the same time of the day (between 10 am and 12 pm) and in the same progression across trials (i.e. the first mouse on session one was also the first on session two). Evaluation of performance was calculated as the average HI of three best trials. Baseline endurance capacity was evaluated prior to experimentation.

For muscle regeneration experiments, injured animals treated with AAV-GFP or AAV-Kl, hang tests were performed before and after AAV- injections, immediately prior to injury, and then again at 1, 7, and 14 days post-injury. For uninjured animals, hang tests were conducted prior to AAV injection, then again at 1, 7, and 14 days after injection.

In Vivo contractile testing

Request a detailed protocol

TA muscle strength of animals in the aged cohorts and treatment cohorts was evaluated using an in vivo testing apparatus (Model 809B, Aurora Scientific, Aurora, ON, Canada). Anesthetized mice underwent surgery to expose the peroneal nerve for stimulation and the Achilles tendon was severed to prevent a counteracting response. The foot was secured onto a footplate in 20° of plantar flexion for maximal responses (Distefano et al., 2013). Initially, single twitch stimulation tests were conducted to measure the peak twitch, time to peak twitch, and the half-relaxation time. Shortly after, tetanic stimulations (350 ms train) at 10, 30, 50, 80, 100, 120, 150, 180, and 200 Hz with 2-min intervals between contractions were induced to obtain the force-frequency curves. Post tetanic stimulation, animals were given a recovery period of 10 min before proceeding to the high-frequency fatiguing protocol wherein muscles were activated every four seconds by a stimulus train at 100 Hz for 7 min. After completion, force recovery was evaluated at 5 and 10 min timepoints. Results of both the single and tetanic stimulations were collected in torque (mN-m) and the absolute force values (mN) were extrapolated by dividing by the length of the footplate (0.03 m). Values were then normalized to physiological muscle cross-sectional area (CSA measured in mm2), as described by Brooks and Faulkner, 1988. Physiological CSA was calculated as musclemassmuscledensitymusclelength (Sacks and Roy, 1982), with muscle density approximated as 1.06 g/cm3 as per previous reports (Méndez, 1960). Results of the high-frequency fatigue and recovery test were analyzed as percentage values of the initial response.

Hallmarks of aging genes classification

Request a detailed protocol

First, we compiled a list of key functions and changes for all systemic hallmarks of aging and cellular hallmarks of aging (López-Otín et al., 2013; Frenk and Houseley, 2018; Rebelo-Marques et al., 2018; DiLoreto and Murphy, 2015). Next, we created multiple gene lists by searching for different combinations of gene ontology (GO) terms on the MGI GO search database (Supplementary file 1). Batch query forms were used to identify genes for each hallmark. A table of (genes) x (hallmarks) was formed by genes classified into different hallmarks of aging. Note that GO terms associated with multiple hallmarks of aging were accounted for in this process.

From this step onwards, we focused on each hallmark of aging, as opposed to a specific gene or a GO term. The goal of the study was to identify and target the genes associated with the hallmarks of aging with or without intervention. We created an R ShinyApp webpage for the research community to classify genes into corresponding hallmarks of aging and vice versa (https://sruthisivakumar.shinyapps.io/HallmarksAgingGenes/).

RNA-seq analysis

Request a detailed protocol

We performed RNA sequencing on gastrocnemius muscle collected from female mice injected with AAV-GFP and AAV-Kl. Four animals were used for each of the five groups: young GFP, old GFP, old Klotho, oldest-old GFP, and oldest-old Klotho. Raw data has been submitted to GEO database (GSE156343). Raw fastq files were converted to gene counts after alignment using STAR_2.7.0a to reference genome GRCm38/mm10 (chromFa.fa, UCSC). The visualizations were generated using R packages, ggplot2, clusterProfiler, VennDiagram, and igraph. DE analysis was performed with Deseq2. DE genes with p-value less than 0.05 were defined as significantly different.

To identify genes associated with the hallmarks of aging, a list of 77,708 genes derived from Mouse Genome Informatics’ batch query was generated based on the GO terms related to cell biology of aging. The ratio of DE genes that changed under each hallmark was visualized in the barplot and then network plot. Shannon network entropy was computed after mapping genes to proteins using Biomart in R.

Network entropy methodology and interpretation

Pre-processing RNA-seq to protein-protein interaction (PPI) networks

Request a detailed protocol

Raw gene count matrix was used to generate PPI networks. For each sample, we generated a node-list and an edge-list. Absolute gene expression values were used to create the node-list, while the edge-list was generated with mouse PPI networks obtained from STRING database version 11.0. If the gene counts were zero, then the nodes were not considered part of the PPI network. This process was done for all the genes, as well as for only the genes classified into hallmarks of aging.

Network entropy interpretation

Request a detailed protocol

In the field of information theory, Shannon entropy is the measure of surprise of an outcome. For example, a fair coin toss has two equally probable outcomes. This would be the case of maximum surprise, or maximum Shannon entropy. On the other hand, if the coin is rigged, and probability of obtaining heads is 0.9 then one will not be as surprised to get heads. This is the case of lower Shannon entropy. When this concept is extrapolated to graph theory, we aimed to predict the presence or absence of an edge in the graph. If network entropy is high, this means that it is harder to predict the network connections with the knowledge one has. A higher network entropy is bound to happen when the knowledge one has about the system is insufficient or highly variable.

The principle of maximum entropy states that the probability distribution best representing the current state of knowledge on a system is the one that maximizes the Shannon entropy, given the prior knowledge about the system (Cimini et al., 2019). Using this method, we gain maximally unbiased information in the absence of complete knowledge. In the content of network biology, the ensemble of most probable network configurations are the ones that have maximum entropy, given certain constraints.

Our goal was to capture the change in gene expression distribution across age groups using a network entropy estimate, as has been previously described (Menichetti et al., 2015; Menichetti and Remondini, 2014). We obtained nodelists from the RNA-seq experiment, and edgelists obtained from existing databases (e.g. STRING) that have PPI network. In addition, we accounted for relative gene expressions by weighting the edges of the network to incorporate the current knowledge we have about the PPI network connections. Hence, we hypothesized that changes in gene expressions will affect the number of nodes, number of edges and the edge weights–a phenomenon that could potentially be captured by the network entropy metric.

Conventionally, maximum entropy is used to predict the probability distribution. But in this study, we used network entropy values corresponding to the most probable ensemble of PPI networks. This allowed us to quantify the change in transcriptional noise and variation in gene expressions into a single integrative metric.

Network entropy calculation

Network entropy is a constraint-based optimization problem that is solved using Lagrangian multipliers. The objective function to maximize is the Shannon entropy of the network. A value of network entropy will correspond to a probability distribution for edges. There are two constraints for the optimization problem. The first constraint imposes a degree sequence of the network, whereas the second constraint imposes restriction on the edge weights of the network. The concept is explained with an arbitrary simple example, as shown below.

Gene counts for node-list (15 genes)Age and # nodes (n)
50070000000100583500675105090Young (n = 8)
502704012309559375045004580Young (n = 10)
6006900105011065378256794500Old (n = 11)
595721000901000300060404250Old (n = 7)
  1. Let edge-list be: (1-3), (3-7), (5-8), (8-9), (9-12), (10-14).

Constraint 1

Request a detailed protocol

The degree sequence was fixed based on the edge-lists. Zeroes are not considered in the network as nodes.

Constraint 2

Request a detailed protocol

The edge weights were calculated by taking the difference between gene counts between the connected nodes. To normalize for different number of nodes in the network, the edge weights are binned with number of bins equal to square root of the number of nodes in the network.

In the arbitrary example above, there would be sqrt(n)~3 bins. [0–240, 241–482, 483–721]. The binning distribution will vary according to the relative gene expression changes. If the binning is similar for all the samples, then network entropy values will be very close to each other. If the binning is different, network entropy values will be different.

Network characterization

Request a detailed protocol

The main global features are described for both Figure 3E and Figure 3—figure supplement 2 in Supplementary file 2. These global features include: number of nodes, number of edges, spatial entropy, link density, the number of disconnected nodes (nodes with no edges), and connected components (nodes with greater than or equal to one edge).

Serum collection

Request a detailed protocol

Animals were anesthetized using isoflurane and remained anesthetized throughout the duration of serum collection. The animal was placed in a supine position and its paws were secured to spread out the torso. A small incision was made with scissors and forceps in the skin above the xiphoid process of the sternum and expanded to reveal the bottom of the rib cage. The peritoneum in the same area was then cut to reveal the underside of the diaphragm. The diaphragm was cut and cleared away with scissors, thus exposing the bottom of the heart. A clamp was used to hold the ribcage out of the way, and a 25 ½ G needle attached to a 1 mL insulin syringe was used to collect blood directly from the apex of the heart. Using this method, roughly 800 µL of blood could be collected from each animal. After collection, the animals were euthanized via cervical dislocation. The collected blood was allowed to clot in a 2 mL tube at room temperature for one-hour. Then, the blood was spun at 16,100 g for 15 min at 4°C in a microcentrifuge. The serum was collected and aliquoted into 50 µL tubes and stored at −20°C until used.

