Abstract

   Aging is associated with an increased risk of frailty, disability, and
   mortality. Strategies to delay the degenerative changes associated with
   aging and frailty are particularly interesting. We treated old animals
   with small extracellular vesicles (sEVs) derived from adipose
   mesenchymal stem cells (ADSCs) of young animals, and we found an
   improvement in several parameters usually altered with aging, such as
   motor coordination, grip strength, fatigue resistance, fur
   regeneration, and renal function, as well as an important decrease in
   frailty. ADSC-sEVs induced proregenerative effects and a decrease in
   oxidative stress, inflammation, and senescence markers in muscle and
   kidney. Moreover, predicted epigenetic age was lower in tissues of old
   mice treated with ADSC-sEVs and their metabolome changed to a
   youth-like pattern. Last, we gained some insight into the microRNAs
   contained in sEVs that might be responsible for the observed effects.
   We propose that young sEV treatment can promote healthy aging.
     __________________________________________________________________

   Restoring defective intercellular communication with young sEVs
   reverses aging hallmarks and improves fitness in old mice.

INTRODUCTION

   To add health to the years gained, as well as to promote the ability to
   live autonomously, is a public health priority in modern societies. The
   search for strategies to delay the degenerative changes associated with
   aging and frailty is particularly interesting.

   Aging is accompanied by an impairment in the physical condition and an
   increased risk of frailty ([54]1). It is characterized by several
   changes at the cellular and organismal levels, such as oxidative
   stress, telomere attrition, or cellular senescence ([55]2), that result
   in a decreased functionality of tissues. These changes are proposed as
   the main factors that drive age-related conditions, such as sarcopenia
   or renal failure ([56]3, [57]4). One of the traits associated with
   aging is the alteration in intercellular communication. As an
   individual ages, the extracellular environment changes, with
   deregulation of important signaling pathways [i.e.,
   insulin–insulin-like growth factor 1 (IGF1) or renin-angiotensin]
   ([58]5) and a chronic proinflammatory status ([59]6, [60]7). Therefore,
   cellular function not only depends on cell-autonomous factors but also
   is affected by the extracellular environment, and its modification can
   have a great effect on the performance of several tissues ([61]5,
   [62]8).

   Parabiosis experiments conducted in mice demonstrated that factors
   present in the blood of a young organism are beneficial for an aged
   one, improving several parameters affected by aging ([63]9–[64]11).
   Moreover, interventions that delay aging on a specific tissue can
   induce similar changes in distant tissues ([65]12–[66]14).

   Extracellular vesicles (EVs), small vesicles that are released by
   virtually all cell types, with an innate ability to mediate the
   transmission of signaling molecules (proteins, small RNAs, and DNA)
   between cells are among the factors that are involved in the
   communication between cells ([67]15).

   Stem cells have intrinsic regenerative effects that are not only
   mediated by the repopulation of damaged tissue. The releasing of
   regulatory molecules is also proposed as one of the most important
   mechanisms in stem cell therapies ([68]16, [69]17). More specifically,
   small EVs (sEVs) derived from multiple stem cells have demonstrated
   their capacity to promote tissue regeneration after several types of
   damage ([70]18, [71]19). Compared to stem cells, sEVs are more stable,
   have no risk of aneuploidy, have a lower chance of immune rejection,
   and can provide an alternative therapy for various diseases
   ([72]20–[73]23).

   It is important to consider the way culturing stem cells affects
   markedly paracrine signals. We previously showed that oxygen tension in
   culture is one of the most important factors affecting the biological
   function of sEVs released by these cells ([74]24–[75]26) and that sEVs
   derived from mesenchymal stem cells (MSCs) cultured under hypoxic
   conditions can mediate the transfer of microRNAs (miRNAs) to senescent
   cells and improve cellular function ([76]25). A recent study
   demonstrated that hypothalamic stem cells modulate aging speed through
   the release of miRNAs contained in EVs ([77]27). sEVs are emerging as a
   potential therapy in the aging field. It has been recently shown that
   sEVs can exert proregenerative effects in tissues of old mice and
   decrease senescence-related damage ([78]28–[79]31). However, the effect
   of MSC-derived sEVs on the health span of an aged organism has not been
   fully addressed. Here, we show that sEVs from young adipose-derived
   stem cells (ADSC-sEVs) improve several functions that are impaired in
   old mice. Old mice that received young ADSC-sEVs showed lower levels of
   frailty and improvements on physical condition tests, fur regeneration,
   and renal function. ADSC-sEVs induced proregenerative effects in muscle
   and kidney of aged mice, as well as a decrease in oxidative stress,
   inflammation, and senescence markers. Moreover, predicted epigenetic
   age was lower in tissues of old mice treated with ADSC-sEVs and the
   metabolome of old mice treated with ADSC-sEVs changed from an old-like
   pattern to a youth-like one. Last, we gained some insight into the
   miRNAs contained in sEVs that might be, at least in part, responsible
   for the effects observed.

RESULTS

ADSCs release sEVs to the culture medium that are delivered to the kidney and
muscle of old mice

   In the first place, we performed a characterization of the vesicles
   isolated from the cell culture medium of ADSCs, following the minimal
   information for studies of extracellular vesicles (MISEV)
   recommendations for the characterization and functional studies of sEVs
   ([80]32). For this purpose, we performed a nanoparticle tracking
   analysis (NTA) of the isolated samples from differential
   ultracentrifugation, showing a mean of 169 nm and a mode of 176 nm
   ([81]Fig. 1A), concurrent with the definition of sEVs ([82]32).

Fig. 1. Characterization of sEVs from ADSC culture medium, which are
delivered to kidney and muscle of old mice.

   [83]Fig. 1.
   [84]Open in a new tab

   (A) Nanoparticle tracking analysis (NTA) of sEVs from young ADSCs. (B)
   Western blot image of staining with TSG101 and bovine serum albumin, a
   sample of purified sEVs from the cell culture medium of ADSCs, and a
   sample of FBS present in the same medium are shown. (C) Flow cytometry
   determination of CD63 presence in sEVs obtained from the cell culture
   medium of young ADSCs. (D and E) Representative image of ADSC-sEVs as
   observed by transmission electron microscopy (TEM) and CD63
   determination with immuno-gold labeling. (F) Images of kidney and
   muscle from old mice treated with PKH26-labeled sEVs from young ADSCs
   and vehicle as a control.

   Second, we confirmed the presence of TSG101, a cytosolic protein
   present in sEVs, using immunoblotting in our samples ([85]Fig. 1B). As
   we used a little amount (2%) of exosome-depleted fetal bovine serum
   (FBS) in the culturing of ADSCs, we used it as a negative control of
   the presence of TSG101. As we were isolating sEVs from the conditioned
   medium of ADSCs, FBS was the only source of potential contaminating
   proteins from plasma such as albumin, so we used it as a positive
   control in the determination of albumin as a contaminant in our samples
   ([86]Fig. 1B).

   Continuing with the characterization of sEVs, we checked for the
   expression of CD63, a widely used marker from the tetraspanin family in
   the membrane of sEVs ([87]33). Using flow cytometry, we were able to
   confirm the presence of this marker in the samples ([88]Fig. 1C).
   Second, we used transmission electron microscopy to measure the size
   and morphology of the isolated vesicles. We could observe round-shaped
   vesicles that ranged from 50 to 200 nm in diameter ([89]Fig. 1D). In
   addition, we found the presence of CD63 in the membrane of the isolated
   vesicles with immunogold labeling ([90]Fig. 1E).

   Last, to demonstrate a direct effect and the uptake of sEVs in our main
   tissues of interest, we labeled our sEVs with a lipophilic dye (PKH26)
   and injected them into the bloodstream of old mice. We then fixed the
   kidney and muscle and evidenced an uptake of the dye in these tissues
   with immunofluorescence ([91]Fig. 1F).