Circulating Klotho and FGF23 quantification

Blood collected from experimental and control animals was clotted at room temperature for 1 hr, after which it was spun at 16,100 G for 15 min in a fixed rotor centrifuge at 4°C in order to separate serum. This serum was diluted 1:25 and 1:6 in diluent buffer to quantify Klotho (Cloud Clone Corp), and FGF23 (Abcam), respectively using commercially available ELISA kits.

Klotho ELISA

Request a detailed protocol

The Klotho ELISA was conducted according to the protocol using the Cloud Clone Corp. ELISA Kit for mouse Klotho (SEH757Mu, Lot: L180223640). This kit was validated by Sahu et al., 2018. Briefly, blood serum samples and standards were diluted in PBS (1:25) and 100 µL of the samples were added in duplicates to a 96-well plate pre-coated with a biotin-conjugated antibody specific for Klotho detection. This plate was then incubated at 37°C for 1 hr. Samples and standards were removed and 100 µL of biotin-conjugated antibody cocktail was added to each well and incubated at 37°C for 1 hr. The microplates were washed three times with washing buffer provided by the kit (2 min per wash), using an automated plate washer (BioTek 50TS). Following this, 100 µL of Avidin conjugated Horseradish-Peroxidase was added to each well and incubated at 37°C for 30 min. The plate was washed with buffer five times, following which 90 µL of tetramethylbenzidine (TMB) substrate was added to each well and the plate was incubated at 37°C for 20 min. A sulfuric acid-based stop solution was added to terminate the reaction and optical density of each well was measured with a Spectramax M3 plate reader (Molecular Devices) at a wavelength of 450 nm. The analysis was conducted using Microsoft Excel by generating a 2P equation for concentration based on the standard curve and plotting the averaged absorbance values for each sample along this curve.

Any sample displaying evidence of hemolysis was not included in data analysis, as we have seen that hemolysis interferes with values. Samples were not subjected to multiple free-thaw cycles, as we have observed that this may affect α-Klotho detection (Cheikhi et al., 2019).

FGF23 ELISA

Request a detailed protocol

The FGF23 ELISA was performed according to the protocol provided by the manufacturer (Abcam, ab213863, Lot: GR3326863). Briefly, blood serum samples and standards (diluted 1:6 in sample diluent buffer) were added in duplicates to a 96-well plate pre-coated with antibody against FGF23 and incubated for 90 min at 37°C. The plate content was then discarded and the plate was blotted. Following this, a biotinylated anti-mouse FGF23 antibody was added to all wells and incubated for 60 min at 37°C. Next, the plate was washed three times with 0.01M PBS using the same plate washer as described above. A total of 100 µL of avitin-biotin-peroxidase complex was added to each well and incubated again for 30 min at 37°C. The plate was washed five more times, following which 90 µL of TMB substrate was added to each well and incubated in the dark for 20 min at 37°C. Finally, 100 µL of TMB stop solution was added and the plate was immediately read at 450 nm using the same plate reader as described above. Analysis was conducted as described above. Hemolyzed samples were not included, and each sample was frozen once prior to the experiment.

Eleven of the samples returned absorbance values lower than the lowest value of the standard curve. Rather than extrapolating and recording negative concentration values for these samples (which would not be biologically coherent), we assumed that these samples did not have detectable levels of FGF23. Thus, their concentration was recorded as zero. These samples have been marked with red symbols in the figure.

Meso-Scale discovery (MSD) Klotho ELISA

Request a detailed protocol

Detection of circulating Klotho levels from mouse blood samples was performed by using an ELISA assay on the MSD platform. Standard-bind MSD plates (#L15XA-1, MSD) were coated with mouse Klotho capture antibody (#AF1819, R and D) at a final concentration of 4 µg/mL in PBS for 1 hr at room temperature. Subsequently, plates were washed three times using wash buffer (PBS + 0.05% Tween-20) at 300 µL/well followed by an incubation with 3% blocker A solution (R93BA-2, MSD) for 1 hr at room temperature. Dilutions of serum samples and murine recombinant Klotho standard (#1819 KL-050, R and D) were prepared in 1% blocker A solution and added to the MSD plate in a final volume of 25 µL/well after the plate was washed three times. Samples were incubated for 1 hr at room temperature followed by three washing steps. Detection antibody (#BAF1819, R and D) was diluted to 1 µg/mL in 1% Blocker A solution in PBS and SULFO-tag labeled streptavidin (#R32AD-5, MSD) was diluted to 0.5 µg/mL in 1% Blocker A in PBS. Both dilutions were added to the plate simultaneously (25 µL/well each) and incubated for 1 hr at room temperature. After three washing steps, 1x Read Buffer (#R92TC-2, MSD) diluted in water was added at 150 µL/well. Electrochemiluminescence was detected in the MSD Sector Imager 600.

Metabolite analysis

Request a detailed protocol

Serum metabolites were analyzed by the Mouse Metabolic Phenotyping Center at the University of Cincinnati.

AAV production

Request a detailed protocol

Recombinant AAV8 was used as vector for expressing murine Klotho or enhanced green fluorescent protein (GFP). The genomic constructs contained the LP-1 promoter, GFP or a codon-usage optimized sequence of murine alpha klotho (Geneart, Regensburg, Germany) with an N-terminal secretion signal originating from human CD33 (sequence: MPLLLLLPLLWAGALA) and a C-terminal V5-tag epitope (Southern et al., 1991), followed by a WPRE sequence and an SV40 polyA signal. The expression cassettes were flanked by AAV2-derived inverted terminal repeats. The recombinant AAV8-LP1-mKlotho and AAV8-LP1-eGFP vectors were prepared as previously described (Strobel et al., 2019). Briefly, human embryonic kidney cells (HEK-293H cells, Thermo Fisher Scientific) were cultured in Dulbecco's modified Eagle's medium +GlutaMAX I+10% fetal calf serum (Gibco/Thermo Fisher Scientific) and transfected as previously described (Strobel et al., 2015). AAV purification via polyethylene glycol precipitation, iodixanol gradient, ultrafiltration, and sterile filtering was conducted as described previously (Strobel et al., 2019). Genomic titers of purified AAV8 vector stocks were determined by isolation of viral DNA and subsequent qPCR analysis using primers specific for the LP-1 promoter with the following primer sequences; forward: GACCCCCTAAAATGGGCAAA. reverse: TGCCCCAGCTCCAAGGT.

AAV tail vein injection

Request a detailed protocol

Aliquots of AAV-Kl (1.79e13 Vg/ml) and AAV-GFP (3.48e12 Vg/ml) were stored in −80°C and thawed within an hour of use. Animals were restrained using a tail illuminator restrainer from Braintree Scientific Inc and injected with either a GFP expressing non-targeting control or AAV-Kl via the tail vein using a 31G insulin syringe. Viral solutions were diluted in Dulbecco’s phosphate-buffered saline (DPBS) to a final volume of 100 µL/mouse. A successful injection entailed full administration of the viral solution into the vein smoothly, without swelling or bleeding following syringe removal. Only successfully injected mice were included in the study (a total of six were excluded).

AAV-Klotho in vivo dose response study

Request a detailed protocol

Female C57Bl/6 mice (9–12 weeks old) with a body weight of 19–21 g were purchased from Charles River Laboratories. AAVs were diluted to the desired concentrations in Dulbecco’s phosphate-buffered saline (DPBS) and administered into the tail vein under light isoflurane anesthesia. The doses tested were 1 × 108, 3 × 109, and 1 × 1010 vg/mouse AAV-Kl compared with 1 × 1010 AAV-GFP control. The final volume for injection was 100 μL per mouse. Three weeks after injection with AAV-GFP, AAV-Kl or DPBS, animals were sacrificed and blood samples were collected for the detection of circulating Klotho levels by MSD-ELISA. Livers were also collected to quantify AAV vector genomes. For vector DNA isolation flash-frozen liver samples, were homogenized in 900 μL of RLT buffer (79216, Qiagen), using a Precellys-24 homogenizer and ceramic bead tubes (KT03961-1-009.2, VWR) at 6000 rpm for 30 s. Homogenates were further processed by applying standard Phenol-chloroform extraction. Finally, vector DNA was purified by using the AllPrep DNA/RNA 96 kit (80311, Qiagen) according to the manufacturer’s instructions. Detection of AAV vector genomes was performed by qPCR using primers specific for the LP-1 promoter (forward: GACCCCCTAAAATGGGCAAA. reverse: TGCCCCAGCTCCAAGGT). For quantification, a standard curve generated by serial dilutions of the respective Klotho expression plasmid was used. qPCR runs were performed on an Applied Biosystems ViiA 7 Real-Time PCR System. Animal experiments in this study were approved by the local German authorities (Regierungs-präsidium Tübingen) and conducted in compliance with the German and European Animal Welfare Acts.

AAV-administration, injury, and sarcopenia experimental models

Request a detailed protocol

For the muscle regeneration experiment, the dose of GFP and AAV-Kl used was 1 × 1010 vector genomes(vg) per animal. Five days following injection, each mouse received a bilateral intramuscular injury to the tibialis anterior muscle (TA) using 10 µL of 1 mg/mL cardiotoxin (Naja pallida, Sigma, molecular weight 6827.4), and animals were euthanized 14 days after injury. In uninjured animals, the dose used was 3 × 108 vector genomes per mouse, and mice were euthanized 14 days after AAV injection.

Flow cytometry

Request a detailed protocol

To quantify the ratio of fibroadipogenic progenitor cells (FAPs) to MuSCs (MuSCs) following muscle injury, old mice were injected with GFP or AAV-Kl at 1 × 1010 vg/animal (9 animals per group) via tail vein. As with the previous injury experiment, animals received a bilateral cardiotoxin injury to the TA 5 days following AAV injection. After 3 days, animals were euthanized via cardiac puncture, and the TA muscles from each animal were collected and digested according to a previously established protocol for flow cytometry (Yi and Rossi, 2011; Liu et al., 2015). Though we began with nine mice in each group, we pooled together sets of TAs from 3 mice to ensure an adequate amount of tissue for digestion and sorting . Thus, the final sample size for each group was n = 3 (3 pools of 3 sets of TAs). Following digestion, FAPs (CD31-, CD45-, Sca1+, α−7-) were sorted from MuSCs (CD31-, CD45-, Sca1-, α−7+) via fluorescence-activated cell sorting (FACS). The ratio of FAPs:MuSCs was quantified and recorded for each experimental replicate.

Exercise-induced injury

Request a detailed protocol

To determine how muscle administration of AAV-Kl affects adaptability to exercise, we adopted the eccentric exercise protocol created by Armand et al., 2003. Mice were acclimated to treadmill (Columbus Instruments) running over 4 days. On day 1 they ran at 6 m/min for 20 min. Day 2 was at 8 m/min for 20 min. Day 3 was at 8 m/min for 20 min on a 15-degree decline. Day 4 was at 10 m/min for 20 min on a 15-degree decline. The mice then were given 1 day of rest after acclimation before the eccentric injury protocol, which consisted of running at 10 m/min for 60 min on a 15-degree decline. A slight electric shock and gentle prodding were used to ensure adherence. We also assessed exercise protocol adherence according to a previously established method (Ríos-Kristjánsson et al., 2019) to ensure equal performance between groups. Mice that scored below 0.3 were excluded from further analysis due to their inability to complete the injury protocol. Other studies used longer runs and faster speeds to induce injury, but these previous studies were conducted in young mice. Old mice were not capable of running at such speeds and times. Still, to confirm the efficacy of the protocol, histology was conducted to visualize the extent of injury.

On the first day of acclimation, mice were randomized into two groups (n = 6 per group, repeated in two cohorts), one receiving AAV-Kl at 1 × 1010 vg per mouse and one receiving the GFP control via tail vein injection as described above. The timing of administration was decided so that AAV injection would fall 5 days prior to muscle injury, similar to our cardiotoxin injury model. Seven days after the muscle injury protocol, contractile testing was conducted, mice were euthanized, and serum and TA muscle were collected. One mouse in the AAV-Kl group was removed from analysis for maximum specific tetanic force production due to an error in testing. TA muscles were sectioned, stained for DAPI and laminin, and imaged at 10x using a Zeiss Observer Z1. Centrally nucleated fibers were quantified using the MuscleJ macro (Mayeuf-Louchart et al., 2018). The mean value for each sample was calculated from three separate images, and the mean percentages were compared between the mice that had undergone the eccentric protocol vs. uninjured control mice.

Statistical analysis

Request a detailed protocol

Analyses were performed using GraphPad Prism version nine software. Shapiro-Wilk and Levene’s tests were initially performed to assess normality of data and equality of variances, respectively. If assumptions of normality and homogeneity of variances were met, a Student’s t-test was performed while comparing two groups, or a one-way analysis of variance (ANOVA) test with Tukey post-hoc comparison. When conditions for normality were not met in two group comparisons, the groups were compared using Mann-Whitney U test. A Welch’s test was applied when there were differences between the standard deviations of two groups. When conditions of normality or homogeneity were not met for comparisons of three or more groups, a Kruskal-Wallis test with Dunn’s post-hoc comparison was used All results were expressed as mean ± standard deviation. Statistical significance was established, a priori, at p≤0.05.

Data availability

Sequencing data has been deposited in GEO accession: GSE156343.

The following data sets were generated
    1. Clemens Z
    2. Sivakumar S
    3. Pius A
    4. Sahu A
    5. Cui J
    6. Barchowsky A
    7. Ambrosio F
    (2021) NCBI Gene Expression Omnibus
    ID GSE156343. The biphasic and age-dependent impact of Klotho on hallmarks of aging and skeletal muscle function.
The following previously published data sets were used

References

    1. Augusto V
    2. Padovani CR
    3. Campos GER
    (2017)
    Skeletal muscle fiber types in C57BL6J mice
    Journal of Morphological Sciences 21:01.
  1. Conference
    1. Carlson G
    (2011)
    The use of four limb hanging wire tests to monitor muscle strength and condition over time
    Treat-NMD Neuromuscular Network SOP.
    1. Lexell J
    2. Taylor CC
    (1991)
    Variability in muscle fibre Areas in whole human quadriceps muscle: effects of increasing age
    Journal of Anatomy 174:239–249.
    1. Méndez J
    (1960)
    Density and composition of mammalian muscle
    Metabolism 9:184–188.
    1. Menichetti G
    2. Remondini D
    (2014)
    Entropy of a network ensemble: definitions and applications to genomic data
    Theoretical Biology Forum 107:77–87.
    1. Taaffe DR
    (2006)
    Sarcopenia-exercise as a treatment strategy
    Australian Family Physician 35:130–134.
    1. Woo J
    (2017) Sarcopenia
    Clinics in Geriatric Medicine 33:305–314.
    https://doi.org/10.1016/j.cger.2017.02.003

Decision letter

  1. Yousin Suh
    Reviewing Editor; Columbia University, United States
  2. Jessica K Tyler
    Senior Editor; Weill Cornell Medicine, United States
  3. Daniel Remondini
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The manuscript provides a comprehensive characterization of the trajectory of murine muscle aging and correlates that to the ability of Klotho to exert beneficial effects on muscle health. The novel entropy-based modeling defines the trajectory of murine muscle aging, which will certainly be of interest to the community of researchers studying the biology of aging.

Decision letter after peer review:

Thank you for submitting your article "The biphasic and age-dependent impact of Klotho on hallmarks of aging and skeletal muscle function" for consideration by eLife. Your article has been reviewed by 4 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Jessica Tyler as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Daniel Remondini (Reviewer #3).

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

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

The manuscript by Pius et al. seeks to comprehensively characterize the trajectory of murine muscle aging and to correlate that to the ability of Klotho to exert beneficial effects on muscle health. The authors present a comprehensive profiling of skeletal muscle aging at the histological, functional, and transcriptomic levels across the lifespan (young, middle, old, oldest old age) and network entropy as a measure of molecular dysfunction during aging. The authors identify a network entropy inflection point and find it correlate with the ability of Klotho delivery to attenuate age-related changes. Interestingly, network entropy does not further increase from old to oldest old and Klotho in oldest old fails to exert a benefit at that age. The study presents a large amount of work, particularly in compiling the various datasets and developing a novel entropy-based modeling to define the trajectory of murine muscle aging, which will certainly be of interest to the community of researchers studying the biology of aging. All four reviewers agreed that the study provides interesting insights into aging. However, as detailed below, various concerns were also raised, relating in large part to insufficient clarity in the current version with respect to the authors' methods and the limitations of their approach and data.

Essential revisions:

1. As the transcriptomes and network entropy analysis are the cornerstone of the authors' approach, it is essential that they provide clear and complete details regarding the experimental strategies and rationale underlying each analysis. In particular, detailed answers to the following key questions must be available in the present manuscript:

2. There were mixed reviews on the validity and advantage of the entropy network method: two positive and two negative reviews. The entropy network model was built on protein-protein interaction (PPI) networks. Given that the authors used transcriptomes as input data, provide the validity of the methodology and discuss the limitation of the approach. For example, numerous reports show that the lack of positive linear correlation between protein levels and the amount of mRNA, e.g. Cell 165: 535-550; Molecular and Cellular Proteomics 1: 304-313, Scientific Reports Article number: 3272.

3. The authors should provide justification as to why the entropy change from 522 to 528 (only 1.1%, p=0.07) in Figure 3D is deemed meaningful and important.

4. pg 7 line 123-130 and Figure 3. From the figure it emerges that the first component of PCA seems to characterize aging progression: which are the genes mostly associated to this component? Further analysis (e.g. gene selection based on loadings, possibly after Factor Analysis) could help to emphasize these aspects and identify the key genes associated to aging progression. Moreover, if biologically relevant, these gene set could be characterized throughout the study in relation to aging progression and to the effects of Klotho. Further: did authors try other dimensionality reduction methods such as t-SNE, that in principle could identify other peculiar aspects within the data?