ADSC-sEVs improve health span and prevent frailty in old mice

   As aged organisms typically show impairment in their physical condition
   and an increased risk of frailty, we designed an experiment to test
   whether ADSC-sEVs could have a beneficial effect on the physical
   performance of old mice. Before treatment, we measured total body
   weight, grip strength, motor coordination, and resistance to fatigue in
   aged mice, which were recorded as baseline values. Twenty-four hours
   later, each mouse received an injection of 20 μg of ADSC-sEVs or
   phosphate-buffered saline (PBS) as a control in the tail vein (day 1).
   One week later (day 7), the mice received a second injection (as
   outlined in [92]Fig. 2A).

Fig. 2. ADSC-sEVs improve health span and prevent frailty in old mice.

   [93]Fig. 2.
   [94]Open in a new tab

   (A) Schematic representation of the experimental design. (B)
   Quantification of the change in total body weight; data are shown as
   the decrease in percentage from the baseline before treatment with
   sEVs/PBS. PBS: day 14, n = 12; day 30, n = 12; sEVs: day 14, n = 14;
   day 30, n = 15. (C to E) Quantification of the change in physical
   condition tests; data are shown as the change in percentage from
   baseline before treatment with sEVs/PBS. Grip strength: PBS: day 14, n
   = 11; day 30, n = 11; sEVs: day 14, n = 13; day 30, n = 11. Motor
   coordination: PBS: day 14, n = 14; day 30, n = 12; sEVs: day 14, n =
   16; day 30, n = 18. Fatigue resistance: PBS: day 14, n = 12; day 30, n
   = 11; sEVs: day 14, n = 14; day 30, n = 15. (F) Quantification of the
   number of frail mice before and 30 days after treatment with sEVs/PBS,
   determined by a score based on the clinical phenotype of frailty. PBS:
   day 0, n = 12; day 30 n = 12; sEVs: day 0, n = 15; day 30, n = 15. (G)
   Quantification of the hair regrowth capacity of dorsal skin in old
   mice. PBS, n = 7; sEVs, n = 7. a.u., arbitrary units. (H)
   Representative image of hair regeneration in both male and female mice
   before and 14 days after treatment with sEVs/PBS. (I) Quantification of
   the changes in plasmatic urea; data are shown as the change in
   percentage from basal levels of urea in plasma before treatment with
   sEVs/PBS. PBS: day 14, n = 5; day 30, n = 4; sEVs: day 14, n = 5; day
   30, n = 6. All data are shown as means ± SD.

   As physical tests depend on the weight of mice, we measured the total
   body weight of mice during the experiment, and we did not find
   differences between groups. In addition, we failed to observe any signs
   of toxicity (excessive weight loss) during the entire experiment
   ([95]Fig. 2B). Physical tests were repeated on days 14 and 30 after
   ADSC-sEV or PBS treatment. On day 14, we could observe an improvement
   in the strength test on ADSC-sEV–treated mice. The maximum benefit was
   observed on day 30: Mice treated with ADSC-sEVs showed an improvement
   in grip strength, motor coordination, and fatigue resistance when
   compared to controls ([96]Fig. 2, C to E). Sixty days after treatment,
   the positive effect was lost, and we could not find any differences
   between groups (table S1). Thus, the protective effects of ADSC-sEVs
   are transitory.

   For a quantitative assessment of frailty, we used a score on the basis
   of the clinical phenotype of frail humans developed by our group
   ([97]34). This enabled us to classify each mouse as frail or nonfrail
   at each time point. On day 30, there were no frail mice in the
   ADSC-sEV–treated group, while the control group shed 40% of frail mice,
   which is in accordance with previous results in mice of this age
   ([98]Fig. 2F) ([99]34).

   Fur density is reduced, and the capacity to regrowth hair becomes
   impaired with aging ([100]35). To check the effect of ADSC-sEVs on the
   growth of hair, we plucked a square of 1 cm by 1 cm of the dorsal fur
   of each mouse on day 1 just before injection. On day 14, we observed
   that most mice treated with ADSC-sEVs had regenerated the entire area.
   In contrast, mice injected with PBS had a much lower capacity for hair
   regrowth ([101]Fig. 2, G and H).

   To determine possible changes in the renal function, we used plasma
   values of urea, as it is a well-known marker of kidney function. We
   obtained plasma samples of each mouse on days 0, 14, 30, and 60 and
   measured urea levels at each time point. Day 0 levels were used as a
   baseline. Fourteen days after being injected with ADSC-sEVs, mice
   showed a decrease in plasma concentration of urea, whereas the control
   group showed no modification. On day 30, urea levels raised 50% from
   baseline in the control group. By contrast, the ADSC-sEV group showed a
   plasmatic urea level like day 0 ([102]Fig. 2I).

   In addition, as a control to further test the possible effect of age on
   the quality of sEVs and ADSCs, we performed a pilot study with old mice
   treated with ADSC-sEVs isolated from old mice, we did not find any
   effect of these old-sEVs on physical performance tests, indicating that
   the age of ADSC donor is essential in the beneficial effects observed
   (table S2). Moreover, we did not find any effect of the non-EVs
   fraction on the physical performance of old mice (table S3). To explore
   the role of ADSC-sEVs on mice of other age groups, we treated young
   mice (4 to 6 months) with sEVs from young ADSCs. We did not find a
   beneficial effect of ADSC-sEVs in the physical performance of young
   mice (table S4), suggesting that sEVs are specifically acting on
   age-related features. Together, we find that ADSC-sEVs isolated from
   young mice induce an improvement in the global health span of aged
   mice.

ADSC-sEVs reverse age-related structural changes in the kidney and muscle of
old mice

   As an organism begins to age, its tissues suffer from several
   structural changes, which are detrimental to the normal function of the
   tissue, such as a loss of regenerative capacity and fibrosis ([103]2,
   [104]36). We selected kidney and muscle to test the effect of ADSC-sEVs
   on tissue structure, as these organs are especially involved in the
   aging process: Kidney function is one of the functions that age faster
   in normal individuals ([105]37), and muscle function is essential for
   the activities of daily life and is directly related to frailty
   ([106]38). Therefore, we obtained kidney and muscle from old mice 30
   days after treatment with ADSC-sEVs/PBS. In the kidney, we performed
   histological analysis of the renal cortex, looking for the presence of
   tubular atrophy and interstitial fibrosis, two distinctive changes
   associated with aging that leads to the loss of renal function
   ([107]4). The macroscopic appearance of kidneys was very different
   between old untreated and treated mice; kidneys isolated from old mice
   treated with sEVs were more like those isolated from young mice
   ([108]Fig. 3A). Control mice showed higher tubular atrophy than
   ADSC-sEV–treated ones, defined by a reduced tubular density and the
   presence of dilated tubules. Tubular dilation was reduced, and tubular
   density was higher in mice treated with ADSC-sEVs. We did not observe
   differences in glomerular density between groups ([109]Fig. 3, B to E).
   For the study of interstitial fibrosis, we used Sirius red staining to
   measure collagen deposition. We noted a mild decrease in the
   ADSC-sEV–treated group, although no statistically significant
   differences were observed ([110]Fig. 3, F and H).

Fig. 3. ADSC-sEVs induce proregenerative effects on kidney and muscle of old
mice.