5. Methods section: more details about the analyzed network are required, e.g. which is the size of the networks used to estimate network entropy? Which are the main global features (e.g. link density, presence of disconnected components,.…). Moreover, there appears to be more than one network that has been analyzed (genes associated to hallmarks of aging in line 146, global PPI network on line 155). A clear description is needed, in the Methods or in Supplementary material. Further: in Figure1D and supplementary Figure S1 the Network entropy values appear with a "confidence interval" (in light blue). In the paper there are no references about how this uncertainty was estimated: is it standard deviation or standard deviation of mean? It is a 95% confidence interval or something else? Also for this point a clear description is requested.

6. In Figure 3, the gene perturbation data presentation is a bit odd; what would be much more helpful would be heat maps showing the progression of gene changes from young to old to oldest-old, and which of these are counter-regulated by Klotho over-expression, in particular, for the pathways shown to be linearly regulated with age, but counter-regulated by Klotho in the old. Figure 3G is redundant. Remove 3G and instead indicate the DEG numbers in each pie chart in Figure 3F. Figure 6C is also redundant. As in the Figure 3D, 3E and 3F, in Supplemental Figure 2B Old vs Young should be compared to Oldest-old vs Young age group, instead of comparing Oldest-old vs Old. For oldest-old it would be interesting to see what Klotho does. The authors get at this in figure 6D but it would important to show the actual degree of perturbation connected to the pathways shown.

7. It seems like the oldest-old Klotho vs oldest-old pathways go in the opposite direct as Old Klotho vs Old. Is that true of every pathway? it would be helpful to connect the gene expression changes to a mechanism.

8. There is a substantial literature on the skeletal muscle transcriptional studies throughout the lifespan. An example in mice (https://doi.org/10.1016/j.biocel.2014.04.025), a very recent example in rats (https://doi.org/10.1016/j.celrep.2019.08.043) and in humans (https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.0020115). Refer to the previous work, which are missing in the current manuscript.

9. The idea that "molecular disorder" underpins age-related muscle dysfunction is clever. Assuming the validity of the approach, an obvious question is whether the "entropy" approach used in this manuscript can be applied to the available datasets to validate the work and to compare differences across species. This translational component and general applicability is important to show and will increase the impact of this work.

10. There is no reference about repository of the RNA-seq data, which is mandatory for eLife. Otherwise authors should justify their decision about why not to make data accessible.

11. Would it be possible to generate an interactive online website tool to probe this dataset (and perhaps the others above), which includes the "entropy" analysis of molecular disorderliness performed here? This could be very useful to the scientific community and reduces the concern that the authors are under-utilizing their novel approach to gene expression analysis.

12. The baseline characterizations of aging are well-done and consistent with prior studies. Likewise, the experiments involving the over-expression of Klotho do not appear novel. Detailed answers to the following key questions must be available in the present manuscript to provide a novelty of the study:

13. The progressive underexpression of Klotho (as shown in Figure 4A) is much smoother than the growth in FGF23 (Figure 4B) that in the paper is claimed to be the major interactor of Klotho: is there any explanation for that? Maybe the group of direct (or second neighbor) interactors of Klotho (and of FGF23 if it is relevant) should be studied specifically, to characterize the surrounding environment of these genes and provide a deeper view of the possible processes associated to aging and to Klotho response. Moreover, it is shown a large difference between growth of FGF23 in oldest-old female as compared to male of the same age group (in female mice the growth starts earlier): can the authors try to explain why? (Maybe FGF23 interactome could provide further insight?)

14. More information on the effectiveness of the AAV is needed. What organs are being affected by this? How much over-expression of Klotho is there with AAV-Kl, and specifically how much in muscle? Explain why the AAV approach was used instead of Klotho protein supplementation, which seems to have been effective in the past and is a safer and more translatable approach.

15. The effects of Klotho are interesting; what would have been helpful would be some pharmacodynamic/pathway studies. Are the effects seen happening as a result of increased FGFR signaling? FGF23 levels go up with age, while Klotho levels go down. This may be a reaction to low FGFR signaling. When Klotho is given back, which are the key pathways that are noted to change? In other words, it would be helpful to connect the gene expression changes to a mechanism.

16. As a comment in the final discussion, since the results about mice treated with Klotho are stated as general, but transcriptomics profiling was performed only on female samples, it should be emphasized that these results could not be generalized independently of sex, since many differences have been observed between the two genders.

17. Following relevant issues emerge that require further analysis, that should not alter the overall results of the paper but in my opinion could reinforce them. Detailed answers to the following key issues must be provided, including justification as to why suggested experiments cannot be done:

18. Page 10 line 215; AAV-KL treatment rather decreased force in oldest-old group (Figure 5F and 5G). Given that Klotho over-expression in the oldest-old is perhaps detrimental, experiments should be undertaken to explore how increasing Klotho earlier in life influences regenerative and transcriptional responses later in life. If the positive attributes of Klotho, when applied to an "unhealthy" oldest-old environment, then become a negative stress, it is an idea that should be explored further. It would also be welcome to see how Klotho influences more translatable outcomes such as adaptability to exercise training, since the link between regenerative potential and age-associated sarcopenia is unclear.

19. For the mouse muscle IHC phenotyping, it is warranted to show other measures such as fiber type-specific CSA, quantification of ECM (via Sirius Red or Masson's), markers of denervation, as well as intramuscular adipose tissue infiltration (IMAT). Extending the analysis to a more oxidative/slow-twitch muscle such as the soleus would also be welcome, given that the TA mostly contains muscle fibers that aren't present in humans (Type IIB).

20. Further characterization of the AAV-Klotho regeneration experiments is warranted. Why is there less fibrosis in AAV-Kl? Are there fewer FAP/fibrogenic cells? Are there more satellite cells? Presenting collagen 4 intensity (Figure 4H) seems unorthodox. Please show Sirius Red or Masson's and normalized to area. In the aged AAV-Kl experiments, why is there less lipid accumulation, which appears to be specifically within (and not between) the muscle fibers? Was lipid accumulation between muscle fibers (IMAT) different? An Oil Red O analysis is warranted.

21. Serum levels of cholesterol and insulin of young mice (3-6 month) is necessary in Figure 4D to compare with those of old mice overexpressing Klotho.

22. As previously reported by the authors in Nature communication (2018), it would be interesting to compare the Klotho's function in the mitochondria of muscle progenitor cell, with the Klotho's function here identified.

23. To verify the author's suggestion in Discussion section, it is important to examine whether F2 or Kng2 plays a bifurcation role in old and oldest muscles.

24. The key experiment that seems to be missing is the manipulation of Klotho earlier in life to try and prevent the onset of "molecular disorder" later in life, especially since acute Klotho over-expression in the oldest old seemed to have the opposite effect.

25. A summary figure that highlight the main findings of this manuscript would be a welcome addition.

26. Abstract has no conclusions and does not properly reflect contents of the results.

https://doi.org/10.7554/eLife.61138.sa1

Author response

Essential revisions:

1. As the transcriptomes and network entropy analysis are the cornerstone of the authors' approach, it is essential that they provide clear and complete details regarding the experimental strategies and rationale underlying each analysis. In particular, detailed answers to the following key questions must be available in the present manuscript:

2. There were mixed reviews on the validity and advantage of the entropy network method: two positive and two negative reviews. The entropy network model was built on protein-protein interaction (PPI) networks. Given that the authors used transcriptomes as input data, provide the validity of the methodology and discuss the limitation of the approach. For example, numerous reports show that the lack of positive linear correlation between protein levels and the amount of mRNA, e.g. Cell 165: 535-550; Molecular and Cellular Proteomics 1: 304-313, Scientific Reports Article number: 3272.

This is a valuable discussion point, and we are grateful for the opportunity to clarify the use of the PPI in our paper. We agree that the nonlinear relationship between gene expression and protein levels precludes one-to-one inferences regarding proteomic muscle changes over time. However, our goal was not to estimate protein levels from transcriptomic data. Instead, we use the PPI network as a means to quantify changes in the interactome of age-related genes. To do this, we map skeletal muscle RNA-seq data using experimentally-validated physical interactions. The advantage of this approach is that it provides a proxy to visualize the functional relationship between genes. On the other hand, edges in the gene co-expression network are based on solely correlation cutoffs, and, hence, may not necessarily represent meaningful interactions. As a result, we did not view the use of a gene co-expression network as an appropriate approach. Previous reports have similarly used the PPI interactome to elucidate how shifts in gene expression network properties may play a role in the onset of complex diseases (Banerji, Miranda-Saavedra et al. 2013, Wu, Xie et al. 2014, Yang, Zhang et al. 2016, Teschendorff and Enver 2017, Fan, Lin et al. 2018).

Despite the advantage of a PPI approach, we acknowledge limitations. Namely, in addition to the fact that mapping of genes to proteins is not one-to-one, there are also a number of proteins for which PPI information has not been documented. We have highlighted both the rationale of our approach as well as its limitations in the Discussion (lines 343-349).

3. The authors should provide justification as to why the entropy change from 522 to 528 (only 1.1%, p=0.07) in Figure 3D is deemed meaningful and important.

This is a particularly intriguing question considering others have similarly observed that entropy changes in the context of stem cell differentiation, cancer, and aging are quite subtle (Park, Lim et al. 2016, Teschendorff and Enver 2017, Kannan, Farid et al. 2020).