   [111]Fig. 3.
   [112]Open in a new tab

   (A) Macroscopic image of kidneys isolated from old mice treated with
   PBS/sEVs. (B to D) Quantification of histological changes in the kidney
   (renal cortex) from old mice induced by sEVs. PBS, n = 6; sEVs, n = 6.
   (E) Hematoxylin and eosin representative images of the effects on the
   structure of the renal cortex of old mice. (F) Representative image of
   Sirius red staining of the renal cortex of old mice. (G) Representative
   immunofluorescence images showing Ki67^+ cells in the renal tubules of
   old mice. Morphology marker refers to background fluorescence to better
   depict tubular morphology. (H) Quantification of extracellular fibrosis
   by Sirius red staining. PBS, n = 6; sEVs, n = 6. (I) Quantification of
   Ki67^+ cells in the renal tubules of old mice. PBS, n = 4; sEVs, n = 4.
   (J) Hematoxylin and eosin representative images of gastrocnemius of old
   mice. (K and L) Quantification of the mean cross-sectional area (CSA)
   of muscle fibers and total protein content in the gastrocnemius of old
   mice. CSA: PBS, n = 6; sEVs, n = 6. Protein content: PBS, n = 11; sEVs,
   n = 6. (M and N) Representative immunofluorescence images of MHC2A and
   LipidTOX staining in gastrocnemius of old mice. (O) Quantification of
   the percentage of MHC2A^+ fibers in gastrocnemius of old mice. PBS, n =
   6; sEVs, n = 6. (P) Quantification of LipidTOX staining intensity in
   gastrocnemius of old mice. PBS, n = 4; sEVs, n = 4. All data are shown
   as means ± SD.

   These observations suggested that ADSC-sEVs may induce the
   proliferation of tubular cells as a means of regeneration, as these
   cells can repopulate renal tubules after damage ([113]39). Using Ki67
   as a proliferation marker, we looked for the presence of Ki67^+ cells
   in the cortex tubules of our mice. We found very low levels of
   proliferating cells in the control group; however, in the treated
   group, we identified a clear population of Ki67^+ tubular cells,
   indicating active proliferation of these cells in treated mice
   ([114]Fig. 3, G and I).

   Muscle loss of function and atrophy lead to a global dysfunction of the
   organism as it ages, as the musculoskeletal system is essential for the
   activities of daily life. The prevalence of sarcopenia (defined by a
   loss of muscle mass and function) increases with age ([115]40). To
   investigate the effect of ADSC-sEVs on structural changes associated
   with aging in muscle tissue of old mice, we measured the
   cross-sectional area (CSA) of muscle fibers, which is a parameter
   related to muscle atrophy and loss of strength. We found an increase in
   the mean CSA with ADSC-sEV treatment, which was supported by a higher
   protein concentration in the muscles of ADSC-sEV–treated mice
   ([116]Fig. 3, J to L). Sarcopenia is also accompanied by other cellular
   changes such as a loss of fast type II fibers and an increase in fat
   deposition ([117]41). To test the effect of ADSC-sEVs on these
   parameters, we determined the percentage of type IIA fibers, finding a
   greater number of these fibers in old mice treated with ADSC-sEVs
   ([118]Fig. 3, M and O). Using a fluorescent marker for lipids, we
   measured the degree of intramuscular fat deposition, and mice that
   received ADSC-sEVs showed a decrease in this parameter when compared to
   control mice ([119]Fig. 3, N and P).

   In summary, these results show that ADSC-sEVs induce proregenerative
   effects in the kidney and muscle, partially reverting structural
   changes associated with aging in these tissues.

ADSC-sEVs mitigate molecular traits associated with aging in the kidney and
muscle and lower senescence in old mice and in an in vitro model

   Multiple molecular traits contribute to the aging process ([120]2).
   Here, we have studied the effect of ADSC-sEV treatment on several
   markers associated with aging: oxidative stress, senescence,
   inflammation, and telomere attrition. We obtained kidneys and muscle
   from old mice 30 days after treatment with ADSC-sEVs/PBS and studied
   how the treatment affected the presence of these markers. In the case
   of oxidative stress, we examined the levels of oxidation in two
   macromolecules, lipids, and proteins. We used malondialdehyde (MDA) as
   a marker of lipid peroxidation, and we observed lower levels of MDA in
   the muscle of mice treated with ADSC-sEVs, with no changes in the
   kidney ([121]Fig. 4A). In protein oxidation levels, determined by
   carbonyl groups present in oxidized proteins, we found similar results
   as to those of MDA, with lower oxidation in the muscle of mice treated
   with ADSC-sEVs, and no changes in the kidney ([122]Fig. 4B and fig.
   S1).

Fig. 4. sEVs from young ADSCs ameliorate molecular traits of aging in the
kidney and muscle and lower senescence in old mice and in an in vitro model.

   [123]Fig. 4.
   [124]Open in a new tab

   (A) Quantification of MDA as a marker of lipid peroxidation in the
   kidney and muscle from old mice. PBS: kidney, n = 10; muscle, n = 11;
   sEVs: kidney, n = 6; muscle, n = 6. (B) Quantification of protein
   carbonylation as a marker of protein oxidation in the kidney and muscle
   of old mice. PBS: kidney, n = 7; muscle, n = 7; sEVs: kidney, n = 6;
   muscle n = 6. (C) Mean telomere length measured as arbitrary units of
   fluorescence (A.U.F.) in the kidney and muscle of old mice. PBS:
   kidney, n = 6; muscle, n = 6; sEVs: kidney, n = 6; muscle, n = 6. (D)
   Quantification of telomere dysfunction–induced foci (TIF) measured as
   number of telomeric probe and DNA damage marker 53BP1 colocalizing foci
   per cell by Immuno FISH in the kidney and muscle of old mice. PBS:
   kidney, n = 5; muscle, n = 5; sEVs: kidney, n = 6; muscle, n = 6. (E
   and F) Quantification of SASP factors IL-6 and IL-1β in the kidney and
   muscle of old mice. PBS: kidney, n = 5; muscle, n = 5; sEVs: kidney, n
   = 5; muscle, n = 5. (G to L) Quantification and representative
   immunofluorescence images of LMNB1 loss and γH2AX^+ cells in the kidney
   and gastrocnemius of old mice. PBS: kidney, n = 4; muscle, n = 4; sEVs:
   kidney, n = 4; muscle, n = 4. (M) Graphical representation of in vitro
   experiment with senescent murine myoblasts. (N and O) Assessment of
   senescence-associated β-galactosidase activity and annexin V by flow
   cytometry in senescent murine myoblasts. Untreated, n = 5; palbociclib,
   n = 5; palbociclib + sEVs, n = 5. All data are shown as means ± SD. (P)
   Quantification of Ki67^+ cells in senescent murine myoblasts.
   palbociclib, n = 3; palbociclib + sEVs, n = 3. (Q and R) Quantification
   of SASP factors IL-6 and IL-1β in senescent murine myoblasts.
   Palbocilib, n = 5; palbociclib + sEVs, n = 5. (S) Assessment of
   senescence-associated β-galactosidase activity by flow cytometry in
   senescent murine myoblasts. Palbociclib, n = 6; palbociclib + 3T3-sEVs,
   n = 6; palbociclib + C2C12-sEVs, n = 6. All data are shown as means ±
   SD.

   Telomere attrition and damage to telomeres are two of the most studied
   factors that lead to genomic instability and loss of proliferative
   capacity. Telomere shortening is observed in many species during normal
   aging, and excessively short telomeres are associated with multiple
   conditions characterized by a loss of regenerative capacity, such as
   pulmonary fibrosis, dyskeratosis congenita, and aplastic anemia
   ([125]42). We explored the effect of ADSC-sEVs on telomere length and
   damage, as possible mediators of the observed effects in mice. For this
   purpose, we measured mean telomere length and DNA damage at the
   telomeres quantified as telomere dysfunction–induced foci (TIF). We did
   not find significant differences between both groups ([126]Fig. 4, C
   and D), although we found that the muscle of treated mice showed
   slightly lower telomere damage than the nontreated ones.