From a mathematical perspective, it is not surprising that the change in entropy over time is relatively small given that the spatial network entropy represents the total number of possible configurations (i.e., flexibility of the system), which has a magnitude on the order of 106. The entropy of the network is a reflection of the network architecture, which is dependent on the number of nodes, the number of edges, and the edge weights (i.e., the difference in gene expression). Since fundamental aspects of the network architecture, including the number of nodes and the number of edges, remains relatively stable over time, the change in the magnitude of network entropy represents only a small fraction of all possible configurations (~1.1%). However, this small percentage captures the change in expression of a large number of genes (~7600 genes), each of which have the potential to elicit a cascade of downstream phenotypic changes.

From a physiological perspective, there is growing evidence that very small-scale mRNA changes can lead to extreme changes in protein expression and cellular phenotype (Ruzycki, Tran et al. 2015, Cheng, Teo et al. 2016). In the context of cancer, for example, a single base mutation can drive extensive cellular remodeling that extends well beyond the predicted downstream responses, a so-called ‘butterfly effect’ (Hart, Zhang et al. 2015). Others have similarly shown that cells are highly sensitive to initial conditions (in our case, gene expression), and that very small changes in transcriptional regulation and equilibrium may exert potent downstream impacts on disease progression (Dorn 2013, Desi and Tay 2019). We have revised the manuscript to highlight these important points (lines 353-368).

4. pg 7 line 123-130 and Figure 3. From the figure it emerges that the first component of PCA seems to characterize aging progression: which are the genes mostly associated to this component? Further analysis (e.g. gene selection based on loadings, possibly after Factor Analysis) could help to emphasize these aspects and identify the key genes associated to aging progression. Moreover, if biologically relevant, these gene set could be characterized throughout the study in relation to aging progression and to the effects of Klotho.

Great suggestion. To address this, first, we evaluated the loading of genes for PC1 and PC2. Interestingly, functional enrichment of top 25 genes associated with each component revealed that 9/25 genes (36%) in PC1, and 17/25 genes (68%) in PC2 are associated with hallmarks of aging. This is consistent with the fact that PC2 best distinguishes age groups (Figure 3B).

We then focused on the top 100 PC2 genes and performed functional gene ontology enrichment analysis (presented in Figure 3—figure supplement 1A). These GO terms were predominantly associated with altered intercellular communication and nutrient sensing deregulation, two of the hallmarks of aging.

Next, we visualized the top 100 genes from PC2 loadings as a heatmap. The heatmap showing all gene trends is presented in Figure 3—figure supplement 1B. Fifteen genes have an increasing trend with aging progression, and 23 genes had a decreasing trend. A heatmap focusing only on these 38 up- and down-regulated genes is presented in Figure 3C. Nutrient sensing deregulation and altered intercellular communication were the two hallmarks most highly associated with the 38 genes displaying an aging progression trend. We have updated the Results sections to present these findings (lines 138-143).

Interestingly, the two hallmarks that emerge different with PC2 genes: altered intercellular communication and nutrient-sensing deregulation also relates to FGF23 signaling that is a known interactor of Klotho. We observed a novel age-dependent effect on FGF23 signaling with Klotho treatment. We have discussed this effect in comments 13 and 15 in greater detail.

Further: did authors try other dimensionality reduction methods such as t-SNE, that in principle could identify other peculiar aspects within the data?

As suggested, we performed t-SNE to complement the PCA presented in Figure 3B. We were encouraged to see a similar clustering pattern for the aging samples (Author response image 1). Specifically, as was seen with the PCA, old and oldest-old displayed some degree of overlap with each other but were distinct from young muscles. Because the information gained by the two-dimensional reduction approaches were similar, we did not include the t-SNE map in the revised paper.

Author response image 1

5. Methods section: more details about the analyzed network are required, e.g. which is the size of the networks used to estimate network entropy? Which are the main global features (e.g. link density, presence of disconnected components,.…). Moreover, there appears to be more than one network that has been analyzed (genes associated to hallmarks of aging in line 146, global PPI network on line 155). A clear description is needed, in the Methods or in Supplementary material. Further: in Figure1D and supplementary Figure S1 the Network entropy values appear with a "confidence interval" (in light blue). In the paper there are no references about how this uncertainty was estimated: is it standard deviation or standard deviation of mean? It is a 95% confidence interval or something else? Also for this point a clear description is requested.

We have now added Supplementary file 2 that describes the number of nodes, number of edges, number of connected components, link density, and spatial network entropy before normalization. The table 2 within the file is divided into several parts: (A) describes the network features for hallmark of aging associated genes presented in the main figure (Figure 3E), and (B) describes network features for Figure 3—figure supplement 2, which is calculated from all genes. Clarification about this in the main text is added (lines 666-670).

The figure legends have been revised to denote that light blue shaded portion represents the standard deviation.

Supplementary file 2 also has features for networks that were generated for male mice, rats and humans (2C, 2D and 2E respectively), later described in response to comment 9.

6. In Figure 3, the gene perturbation data presentation is a bit odd; what would be much more helpful would be heat maps showing the progression of gene changes from young to old to oldest-old, and which of these are counter-regulated by Klotho over-expression, in particular, for the pathways shown to be linearly regulated with age, but counter-regulated by Klotho in the old. Figure 3G is redundant. Remove 3G and instead indicate the DEG numbers in each pie chart in Figure 3F. Figure 6C is also redundant. As in the Figure 3D, 3E and 3F, in Supplemental Figure 2B Old vs Young should be compared to Oldest-old vs Young age group, instead of comparing Oldest-old vs Old.

We characterized aging progression genes by extracting genes with highest PC2 loadings and displayed this as a heatmap (described above in response to comment 4 and Figure 3—figure supplement 1). We did not see any striking trend in the aging progression genes with Klotho over-expression. However, as stated above, we observed an age-dependent effect on FGF23 signaling with Klotho treatment. We have discussed this effect in greater detail comments 13 and 15. We have suggested future studies to explore the relationship (Discussion, lines 400-412).

As per the reviewer’s suggestion, we have expressed the DE gene count on the network plot (current Figures 3G, 7B). We have removed the redundant figures (previously Figures 3F, and 6C), and the revised figures have been added as Figures 3G, and 7B.

Figure 3—figure supplement 3B (previously Supplemental figure 2B) has now been changed to show Old vs Young and Oldest-old vs Young.

For oldest-old it would be interesting to see what Klotho does. The authors get at this in figure 6D but it would important to show the actual degree of perturbation connected to the pathways shown.

We apologize for not being clear in the text and figure legend. Figure 7D (previously, Figure 6D) shows gene perturbation as total accumulation in the bar graph. The top pathways are perturbed in opposite directions in old and oldest-old. We have clarified this in the figure legend (line 232).

7. It seems like the oldest-old Klotho vs oldest-old pathways go in the opposite direct as Old Klotho vs Old. Is that true of every pathway? it would be helpful to connect the gene expression changes to a mechanism.

In the previous Figure 6C (current Figure 7C), we originally highlighted only those genes that displayed opposing responses to Klotho intervention when comparing old and oldest-old groups. However, not every pathway goes in the opposite direction. We have now expanded Supplementary file 1 (previously, supplemental figure 5) to present every pathway.

In order to probe potential mechanisms emerging from alterations in gene expression across the groups, we first performed gene ontology enrichment for the differentially expressed genes obtained after Klotho treatment (Figure 7A). Interestingly, we found that FGF/ion transport/calcineurin/NFAT signaling were amongst the top GO terms for old vs old klotho, but not oldest-old vs oldest-old klotho (Figure 7C, Table 1). We describe this age-dependent mechanism in greater detail in response to comments 13 and 15.

We next evaluated genes associated with the top 3 pathways differentially responsive to Klotho according to age: regulation of actin cytoskeleton, cGMP-PKG signaling pathway, and sphingolipid signaling pathway. In old, but not oldest old, mice, the majority of genes comprising these pathways were associated with calcium ion transport and signaling. These are well-established pathways that are regulated by Klotho (Kuro-o 2019). However, the age-dependency of these pathways in response to Klotho has not been shown. Instead, we found that the major responses to Klotho in the oldest-old group related to mitochondrial and metabolism associated processes (Figure 7C, Table 2). We present the top 20 functional gene ontology terms in Figure 7—figure supplement 2. We have also revised the Discussion accordingly (lines 400-412).

8. There is a substantial literature on the skeletal muscle transcriptional studies throughout the lifespan. An example in mice (https://doi.org/10.1016/j.biocel.2014.04.025), a very recent example in rats (https://doi.org/10.1016/j.celrep.2019.08.043) and in humans (https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.0020115). Refer to the previous work, which are missing in the current manuscript.

These studies are now cited in the Discussion (lines 337-340).

9. The idea that "molecular disorder" underpins age-related muscle dysfunction is clever. Assuming the validity of the approach, an obvious question is whether the "entropy" approach used in this manuscript can be applied to the available datasets to validate the work and to compare differences across species. This translational component and general applicability is important to show and will increase the impact of this work.

This is a great idea. To compare network entropy changes over time, as suggested, we accessed archived transcriptomic datasets from male mice, as well as rats and humans.