   Chronic inflammation is a condition closely related to aging and
   senescence; an increase in proinflammatory cytokines in several tissues
   has been linked to aging, frailty, and age-related diseases ([127]7).
   Interleukin-6 (IL-6) and IL-1β are two of the most important factors of
   the senescence-associated secretory phenotype (SASP), so we measured
   the levels of these molecules in the kidney and the muscle of old mice,
   observing a decrease in IL-6 and IL-1β levels in both tissues from
   ADSC-sEV–treated mice ([128]Fig. 4, E and F).

   Senescent cells accumulate in aged tissues, having a detrimental effect
   on their function, as they are unable to enter the cell cycle. In
   addition, these cells act in a pleiotropic manner releasing several
   factors that make up the SASP, contributing to the development of
   age-associated diseases ([129]43, [130]44). We measured the loss of
   lamin B1 (LMNB1) and γH2AX as senescence markers ([131]45) in muscle
   and kidney of old mice 30 days after treatment with ADSC-sEVs/PBS and
   noticed an important decrease in the number of senescent cells in both
   tissues in mice treated with ADSC-sEVs ([132]Fig. 4, G to L).

   Last, to test the effect of ADSC-sEVs on a more controlled environment,
   we developed an in vitro model of senescence in muscle progenitor
   cells. For this purpose, we used C2C12 mouse myoblasts and induced
   cellular senescence with palbociclib (5 μM). We treated the senescent
   cells with young ADSC-sEVs (5 μg/ml), and we measured
   senescence-associated β-galactosidase (SABG) as a senescence marker and
   annexin V as an apoptosis marker (see [133]Fig. 4M for experimental
   design). We found that cells that received ADSC-sEVs showed a decrease
   in senescence and a decrease in the percentage of apoptotic cells
   ([134]Fig. 4, N and O). As one of our main hypotheses was that
   ADSC-sEVs could be promoting the proliferation of specific cells in old
   individuals, we quantified Ki67^+ cells, showing an increase in cells
   treated with ADSC-sEVs ([135]Fig. 4P). We also found a decrease in the
   release of proinflammatory molecules of the SASP (IL-6 and IL-1β) by
   the senescent cells treated with ADSC-sEVs ([136]Fig. 4, Q and R).
   Last, we checked the effect of sEVs from other origins in senescent
   murine myoblasts (C2C12), using sEVs from murine fibroblasts (NIH3T3
   cell line) and nonsenescent C2C12 myoblasts. We could not find any
   effect on the proportion of senescent cells using these sEVs ([137]Fig.
   4S), suggesting that sEVs from ADSCs contain specific factors that play
   a role in senescence. Overall, these findings prove that treatment with
   ADSC-sEVs can alleviate molecular and cellular traits of aging in the
   kidney and muscle of aged mice, inducing proliferation and a decrease
   in senescence, apoptosis, and SASP factors in an in vitro model of
   senescence.

Predicted epigenetic age is lower in tissues of old mice treated with
ADSC-sEVs

   We used multiple mouse clocks trained on different tissues to estimate
   the age of our treated and untreated animals. We grouped the clocks
   into two groups. The general clocks were trained across a wide spectrum
   of animals. The intervention group of clocks was trained on animals
   with age-affecting interventions. We used mixed-effects modeling to
   discern a trend across multiple clocks. In the kidney tissue samples,
   age estimates after the ADSC-sEV treatment were significantly lower for
   both general and intervention clocks (P = 0.006 and P = 0.0002,
   respectively). Similarly, in the liver tissue samples, both models
   showed a reduction in estimated age with the interventions group
   passing the statistical significance threshold (P = 0.06 and P = 0.001
   for general and intervention clocks, respectively). Muscle and spleen
   tissues did not show a significant difference in age prediction between
   treated and control animal groups ([138]Fig. 5A).

Fig. 5. ADSC-sEVs decrease the epigenetic age of several tissues in old mice.

   [139]Fig. 5.
   [140]Open in a new tab

   (A) ADSC-sEV treatment effect on epigenetic age estimation
   mixed-effects modeling across several tissues. (B to E) Biological
   process enrichment analysis of differentially methylated cytosines in
   ADSC-sEV–treated tissues (liver, muscle, kidney, and spleen,
   respectively). PBS, n = 6; ADSC-sEVs, n = 6. All data are shown as
   means ± SD. BMP, bone morphogenetic protein; GTPase, guanosine
   triphosphatase; MAP, mitogen-activated protein; MAPK, MAP kinase.

   Linear regression analysis revealed abundant differentially modified
   loci in all tissues [n = 5.546, 36.163, 7.247, and 11.092 in the
   spleen, liver, kidney, and muscle, respectively; robust linear
   regression, false discovery rate (FDR), q < 0.05]. We assigned each
   locus the nearest gene name and performed gene set enrichment analysis
   using g:profiler ([141]46). Although affected pathways were different
   across tissues, we observed developmental, neuronal, and synaptic
   pathways similarly affected in all tissues ([142]Fig. 5, B to E).

ADSC-sEVs induce a change in the metabolome of old mice to a youth-like
pattern

   We quantified 72 metabolites in blood plasma samples from seven
   PBS-treated and six ADSC-sEV–treated old mice from four control young
   mice. Fourteen of these metabolites show statistically significant
   differences between ADSC-sEV–treated and untreated old animals
   ([143]Fig. 6, A and B, and table S5). Most of the significant
   metabolites were amino acids including essential (isoleucine,
   tryptophan, threonine, and valine) and nonessential (aspartate,
   arginine, tyrosine, glycine, and proline). Some of these amino acids
   have been previously associated with metabolic health like the
   branched-chain amino acids isoleucine and valine ([144]47) as well as
   tryptophan ([145]48). Lactate is formed by the anaerobic glycolysis in
   most mammalian tissues. Together with acetate, it is also a product of
   fermentation of fructans and other prebiotics by gut microbiota
   ([146]49). They are further processed into butyrate and other
   short-chain fatty acids by some strains of bifidobacteria and
   lactobacilli, which suggest some impact of ADSC-sEV treatment in gut
   microbiota ecosystems. The heatmap of these metabolites shows a
   clustering trend among groups ([147]Fig. 6A) and some resemblance
   between ADSC-sEV–treated old mice and young mice ([148]Fig. 6B), i.e.,
   red dots (old-sEV mice) differ from black ones (old-PBS mice) but
   coincide with blue dots (young mice).

Fig. 6. The metabolome of ADSC-sEV–treated mice resembles that of young
animals.

   [149]Fig. 6.
   [150]Open in a new tab

   (A) Heatmap of metabolite z scores with statistically significant
   differences between old-PBS (black bar samples) and old-sEV (red bar
   samples) mice compared with young untreated mice (blue bar samples).
   (B) Mean and 90% confidence intervals for z scores in SD units for the
   same metabolites that (A) in old-PBS (black), old-sEVs (red), and young
   untreated (blue) mice. (C) Score plot for a metabolome partial
   least-squares discriminant analysis (PLS-DA) model (72 metabolites) for
   discrimination between young (blue squares) and old-PBS (black circles)
   mice. Old mice treated with sEVs (red circles) are projected in the
   score plot. (D) Variable importance in projection (VIP) scores in the
   metabolome PLS-DA model of (C) for those metabolites significantly
   associated with treatment. (E) Enrichment ratio (size of the circle)
   and P values (color of the circle) for the 25 most enriched metabolite
   sets from a metabolite set enrichment analysis (MSEA) on the Small
   Molecule Pathway Database (SMPDB) and our 14 significant metabolites.
   The n used for all metabolomic analyses was young mice, n = 4; old-PBS
   mice, n = 6; and old-sEV mice, n = 6.