The male mice transcriptomic data was obtained from Tabula Muris Senis limb muscle (Schaum, Lehallier et al. 2020). We excluded females due to insufficient sample size. Age grouping was done as in our manuscript, with 3-6 months as young, 12-15 months as middle-aged, 21-24 months as old, and 27 months as oldest-old. The spatial entropy obtained from limb muscle sequencing is presented in Figure 3—figure supplement 4A. Network properties are presented in Supplementary file 2 (Table 2C). The blue shaded region denotes standard deviation (n=6, 7, 6, and 4 respectively). Unlike females, we saw that entropy continued to increase from old to oldest-old timepoints, which we found interesting considering that male mice showed less severe sarcopenic declines at the oldest-old age. Also interesting, entropy was lowest at the middle-aged timepoint, then gradually increased.

The rat spatial entropy is obtained from the dataset provided in comment 8, gastrocnemius sequenced from aging male rats ranging from young to middle-aged to old. The blue shaded region denotes standard deviation (n=7, 7, 8, 7, 7, 8, and 8 respectively) (Figure 3—figure supplement 4B). (Shavlakadze, Morris et al. 2019). Network properties are presented in Supplementary file 2 (Table 2D). These youngest and oldest age groups are roughly equivalent to 20, and 55 years in humans, respectively (Tidball et al., Experimental Gerontology, 2021). RNA-seq data was not available at the oldest-old timepoint (the average lifespan of Rattus Norvegicus (Sprague Dawley rat) used in this study is close to three years). Since we had more time points in rats, we observed entropy decreased from young to middle-age, after which time it increased into old age, consistent with our mice data.

For humans, the dataset referenced in comment 8 was not available, and the corresponding author of the study did not respond to our request for the data . As an alternative, we used RNA-seq data from a recent unpublished study on aging skeletal muscle obtained from vastus lateralis (GSE164471) (Tumasian RA 2021). Similar to the mice and rat trajectory, in humans, entropy decreased from young adulthood into the fourth decade of life, after which time it increased, reaching a tipping point in the sixth decade of life, then decreasing slightly. This is consistent with studies that show that sarcopenic declines begin in the fourth decade of life (Walston 2012, Yazar and Olgun Yazar 2019). PPI network entropy obtained from all genes in human males with ages ranging from young to middle-aged to old to oldest-old, grouped as decades, namely, 20-29 years, 30-39 years, and so on till >80 years old. We did not present females because of insufficient sample size. The blue shaded region denotes standard deviation (n=4, 7, 4, 5, 4, 7, and 4 respectively) (Figure 3—figure supplement 4C). Network properties are presented in Supplementary file 2 (Table 2E). Entropy trend is generally consistent with our mouse data. We have described this is Results (lines 176-187).

10. There is no reference about repository of the RNA-seq data, which is mandatory for eLife. Otherwise authors should justify their decision about why not to make data accessible.

We apologize for the oversight. The GEO database ID, GSE156343, has been added to the methods sections (line 607).

11. Would it be possible to generate an interactive online website tool to probe this dataset (and perhaps the others above), which includes the "entropy" analysis of molecular disorderliness performed here? This could be very useful to the scientific community and reduces the concern that the authors are under-utilizing their novel approach to gene expression analysis.

As recommended, we have created a web-platform that allows for easy estimation of network entropy using both transcriptomic and proteomic datasets: https://network-entropy-calculator.herokuapp.com/. The steps for preprocessing have been described in (https://github.com/sruthi-hub/sarcopenia-network-entropy). We added this link to the main text (lines 185-187 and Key Resources Table).

12. The baseline characterizations of aging are well-done and consistent with prior studies. Likewise, the experiments involving the over-expression of Klotho do not appear novel. Detailed answers to the following key questions must be available in the present manuscript to provide a novelty of the study:

We agree that the characterization of murine muscle aging is not the most novel aspect of our paper. Still, we felt that it was important to first thoroughly characterize age- and sex-related changes in muscle phenotype and transcriptomic profile over time ranging from young to very old age (a characterization we could not find in the literature) so as to provide a solid foundation for the experimental studies that followed.

Along these lines, to the best of our knowledge, no studies to date have demonstrated the age-dependent ability of Klotho overexpression to attenuate sarcopenia, nor has previous work identified the biphasic gene expression profiles that accompany these disparate responses to Klotho intervention when comparing old and oldest-old mice. It is our goal that this information may provide insights into the potential therapeutic window of Klotho overexpression in the treatment of sarcopenia as well as potential molecular mechanisms that may underlie anabolic resistance to therapeutic intervention in older individuals.

13. The progressive underexpression of Klotho (as shown in Figure 4A) is much smoother than the growth in FGF23 (Figure 4B) that in the paper is claimed to be the major interactor of Klotho: is there any explanation for that? Moreover, it is shown a large difference between growth of FGF23 in oldest-old female as compared to male of the same age group (in female mice the growth starts earlier): can the authors try to explain why? (Maybe FGF23 interactome could provide further insight?)

To address this question, first, we increased the sample size used for the FGF23 analyses. Re-running the analyses with additional samples further confirmed the trend initially observed (Figure 4B).

As pointed out by the reviewer, the increase in FGF23 is not as smooth as the decline in Klotho over time, a disconnect that is driven primarily by the female data. Males, on the other hand, display a relatively steady increase with age that coincides with steadily decreasing Klotho levels. We note that approximately 40% of the female population express exceedingly high levels of FGF23 at oldest-old age, whereas the remaining mice display a more linear increases in FGF23 compared to the other age groups. We posit that this discrepancy may be a result of varied hormonal status in aged female mice. Indeed, it has been reported that 25-40% female mice undergo menopause in older age, but shortly after, spontaneously regenerate their follicles, making them hormonally more similar to pre-menopausal mice (Review, Brinton RD, Endocrinology 2012). Since estrogen levels directly regulate circulating FGF-23 in both humans and rodents (Ix et al., Am J Kidney Dis 2011; Saki et al., Gynecologic Endocrinology and Reproductive Medicine 2020), the variability in FGF-23 levels may be a result of some of the mice having a menopausal hormonal profile while others do not. Unfortunately, we do not have serum from these animals in order to measure estrogen levels. Instead, we have raised this important point in the text (Results, lines 202-207).

However, this does not fully clarify possible relationships/interactions between Klotho and FGF23. First, there is no linear interaction in transcriptional regulation of the two genes, as they are on separate chromosomes and are predominantly expressed in different tissues. However, FGF23 protein will increase in response to loss of Klotho or transgenic global overexpression of Klotho (Xiao et al., JCI Insight 2019). The important interaction between Klotho and FGF23 is at the protein level where they act as co-ligands for FGFR1 (Chen, Liu et al. 2018). At the same time, Klotho reduces FGF23 interactions with FGFR4 (Han et al., J Mol Cell Cardiol 2020, Xiao et al., JCI Insight 2019). We discuss this interaction in the context of the observation that many of the genes in the GO terms that are decreased by AAV-Klotho in the old mice are related to inhibition of FGFR4/PLCγ/calcineurin/NFAT transcriptional program. This may mechanistically account for some, but not all, of the improvements of old muscle and reduced adiposity. On the other hand, top GO terms when comparing oldest-old and oldest-old Klotho were not related to inhibition of FGFR4/PLCγ/calcineurin/NFAT signaling. Instead, we saw no patterns in these GO terms, which is indicative of a dysregulated response Klotho intervention in the oldest-old animals. This has been added to Results (lines 285-294), and discussion (lines 400-412). These pathways also presented in Figure 7C.

Maybe the group of direct (or second neighbor) interactors of Klotho (and of FGF23 if it is relevant) should be studied specifically, to characterize the surrounding environment of these genes and provide a deeper view of the possible processes associated to aging and to Klotho response.

Primary interactors with Klotho include Fgf23 and Fgfr4, which did not display a change in mRNA levels following Klotho treatment. We also probed for effects of AAV-Klotho treatment on several interactors of FGF23 and Klotho (Fgf6, Fgfr4, Fgf22, Dmp1). Again, we observed no significant changes in the expression of these interactors (Figure 7—figure supplement 1C). This suggests that the mechanism of Klotho in modulating function in the context of sarcopenia may be independent of transcriptional regulation of FGF23.

However, when we looked at the GO terms for pathways which were differentially expressed with Klotho treatment in old vs. oldest-old age groups, the changes in old were consistent with an inhibition of non-canonical FGF23 signaling through FGFR4/calcineurin/NFAT (Han, Cai et al. 2020). Importantly, this inhibition was not observed in the oldest-old age group. These results are presented in Figure 7C.

14. More information on the effectiveness of the AAV is needed. What organs are being affected by this? How much over-expression of Klotho is there with AAV-Kl, and specifically how much in muscle? Explain why the AAV approach was used instead of Klotho protein supplementation, which seems to have been effective in the past and is a safer and more translatable approach.

The main target organs for AAV8-mediated gene delivery in mice are the liver, heart, and skeletal muscle. However, by using the liver-specific LP1 promoter, Klotho expression is limited to hepatocytes. Circulating Klotho levels of AAV-Kl treated animals were approximately 50-100-fold higher when compared to untreated or AAV-GFP treated control groups, as assessed by Klotho-specific ELISA measurements in serum samples (Figure 4E). To address the question of Klotho overexpression, we also quantified vector genomes in the liver to quantify expression and included this data in Figure 4D. Klotho expression was not expected in skeletal muscle due to the liver-specificity of the AAV construct. Nevertheless, we extracted Klotho expression in gastrocnemius from our RNA-seq data and confirmed that levels were unchanged with AAV-Kl treatment, seen in Figure 4F.