   To further explore this resemblance, we built a partial least-squares
   discriminant analysis (PLS-DA) model with all the metabolites for
   discriminating between old and young untreated animals and projected
   into this model the samples from ADSC-sEV–treated animals. The score
   plot of all the samples in this model shows the same resemblance
   between old treated and young animals along the latent variable 2 (LV2)
   dimension, suggesting that ADSC-sEV treatment partially ameliorates the
   impact of aging on the metabolome ([151]Fig. 6C). Thirteen of the
   fourteen metabolites significantly associated with treatment also show
   a PLS-DA VIP (variable importance in projection) score higher than 1,
   which confirms interactions between age-related metabolomic changes and
   ADSC-sEV treatment ([152]Fig. 6D).

   Last, we performed a metabolite set enrichment analysis (MSEA)
   concerning the Small Molecule Pathway Database (SMPDB) for mammal’s
   metabolism to gain biological insight into the metabolome changes
   induced by the ADSC-sEV treatment. The only three metabolite clusters
   enriched in our data with statistical significance were the arginine
   and proline metabolism cluster, the glycine and serine metabolism
   cluster, and the aspartate metabolism cluster with five, four, and
   three hits, respectively ([153]Fig. 6E). These three clusters,
   especially the arginine and proline clusters, have been previously
   associated with cardiometabolic health in humans ([154]50). Besides the
   well-known role of arginine as urea or nitric oxide precursor, arginine
   is also a precursor of proline and is relevant for collagen
   biosynthesis ([155]51). Proline and glycine, the two most abundant
   amino acids in collagen accounting for 23% of the molecule, are also
   represented in the first three clusters of our MSEA suggesting some
   involvement of collagen metabolism in the potential beneficial effects
   of ADSC-sEV treatment.

Analysis of sEV miRNA content reveals that up-regulated miRNAs in young
ADSC-sEVs govern tissue development and regeneration pathways

   Motivated by these results, we attempted to investigate candidate
   factors present in ADSC-sEVs that could be responsible for the
   beneficial effects in old mice. miRNAs are an important fraction of
   several types of EV content and can regulate transcription in several
   tissues ([156]52). Here, we used miRNA profiling (Affymetrix GeneChip
   miRNA 4.0 Array) to study miRNAs contained in sEVs. We obtained sEVs
   from three different conditions: supernatant of young and old ADSCs,
   and plasma from aged mice. In the analysis, miRNAs with a twofold
   increase or decrease in expression and a P value < 0.01 were defined as
   significantly different. Principal components analysis (PCA) showed
   three differentiated groups of samples ([157]Fig. 7A). Analysis of 3196
   miRNAs in sEVs from young ADSC cultures showed several differentially
   expressed miRNAs when compared to sEVs from old ADSC cultures (9
   up-regulated and 1 down-regulated) and from plasma of aged mice (25
   up-regulated and 4 down-regulated; [158]Fig. 7, B to D). In all
   comparisons, sEVs from young ADSCs showed an increased miRNA expression
   profile, in accordance with previous findings that showed a deregulated
   miRNA biosynthesis with aging, especially in the adipose tissue
   ([159]53–[160]55). These miRNAs are shown in tables S6 to S8.

Fig. 7. Tissue development and regeneration are major targets of miRNAs
enriched in young ADSC-sEVs, with an effect on senescence in vitro.

   [161]Fig. 7.
   [162]Open in a new tab

   (A) PCA of miRNA profiling in sEVs from young ADSCs, old ADSCs, and
   plasma of aged mice. Old plasma, n = 4; old ADSC-sEVs, n = 4; and young
   ADSC-sEVs, n = 4. (B) Summary of differentially expressed miRNAs under
   the three conditions. (C) Heatmap of differentially expressed miRNAs in
   sEVs from young and old ADSCs. (D) Heatmap of differentially expressed
   miRNAs in sEVs from young ADSCs and plasma of aged mice. (E) Volcano
   plot showing the different miRNA expression between plasma of aged mice
   and young ADSC-sEVs. NS, not significant. (F and G) Assessment of
   senescence-associated β-galactosidase activity by flow cytometry in
   senescent murine myoblasts treated with miRNA mimics. Vehicle (VEH), n
   = 13; small interfering RNA control, n = 13; miR-125b-5p, n = 9;
   miR-let7c-5p, n = 17; miR-214-3p, n = 15; miR-145a-5p, n = 17;
   miR-135a-3p, n = 17; miR-143-3p, n = 9.

   As we injected the young ADSC-sEVs into the blood of old mice, we
   reasoned that miRNAs up-regulated in the young ADSC-sEVs and with low
   levels in plasmatic sEVs from aged mice may be, at least in part,
   responsible for the beneficial effects of ADSC-sEVs. Thereafter, we
   explored the predicted mRNA targets for the up-regulated 25 miRNAs and
   their role in biological processes and molecular pathways. However, as
   the number of predicted targets was too high for further analysis, we
   outlined six miRNAs as plausible biologically relevant, i.e., that can
   be considered relevant because these miRNAs (highlighted in red in
   [163]Fig. 7E) are the ones that share their nucleotide sequence among
   several species, including humans, as many features of the aging
   process are highly conserved across species ([164]56).

   Together, these six miRNAs were predicted to target 5.379 genes with
   very high stringency (top 1% in confidence, present in at least five
   databases as curated by miRDIP). The target gene dataset was then
   introduced in PANTHER to study overrepresented biological processes and
   pathways ([165]57). A great number of genes targeted by our miRNAs were
   involved in cell proliferation, tissue development, growth, and
   regeneration (fig. S2). Regarding molecular pathways, these miRNAs
   regulated important pathways related to tissue regeneration, fibrosis,
   and inflammation, such as angiogenesis, Wnt signaling, platelet-derived
   growth factor, fibroblast growth factor, transforming growth factor–β,
   IL, and interferon-γ signaling. They targeted critical pathways in the
   aging process, including apoptosis, p53 signaling, p38
   mitogen-activated protein kinase, oxidative stress response, and
   insulin/IGF pathway (fig. S2). In conclusion, predicted targets of the
   selected miRNAs up-regulated in young ADSC-sEVs compared with plasma
   sEVs of old mice are probably involved in several relevant processes
   and pathways affected during aging.

   To explore the role of these miRNAs in a biological process related to
   aging, we treated senescent murine myoblasts with miRNA mimics
   ([166]Fig. 7F). We found that three of our selected miRNAs—miR-125b-5p,
   miR-let7c-5p, and miR-214-3p—were able to lower senescence in vitro
   ([167]Fig. 7G), indicating a plausible role of miRNAs contained in sEVs
   as modulators of this process.

DISCUSSION

   Aging is the largest risk factor for most diseases that affect people
   at the late stages of their lives. Our understanding of the aging
   process has grown greatly in the past decades, providing us with
   several biochemical changes that are considered drivers of this
   process; oxidative stress, telomere attrition, and cellular senescence
   are some of these biological processes that are proposed as drivers of
   age-related morbidities ([168]2). An altered intercellular
   communication is one of the ultimate culprits of tissue and organismal
   aging, and several players of intercellular communication have been
   proposed as possible antiaging factors in parabiosis experiments
   ([169]10, [170]11, [171]58). sEVs could be one important fraction of
   the factors present in young environments that can have a beneficial
   effect on aged individuals. sEVs from MSCs have a regenerative capacity
   themselves, can help tissue to regenerate after damage, and are
   potential cell-free therapeutic agents ([172]59). Nicotinamide
   phosphoribosyl transferase was shown to be present within plasmatic
   EVs, and these EVs were able to ameliorate several parameters affected
   by aging and increased the life span of old mice ([173]31). Similarly,
   sEVs have been described as carriers of glutathione S-transferase
   activity, which can ameliorate senescence and oxidative stress–related
   damage in old mice ([174]29). More recently, sEVs from MSCs were able
   to extend life span in a transgenic mouse model of accelerated aging
   and mitigated several senescence-associated markers in vivo and vitro
   ([175]60).

   Given this background, the effect of sEVs from young MSCs on the
   function of tissues affected by aging in physiologically aged mice
   remained to be elucidated. In the present work, we have shown that
   young ADSC-sEVs have great potential as antiaging factors, as old mice
   injected with these sEVs showed an improvement in several functions
   affected by aging, such as physical condition, fur regeneration, and
   renal function, as well as a reduction of frailty. As we based our
   determination of renal function on the concentration of urea, it is
   important to mention that this might be a result of a decreased
   catabolic rate in skeletal muscle, as we have observed an increase in
   CSA and protein content in muscle of ADSC-sEV–treated mice. The effect
   of ADSC-sEVs seems to be finite, as we observed in a more long-term
   experiment that the effect on the functionality of old mice was lost 2
   months after treatment with ADSC-sEVs. Moreover, sEVs induced
   structural changes in tissues from old mice, exhibiting a potent
   proregenerative effect in these tissues, which is in accordance with
   previous results that proposed sEVs as a regenerative tool ([176]29).
   We observed a reduction of senescence in tissues and in vitro when sEVs
   were introduced; however, the mechanism of action remains unclear, as
   we did not find senolytic activity. They may probably act as
   senomorphics, molecules that suppress the senescent phenotype without
   the specific induction of apoptosis in senescent cells, probably
   through the inhibition of the SASP, as has been suggested recently
   ([177]60). We found a reduced level of IL-6 and IL-1β, two of the main
   SASP factors, in the tissues from ADSC-sEV–treated mice and in the in
   vitro model of senescence. IL-6 is a key factor of the SASP, but its
   role in muscle physiology and function goes beyond inflammation, as it
   is an important myokine that stimulates its expression with exercise
   ([178]61); thus, the modulation of this factor may be key in the
   effects observed in the muscle of treated mice. All these effects found
   in tissues of ADSC-sEV–treated mice were accompanied by a decrease in
   the epigenetic age estimation from epigenetic clocks, which are
   recognized as a feasible estimation of biological age ([179]62). Some
   of the most studied interventions in aging, such as caloric restriction
   or rapamycin treatment, have demonstrated a strong impact on predicted
   epigenetic age ([180]63–[181]65). To our knowledge, this study is the
   first to show that intervention with sEVs can decrease epigenetic age
   in an old organism.

   sEVs induced effects not only in individual tissues but also in the
   whole metabolome of old mice, which turned to a youth-like pattern,
   indicating that the effects observed are probably part of a pleiotropic
   effect on the organism. Now as it is time to continue investigating the
   molecules that are included in sEVs, we have explored miRNAs contained
   in young ADSC-sEVs and found that they are involved in several
   processes and pathways affected by aging, thus proposing miRNAs as
   possible mediators of the effects shown in mice. Taking into account
   our results and other preliminary studies, miR-214-3p may play a role
   in senescence ([182]66, [183]67). It is important to point out the
   debate on the relevance of miRNAs in the function of EVs, as recent
   studies have shown low levels of miRNAs in EVs, along with a limited
   delivery into target cells ([184]68, [185]69). More studies are needed
   to identify factors derived from stem cells that can assist tissue
   function and regeneration, as they could have an enormous impact on
   age-related pathologies, such as frailty or renal failure.

MATERIALS AND METHODS

Animal model

   This study was performed in strict accordance with all applicable
   federal and institutional policies. The protocol was approved by the
   University of Valencia Animal Ethics Committee following European Union
   (EU) regulations on animal research (identification numbers
   A1490612630929 and A1508582840889). All the mice used in this study
   were of a C57BL/6J background. Aged mice used were treated at 20 to 24
   months of age. Both sexes were used throughout the study. Where
   feasible, littermates of the same sex were used. The sample size needed
   was calculated using G*Power. These were randomly assigned to
   experimental groups. All experiments were addressed blindly. The
   experimental procedure was replicated throughout the study, as mice
   from six batches (sEVs versus PBS) were used. After mice were
   euthanized, plasma and organs were obtained and stored at −80°C. For
   all the experiments conducted in the kidney and gastrocnemius, we used
   organs obtained from mice 30 days after treatment with sEVs/PBS. The
   mean age of mice that received sEVs was 21,925 months and that of mice
   that received PBS was 21,975 months. For the mice that were euthanized
   on day 30 to obtain tissues for the analysis, the mean age was 22.15
   months for the sEV group and 21.75 months for the PBS group.

Stem cell culture

   Mice at 3 to 6 months of age from both sexes were used to obtain MSCs
   from both inguinal fat pads with the previously described protocol
   ([186]70). All cells were used in passages 2 and 3 for the experiments.
   The culture medium used for the expansion and maintenance of the cells
   was high-glucose Dulbecco’s modified Eagle’s medium (DMEM) with 10% FBS
   and 1% penicillin/streptomycin (P/S). Cells were maintained at 37°C, 5%
   CO[2], and 3% O[2] (Whitley H35 HEPA Hypoxystation).

ADSC characterization

   For the characterization of ADSCs, we analyzed the presence of typical
   MSCs markers (CD29, CD44, CD90, and CD105) by flow cytometry; CD45 and
   CD31 were used as negative controls. When cells reached 80% confluency,
   they were trypsinized and resuspended in fluorescence-activated cell
   sorting (FACS) buffer (10% FBS and 1% sodium azide in PBS). A total of
   100,000 cells were used for each condition; they were stained with
   antibody (1 μg/μl) and incubated for 30 min at 4°C. All antibodies were
   acquired from BioLegend: CD29 R-phycoerythrin/Cyanine7 (PE/Cy7), 10222;
   CD31 PE, 102507; CD44 Alexa Fluor 488, 103015; CD45 PE, 103106; CD90
   fluorescein isothiocyanate (FITC), 140303; and CD105 allophycocyanin
   (APC), 120413.

   ADSCs isolated from inguinal fat pads of young mice showed a strong
   expression of MSCs markers; cells were positive (>80%) for CD29, CD44,
   CD90, and CD105 and negative for hematopoietic (CD45) and endothelial
   (CD31) markers (fig. S3).

ADSC-derived sEV isolation

   ADSCs at passages 2 and 3 were expanded until they reached 80%
   confluency; at that moment, medium was changed to high-glucose DMEM
   with 2% exosome-depleted FBS (Gibco, A2720803) and 1% P/S. After 48
   hours, conditioned medium was collected, and sEVs were isolated by
   differential ultracentrifugation. Medium was centrifuged at 2000g for
   10 min and then at 20,000g for 30 min to remove whole cells, cell
   debris, and bigger EVs. The supernatant was then ultracentrifuged at
   100,000g for 70 min. Pelleted vesicles were suspended in PBS,
   ultracentrifuged again at 100,000g for 70 min for washing, resuspended
   in PBS, and prepared for treatment, NTA, electron microscopy, or flow
   cytometry analysis. For the non-EV fraction injections, we used 20 μg
   of protein from conditioned cell culture medium obtained after the
   first ultracentrifugation step.

ADSC-derived sEV NTA

   The size distribution of vesicles was determined by NTA in a NanoSight
   LM10 (Malvern Instrument Ltd., Malvern, UK), using a 405-nm laser and a
   scientific complementary metal-oxide semiconductor camera. Data were
   analyzed with the NTA software version 3.3. (Dev Build 3.3.104), with
   Min track Length, Max Jump Distance, and Blur set to auto, and the
   detection threshold was set to 5. The camera level was set to 15 and
   five readings of 30 s at 30 frames per second. Images were taken with
   manual monitoring of temperature. Samples were diluted at 1:200 with
   filtered PBS to reach the concentration (20 to 120 particles per frame)
   recommended by the manufacturer.