We selected an AAV approach in order to allow for robust and stable Klotho levels over a three-week time frame without the need for frequent re-administration of recombinant Klotho. Due to the observed challenges of recombinant Klotho production, which typically results in very low protein yields, thereby raising therapeutic obstacles. AAV8-based gene therapy circumvents this issue. Of note, similar approaches using various AAV capsid variants, including AAV8, are currently in advanced clinical trials in the context of hemophilia and glycogen storage diseases (George 2017, Kishnani, Sun et al. 2019), thereby underscoring the feasibility of AAV-based gene therapy approaches for the continuous expression of secreted therapeutic proteins. We have added a rationale for the use of AAV in the discussion (Page 17, lines 382-386).

15. The effects of Klotho are interesting; what would have been helpful would be some pharmacodynamic/pathway studies. Are the effects seen happening as a result of increased FGFR signaling? FGF23 levels go up with age, while Klotho levels go down. This may be a reaction to low FGFR signaling. When Klotho is given back, which are the key pathways that are noted to change? In other words, it would be helpful to connect the gene expression changes to a mechanism.

As a first step to address the potential effects of Klotho administration on FGFR activity, we conducted an ELISA for FGF23 in the serum of the 21-24-, and 27–29-month-old female mice that had been treated with AAV-Klotho or GFP, as in Figure 6. We found no significant difference in FGF23 levels with Klotho treatment versus the age matched GFP treatment group (Figure 7—figure supplement 1A).

Klotho increases canonical FGF23 signaling by increasing the affinity of FGF23 to FGFR1 by 20-fold (Ho and Bergwitz, J Mol Endocrinol 2021). At the same time, increased Klotho levels reduce interactions of FGF23 with FGFR4, which initiates the pathogenic calcineurin/NFAT transcriptional program (Ho and Bergwitz, J Mol Endocrinol 2021). Thus, the data suggest that, in old animals, the changes observed are not a result of increased FGFR signaling per se, but rather, a change in the receptor activated (FGFR4 to FGFR1) and subsequent downstream signaling by FGF23. These points are addressed in the results and in the discussion (lines 299-307, lines 400-412).

16. As a comment in the final discussion, since the results about mice treated with Klotho are stated as general, but transcriptomics profiling was performed only on female samples, it should be emphasized that these results could not be generalized independently of sex, since many differences have been observed between the two genders.

We have revised the text to highlight this limitation (lines 314-315).

17. Following relevant issues emerge that require further analysis, that should not alter the overall results of the paper but in my opinion could reinforce them. Detailed answers to the following key issues must be provided, including justification as to why suggested experiments cannot be done:

18. Page 10 line 215; AAV-KL treatment rather decreased force in oldest-old group (Figure 5F and 5G). Given that Klotho over-expression in the oldest-old is perhaps detrimental, experiments should be undertaken to explore how increasing Klotho earlier in life influences regenerative and transcriptional responses later in life. If the positive attributes of Klotho, when applied to an "unhealthy" oldest-old environment, then become a negative stress, it is an idea that should be explored further.

We agree with the reviewer that this is a very interesting line of investigation and one of potentially great clinical relevance. Our team carefully weighed the feasibility of conducting such experiments. Unfortunately, considering the time to perform dose-testing pilot studies and power analysis as well as time required to age the animals following intervention, we estimated such studies would require almost one year. As an alternative, we have added text to the Discussion highlighting the value of such work for future studies (lines 431-433).

It would also be welcome to see how Klotho influences more translatable outcomes such as adaptability to exercise training, since the link between regenerative potential and age-associated sarcopenia is unclear.

To probe the effect of Klotho on exercise response, we evaluated the effects of AAV-Klotho administration on muscle function of mice following an exercise-induced injury (Armand, Launay et al. 2003). Briefly, 22-month-old male mice were randomized into one of two groups and treated with either AAV-Klotho or AAV-GFP via tail vein injection. The AAV-Klotho dose used was the same used for the severe injury protocol. Mice were acclimated to a treadmill for 4 days, given one day of rest, then underwent an eccentric exercise protocol, which consisted of running at 10 m/min for 60 minutes at a 15-degree decline. Though we could not find published studies that utilized this protocol in aged mice, the eccentric treadmill model has been shown to induce exercise-related tibialis anterior damage in young (2 months old) mice (Parise, McKinnell et al. 2008). We ensured that all mice included in the analysis were able to complete the protocol by assessing running performance according to the method established by Ríos-Kristjánsson (Ríos-Kristjánsson, Rizo-Roca et al. 2019). Animals performing under a threshold of 0.35 were removed from analyses (n=2 from each treatment group). One week after the treadmill protocol, an investigator blinded to the treatment group performed in vivo contractile testing of the TA muscle. We repeated this study in duplicate (two cohorts of mice, n=6/group/cohort).

We found that mice treated with AAV-Klotho had a slightly greater force production than those treated with AAV-GFP, though the result was not significant (p=0.11 as determined by a linear mixed effects model and adjusting for cohort). While these results suggest a potentially positive effect on adaptability to exercise for AAV-Klotho administration, a power analysis revealed that 28 mice per group would be needed to detect statistically significant differences between groups. Given that the effect size of the AAV-Klotho intervention is small (roughly 9% force improvement), we did not test additional animals as the initial results suggest that intervention at the dose tested is not likely to be physiologically relevant. These results are described on lines 237-250 and presented in Figure 5M and Figure 5—figure supplement 1.

19. For the mouse muscle IHC phenotyping, it is warranted to show other measures such as fiber type-specific CSA, quantification of ECM (via Sirius Red or Masson's), markers of denervation, as well as intramuscular adipose tissue infiltration (IMAT). Extending the analysis to a more oxidative/slow-twitch muscle such as the soleus would also be welcome, given that the TA mostly contains muscle fibers that aren't present in humans (Type IIB).

As suggested, we performed fiber-type staining of the TA muscle for each sex across age groups (Figure 1C, D, Figure 1—figure supplement 1C-G). Unfortunately, we did not have another muscle type, such as the soleus, available for quantification of fiber-type in a more oxidative muscle, and we did not have oldest-old mice available to repeat the protocol. However, our evaluation of the TA muscle across age groups revealed a shift in fiber characteristics and, specifically, a loss of the more oxidative TA type IIA and IIX phenotypes with age (Figure 1D). These findings are consistent with Augusto et al. and Giacomello et al. (2017, 2020).

We also quantified ECM using Masson’s trichrome staining. Results are presented in Figure 1D. Consistent with Collagen IV data, there was a significant increase in collagen deposition in the 27-29-month-old age groups of both sexes.

Additionally, we quantified both inter- and intramuscular adipose tissue using lipidtox staining, though no differences were observed according to age (Figure 1F, G).

Since we did not have tissues available for whole mount staining of the neuromuscular junctions available, as an alternative, we accessed the publicly available database of skeletal muscle transcriptome Tabula Muris Senis (Schaum, Lehallier et al. 2020) and quantified the change over time using well-established denervation genes Musk, Ncam1, and Runx1 (Aare, Spendiff et al. 2016). We observed an increase with age for Musk and Runx1 These data are presented in Figure 1—figure supplement 2. We also quantified these same genes in our own AAV-Kl treated tissues, with data presented in Figure 6—figure supplement 1C-D.

20. Further characterization of the AAV-Klotho regeneration experiments is warranted. Why is there less fibrosis in AAV-Kl? Are there fewer FAP/fibrogenic cells? Are there more satellite cells?

To answer this question, we quantified and compared the ratio of fibrogenic cells (FAPs) to muscle satellite cells (MuSCs) across experimental groups. Briefly, mice were treated with AAV-GFP or AAV-Klotho, as in other experiments (n=9 per group). Five days following injection, mice received a bilateral TAs were injured via cardiotoxin injection. Three days after injury, mice were euthanized to capture the muscle at the height of its fibrotic response. TA muscles were then digested, and cells were sorted via flow cytometry. We observed a twenty-fold decrease in the ratio of FAPs:MuSCs with Klotho treatment, which we hypothesize contributes to the decreased fibrosis observed in histological sections. This data is presented in Figure 5F.

Presenting collagen 4 intensity (Figure 4H) seems unorthodox. Please show Sirius Red or Masson's and normalized to area.

As suggested, we performed Masson’s Trichrome staining to confirm Collagen IV findings (Figure 5E).

In the aged AAV-Kl experiments, why is there less lipid accumulation, which appears to be specifically within (and not between) the muscle fibers? Was lipid accumulation between muscle fibers (IMAT) different? An Oil Red O analysis is warranted.