ADSC-derived sEV immunoblotting

   sEVs were resuspended and lysed in tris/SDS/glycerol buffer, protein
   concentration was determined using the Lowry method, and 20 μg of
   proteins were separated on SDS polyacrylamide gels and transferred onto
   nitrocellulose membranes; as a control for the sEV immunoblotting, we
   used FBS used in the isolation of sEVs (Gibco). The membranes were
   blocked with 3% bovine serum albumin (BSA) in tris-buffered saline with
   0.05% Tween 20 (TBS-t) for 60 min at room temperature (RT) and
   incubated overnight at 4°C with anti-TSG101 antibody (Proteintech,
   28283-1-AP; 1:1000 dilution) or BSA antibody (Proteintech, 66201-1-Ig).
   Following three washes (10 min) with TBS-t, membranes were incubated
   with secondary antibody (anti-rabbit; Cell Signaling Technology, 70745;
   1:2000 dilution; anti-mouse, BioLegend, 401215; 1:1000 dilution) for 60
   min at RT. Following three washes with TBS-t, membranes were developed
   with Luminol (Sigma-Aldrich) in the ImageQuant LAS4000 system. Images
   were processed in ImageJ.

ADSC-derived sEV flow cytometry

   For flow cytometry analysis, pellets obtained from 10 ml of medium were
   resuspended in 100 μl of PBS. sEVs were stained with APC anti-CD63 (4
   μg/ml; BioLegend, 143905) for 30 min at 4°C in darkness. After
   incubation, positive events were read by FACSVerse flow cytometry. One
   unstained sample of sEVs and one sample with the antibody at 4 μg/ml in
   PBS were used as negative controls. Calibration of the cytometer was
   assessed using fluorescent particles of standardized size (Nano
   Fluorescent Particle Size Standard Kit, Spherotech).

ADSC-derived sEV electron microscopy

   The isolated sEVs were fixed in 2% paraformaldehyde (PFA)–0.1 M PBS for
   30 min. Glow discharge technique (30 s, 7.2 V, using a Bal-Tec MED 020
   Coating System) was applied over carbon-coated copper grids, and
   immediately, these grids were placed on top of sample drops for 15 min.
   Then, the grids with adherent sEVs were washed in a 0.1 M PBS drop, and
   additional fixation in 1% glutaraldehyde was performed for 5 min. After
   washing properly in distilled water, the grids were contrasted with 1%
   uranyl acetate and embedded in methylcellulose. Excess fluid was
   removed and allowed to dry before examination with a transmission
   electron microscope FEI Tecnai G2 Spirit (Thermo Fisher Scientific, OR,
   USA). All images were acquired using a digital camera Morada (EMSIS
   GmbH, Münster, Germany).

Immunogold labeling of sEVs

   In brief, 8 μl of isolated sEVs were fixed in 2% PFA–0.1 M PBS for 30
   min, and carbon-coated nickel grids were placed on top of these sEV
   drops for 15 min. Then, the grids with adherent sEVs were washed in 0.1
   M PBS and blocked in 0.1 M glycine and 0.3% BSA for 10 min. The grids
   were incubated with anti-CD63 (MBL, D263-3) primary antibody (1:100
   dilution) for 1 hour. After another blocking step for 10 min, the grids
   were incubated in Gold 6nm–conjugated goat anti-rat (Abcam, ab105300)
   secondary antibody (1:1000 dilution) for 1 hour. Last, after washing, a
   standard negative staining procedure was made and observed under a
   transmission electron microscope as previously described.

In vivo tracking of sEVs with PKH26 staining

   For the in vivo tracking of sEVs, we isolated sEVs from the conditioned
   medium of young ADSCs and labeled them after the first ultracentrifuge
   with 5 μM PKH26 (Sigma-Aldrich, MINI26-1KT), after that, sEVs were
   resuspended in PBS and washed with another ultracentrifuge. Twenty
   micrograms of PKH26-labeled sEVs were injected in old mice (18 months),
   mice were euthanized 24 hours later, and kidney and gastrocnemius were
   obtained and fixed with 2% PFA for 24 hours. Forty-micrometer slices
   were mounted and stained with Hoechst (Invitrogen) 1:1000 for 30 min at
   RT. Coverslips were mounted with an aqueous mounting medium and sealed
   with nail polish. Images were acquired with Olympus FV1000 confocal
   laser scanning biological microscope. Images were processed in ImageJ,
   all levels were adjusted equally, and the ratios were not altered.

Treatment preparation

   sEVs isolated from the conditioned medium were kept at 4°C and used for
   treatment within 24 hours. To determine the dose of sEVs, we performed
   protein quantification with the Lowry method of the sEV samples. Each
   dose consisted of either 20 μg of sEVs’ protein or PBS as a control in
   a total volume of 100 μl. Mice were intravenously injected with the
   preparation in the tail vein.

Physical condition tests

   The Grip Strength Meter (Panlab, Harvard Apparatus) was used in
   assessing neuromuscular function by sensing the peak amount of force
   that the mice applied in grasping specially designed pull bar
   assemblies. Peak force was automatically registered in gram-force by
   the apparatus. Data were recorded, and two additional trials were
   immediately given. Maximum force was normalized to the weight of the
   mouse. Animals were submitted to a graded intensity treadmill test
   (Treadmill Control LE 8710 Panlab, Harvard Apparatus) to determine
   their endurance (running time) and running speed along with the study.
   After a warm-up period, the treadmill band velocity was increased until
   the animals were unable to run further. The initial bout of 4 min at 10
   cm/s was followed by consecutive 4 cm/s increments every 2 min.
   Exhaustion was reached when a mouse remained on the shock grid for 5 s
   rather than running. Motor coordination was assessed with a Rota Rod
   (Panlab, Harvard Apparatus no. 76-0772), consisting of a 3-cm-wide
   wheel that rotates with an increasing speed. Time to fall was recorded
   for each mouse, with a total of three trials. To assess frailty
   quantitatively, we used a score based on the clinical phenotype of
   frailty developed in our group ([187]34), mice with three or more
   criteria were classified as frail. Criteria used are as follows: loss
   of 5% of body weight, time to fall in Rota Rod under percentile 20
   (p20), time to exhaustion in a treadmill under p20, maximum speed
   reached in a treadmill under p20, and normalized grip strength under
   p20.

Hair regrowth assay

   On day 1, dorsal hair was removed by plucking from a square of
   approximately 1 cm by 1 cm. Hair regrowth was scored 2 weeks later,
   based on digital photographs and a semiquantitative assessment, using
   an arbitrary scale from one to four (where four represents complete
   hair regeneration). Scoring was performed blindly by two independent
   investigators.

Plasmatic urea value measurement

   One hundred microliters of whole blood was obtained from the saphenous
   vein of each mouse just before treatment; after that, whole blood was
   obtained on days 14, 30, and 60. Whole-blood samples were collected in
   a microvette with EDTA for plasma separation and spun for 15 min at
   1500g. The clear supernatant was transferred into regular 1.5-ml tubes,
   snap-frozen in liquid N[2], and stored at −80°C. Urea concentration was
   measured using a QuantiChrom urea assay kit (Gentaur). The samples were
   incubated in a 200-μl reaction mix for 20 min at RT before absorbance
   was measured at 520 nm (Molecular Devices SPECTRAmax Plus 384). All
   samples were assayed in duplicate.