We have now quantified inter- and intramuscular lipid accumulation for these experiments using ImageJ to distinguish IMAT lipid from intramuscular lipid (for detailed see methods lines 518-526). Intramuscular, but not intermuscular, lipid accumulation was significantly decreased in old female mice treated with AAV-Kl. Conversely, in oldest-old female mice, we found that intramuscular lipid increased with AAV-Kl treatment, while intermuscular lipid was unchanged. The results are presented in Figure 6C, G, M. The observation of increased intramuscular lipid in oldest-old mice treated with AAV-Kl is likely a reflection of myofiber metabolic changes (Johannsen, Conley et al. 2012), which is consistent with our findings in Table 2. We now address this in the discussion on lines 417-425.

We used LipidTox staining based-analysis instead of Oil Red O staining because LipidTox has been shown to be more specific to native lipid droplets in fixed tissue (Fam, Klymchenko et al. 2018). Whereas Oil Red O staining has been used as a gold standard for decades, due to its poor solubility in solutions other than ethanol or isopropanol, the abrasive staining procedure can disrupt the native lipid droplets in fixed tissues (Fam, Klymchenko et al. 2018).

21. Serum levels of cholesterol and insulin of young mice (3-6 month) is necessary in Figure 4D to compare with those of old mice overexpressing Klotho.

We now present data showing levels of cholesterol, insulin, and glucose in young mice (Figure 4G, H, I), in addition to the other metabolites presented in the paper (Figure 4—figure supplement 1).

22. As previously reported by the authors in Nature communication ( 2018), it would be interesting to compare the Klotho's function in the mitochondria of muscle progenitor cell, with the Klotho's function here identified.

To address this comment, we performed TEM on muscles across intervention groups, and mitochondrial integrity was analyzed by an investigator blinded to the treatment groups using a method as previously described (Owen, Patel et al. 2019). Our findings reveal that the percent of mitochondria with normal structure is increased in mice treated with Klotho. Results are presented in Figure 5G, H.

23. To verify the author's suggestion in Discussion section, it is important to examine whether F2 or Kng2 plays a bifurcation role in old and oldest muscles.

We have removed the statement about F2 and Kng2 in the discussion. Instead, we focus on FGF signaling as a mediator of Klotho treatment in old and oldest-old animals.

24. The key experiment that seems to be missing is the manipulation of Klotho earlier in life to try and prevent the onset of "molecular disorder" later in life, especially since acute Klotho over-expression in the oldest old seemed to have the opposite effect.

As described above, we agree that these would be very interesting studies to conduct, but the longitudinal nature of these experiments precludes their execution as a part of this current work.

25. A summary figure that highlight the main findings of this manuscript would be a welcome addition.

Thank you for this suggestion. We have added a graphical abstract summarizing the major take-home messages of the work (Figure 8).

26. Abstract has no conclusions and does not properly reflect contents of the results.

We agree that our manuscript presents an abundant data set that is not fully captured in the abstract. With the 150-word limit, we have attempted to highlight and explain the major takeaway messages, and most notably, the novel use of network entropy to characterize integrative transcriptomic changes over time as well as the dependency of Klotho overexpression efficacy to reverse a sarcopenic phenotype according to age. We have revised the abstract in the hopes of making these main messages even more clear and direct.

https://doi.org/10.7554/eLife.61138.sa2

Article and author information

Author details

  1. Zachary Clemens

    1. Department of Physical Medicine & Rehabilitation, University of Pittsburgh, Pittsburgh, United States
    2. Department of Environmental and Occupational Health, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation, Formal analysis, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Sruthi Sivakumar and Abish Pius
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2230-9151
  2. Sruthi Sivakumar

    1. Department of Physical Medicine & Rehabilitation, University of Pittsburgh, Pittsburgh, United States
    2. Department of Bioengineering, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation, Formal analysis, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Zachary Clemens and Abish Pius
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0649-8551
  3. Abish Pius

    1. Department of Physical Medicine & Rehabilitation, University of Pittsburgh, Pittsburgh, United States
    2. Department of Computational & Systems Biology, School of Medicine, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation, Formal analysis, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Zachary Clemens and Sruthi Sivakumar
    Competing interests
    No competing interests declared
  4. Amrita Sahu

    Department of Physical Medicine & Rehabilitation, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation, Formal analysis, Visualization, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Sunita Shinde

    Department of Physical Medicine & Rehabilitation, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  6. Hikaru Mamiya

    Department of Bioengineering, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  7. Nathaniel Luketich

    Department of Bioengineering, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  8. Jian Cui

    Department of Computational & Systems Biology, School of Medicine, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  9. Purushottam Dixit

    Department of Physics, University of Florida, Gainesville, United States
    Contribution
    Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  10. Joerg D Hoeck

    Department of Research Beyond Borders, Regenerative Medicine, Boehringer Ingelheim Pharmaceuticals, Inc, Rhein, Germany
    Contribution
    Validation, Methodology
    Competing interests
    is an employee of Boehringer INgelheim Pharmaceutical Company
  11. Sebastian Kreuz

    Department of Research Beyond Borders, Regenerative Medicine, Boehringer Ingelheim Pharmaceuticals, Inc, Rhein, Germany
    Contribution
    Validation, Methodology
    Competing interests
    is an employee of Boehringer INgelheim Pharmaceutical Company
  12. Michael Franti

    Department of Research Beyond Borders, Regenerative Medicine, Boehringer Ingelheim Pharmaceuticals, Inc, Rhein, Germany
    Contribution
    Validation, Methodology, Writing - review and editing
    Competing interests
    is an employee of Boehringer Ingelheim Pharmaceutical Company
  13. Aaron Barchowsky

    Department of Environmental and Occupational Health, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1268-8159
  14. Fabrisia Ambrosio

    1. Department of Physical Medicine & Rehabilitation, University of Pittsburgh, Pittsburgh, United States
    2. Department of Environmental and Occupational Health, University of Pittsburgh, Pittsburgh, United States
    3. Department of Bioengineering, University of Pittsburgh, Pittsburgh, United States
    4. McGowan Institute for Regenerative Medicine, University of Pittsburgh, Pittsburgh, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    ambrosiof@upmc.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5497-5968

Funding

National Institute on Aging (R01AG052978)

  • Fabrisia Ambrosio

National Institute on Aging (R01AG061005)

  • Fabrisia Ambrosio

Boehringer Ingelheim

  • Fabrisia Ambrosio

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors thank Matthias Duechs (Boehringer-Ingelheim (BI)) who performed MSD-ELISA on samples obtained from the dose testing experiment in young animals. We thank Timothy Lezon, Assistant Professor, Computational and Systems Biology at the University of Pittsburgh for valuable discussions regarding Shannon entropy and network analysis. We also thank Center for Research and Computing at the University of Pittsburgh that provided resources for network entropy computations. Amanda Miller, University of Pittsburgh, contributed to histology and flow cytometry. We are grateful to the Bryan Brown Lab, University of Pittsburgh, for equipment use and technical guidance. Finally, we thank Center for Biologic Imaging, University of Pittsburgh for providing resources to perform histological IF staining. The studies reported in this manuscript are supported by funding from NIA R01AG052978 (FA), NIA R01AG061005 (FA), and BI (FA).

Ethics

Animal experimentation: All animal experiments were performed with prior approval from the Institutional Animal Care and Use Committee of the University of Pittsburgh. These experiments were conducted in accordance with protocol 17080802 (University of Pittsburgh ARO: IS00017744). All surgeries and invasive procedures were performed under isoflurane anesthesia, with painkillers administered afterwards. Every effort was made to minimize suffering.

Senior Editor

  1. Jessica K Tyler, Weill Cornell Medicine, United States

Reviewing Editor

  1. Yousin Suh, Columbia University, United States

Reviewer

  1. Daniel Remondini

Publication history

  1. Received: July 16, 2020
  2. Accepted: April 6, 2021
  3. Accepted Manuscript published: April 20, 2021 (version 1)
  4. Version of Record published: May 13, 2021 (version 2)

Copyright

© 2021, Clemens et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,041
    Page views
  • 266
    Downloads
  • 0
    Citations

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

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cancer Biology
    2. Computational and Systems Biology
    Ioannis Galdadas et al.
    Research Article

    Mutations within the kinase domain of the epidermal growth factor receptor (EGFR) are common oncogenic driver events in non-small cell lung cancer. Although the activation of EGFR in normal cells is primarily driven by growth-factor-binding-induced dimerization, mutations on different exons of the kinase domain of the receptor have been found to affect the equilibrium between its active and inactive conformations giving rise to growth-factor-independent kinase activation. Using molecular dynamics simulations combined with enhanced sampling techniques, we compare here the conformational landscape of the monomers and homodimers of the wild-type and mutated forms of EGFR ΔELREA and L858R, as well as of two exon 20 insertions, D770-N771insNPG, and A763-Y764insFQEA. The differences in the conformational energy landscapes are consistent with multiple mechanisms of action including the regulation of the hinge motion, the stabilization of the dimeric interface, and local unfolding transitions. Overall, a combination of different effects is caused by the mutations and leads to the observed aberrant signaling.

    1. Chromosomes and Gene Expression
    2. Computational and Systems Biology
    Kazuhiro Maeshima, Shiori Iida
    Insight

    DNA loops can be formed by a mechanism in which the cohesin complex pulls DNA strands through its ring structure using biased Brownian motion.