Histological analysis

   Kidneys and gastrocnemius were fixed in 4% PFA for 48 hours and then
   cryoprotected with 30% sucrose in PBS; 10-μm slices were obtained in a
   cryostat for histological staining and immunofluorescence. Slices were
   stained with hematoxylin (Sigma-Aldrich, MHS32) and eosin
   (Sigma-Aldrich, E4009) or with Sirius red (Sigma-Aldrich, 365548),
   mounted, and sealed for further morphometric analysis. Images were
   obtained using an optical microscope (Leica), and three images from
   different areas of each slice were obtained. All levels were adjusted
   equally, the ratios were not altered, and morphometric analysis of
   kidney and muscle sections was performed with ImageJ.

Immunofluorescence

   Ten-micrometer tissue slices were mounted in slides and heated for 30
   min in tris-EDTA buffer (pH 10) for antigen retrieval and permeabilized
   with 1% Triton X-100 in PBS (Ki-67, γH2AX) or 1% Tween 20 in PBS
   (MHC2A, Laminin, LMNB1), depending on the antibody. Sections were
   blocked with 10% normal goat serum (Invitrogen) in PBS with 0.05% Tween
   20 (PBS-t) and incubated with primary antibody overnight at 4°C. After
   incubation with primary antibody, sections were washed three times with
   PBS-t and incubated with secondary antibody for 2 hours at RT, washes
   were repeated, and tissues were counterstained with Hoechst
   (Invitrogen; 1:1000 dilution) for 30 min at RT. Coverslips were mounted
   with an aqueous mounting medium and sealed with nail polish.

   Images were acquired with Olympus FV1000 confocal laser scanning
   biological microscope. Images were processed in ImageJ, all levels were
   adjusted equally, the ratios were not altered, and three images from
   different areas of each slice were obtained.

   Ki67 staining antibodies: anti-Ki67 (Thermo Fisher Scientific,
   PA5-19462; 1:1000 dilution) and Alexa Fluor 647 anti-rabbit (Abcam,
   ab150079; 1:1000 dilution); a minimum of 1000 nuclei were analyzed per
   sample. MHC2A staining antibodies: anti-MHC2A (DSHB, 2F7; 0.5 μg/ml)
   and Alexa Fluor 647 anti-mouse (Cell Signaling Technology, 4410; 1:1000
   dilution); a minimum of 200 fibers were analyzed per sample. Laminin
   staining antibodies: anti-laminin (Thermo Fisher Scientific, PA1-16730;
   1:500 dilution) and Alexa Fluor 488 anti-rabbit (Abcam, ab150079;
   1:2000 dilution); a minimum of 200 fibers were analyzed per sample.
   LMNB1 staining antibodies: anti-LMNB1 (Proteintech, 12987-1-AP; 1:50
   dilution) and Alexa Fluor 488 anti-rabbit (Abcam, ab150079; 1:2000
   dilution); a minimum of 400 nuclei for kidney and 100 nuclei for muscle
   were analyzed per sample. γH2AX staining antibodies: anti- γH2AX (Cell
   Signaling Technology, 9718S; 1:1000 dilution) and Alexa Fluor 488
   anti-rabbit (Abcam, ab150079; 1:2000 dilution); a minimum of 400 nuclei
   for kidney and 140 nuclei for muscle were analyzed per sample. For the
   intramuscular fat staining, slices were not heated for antigen
   retrieval, and saponin in 0.05% PBS was used as a detergent. Slices
   were incubated with LipidTOX (Thermo Fisher Scientific, 10242283;
   1:1000 dilution) for 30 min at RT and counterstained with Hoechst
   (Invitrogen; 1:1000 dilution) for 30 min at RT.

Lipid peroxidation measured using HPLC

   Tissues were lysed with a KPi-EDTA buffer [50 mM KPi and 1 mM EDTA (pH
   7.4)], and lipid peroxidation was determined as MDA levels, which were
   detected using high-performance liquid chromatography (HPLC) as an
   MDA–thiobarbituric acid (TBA) adduct following a method described
   previously ([188]71). This method is based on the hydrolysis of
   lipoperoxides and the subsequent formation of an adduct between TBA and
   MDA (TBA-MDA2). This adduct was detected using HPLC in the reverse
   phase and quantified at 532 nm. The chromatographic technique was
   performed under isocratic conditions, the mobile phase being a mixture
   of monopotassium phosphate at 50 mM (pH 6.8) and acetonitrile (70:30).
   The level of MDA in each sample was divided by the concentration of
   protein determined by the Lowry method.

Immunoblotting for protein oxidation

   To detect total protein carbonylation by immunoblot, we used the
   OxyBlot Protein Oxidation Detection kit (Merck) following the
   manufacturer’s instructions. In brief, tissues were lysed in
   tris/SDS/glycerol buffer, protein concentration was determined by the
   Lowry method, and 20 μg of proteins were separated on SDS
   polyacrylamide gels and transferred onto nitrocellulose membranes. The
   membranes were blocked with 3% BSA in TBS-t for 60 min at RT and
   incubated overnight at 4°C with primary antibody from the kit.
   Following three washes (10 min) with TBS-t, membranes were incubated
   with secondary antibody for 60 min at RT. Following three washes with
   TBS-t, membranes were developed with Luminol (Sigma-Aldrich) in the
   ImageQuant LAS4000 system. Images were processed in ImageJ. Ponceau
   staining of the membranes was used as the loading control.

Telomere length and telomere damage assays

   Telomere length and the number of TIF were calculated using the
   immunofluorescence-FISH (IF-FISH) technique. Briefly, frozen slides of
   the kidney and muscle were permeabilized with 0.5% Triton X-100 and
   washed in PBS. After 1 hour of blocking with 5% BSA-PBS, slides were
   incubated with 53BP1 antibody (Novus Biologicals, NB100-304) overnight
   at 4°C. After several washes in PBS–0.1% Tween 20, slides were
   incubated in donkey anti-rabbit Alexa Fluor 488 antibody (Invitrogen,
   A21206) for an hour and washed again in PBS. Tissues were then fixed in
   4% formaldehyde for 20 min, washed several times in PBS, and dehydrated
   in a 70%-90%-100% ethanol series (5 min each). Slides were left to air
   dry, and 30 μl of telomere probe mix [10 mM tris-Cl (pH 7), 25 mM
   MgCl[2], 9 mM citric acid, 82 mM Na[2]HPO[4], 50% deionized formamide
   (Ambion AM9342), 0.25% blocking reagent (Roche 11096176001), and
   telomeric PNA-Cy3 probe (0.5 μg/ml; Panagene)] was added to each slide.
   A coverslip was added, and slides were incubated for 3 min at 85°C and
   a further 2 hours at RT in a wet chamber in the dark. Slides were
   washed two times for 15 min in 10 mM tris-Cl (pH 7), were vigorously
   shaken with 0.1% BSA in 50% formamide, then washed three times for 5
   min in TBS 0.08% Tween 20, and were rinsed in PBS. Slides were mounted
   in VECTASHIELD with 4′,6-diamidino-2-phenylindole (Vector H-1200-10).
   Confocal images were acquired as stacks using a Leica SP5-MP confocal
   microscope, and maximum projections were done with the LAS-AF software.
   Telomere signal intensity was quantified using Definiens software, and
   sites of colocalization of telomeric PNA-Cy3 probe and 53BP1
   fluorescent signals were counted per cell. Image analysis was performed
   blindly.

SASP factor quantification

   We used two commercial enzyme-linked immunosorbent assay kits for the
   quantitative assessment of IL-6 (Abcam, ab100713) and IL-1β (Abcam,
   ab197742) following the manufacturer’s instructions. In brief, tissues
   were lysed with a specific buffer contained in each kit, protein
   concentration was determined by the Lowry method, and samples were
   diluted at 1:5 for muscle and 1:4 for the kidney. Supernatants from the
   C2C12 cells treated with palbociclib were obtained after treatment with
   sEVs or PBS and were assessed directly. Absorbance at 450 nm was
   measured in Molecular Devices SPECTRAmax Plus 384. All samples were
   assayed in duplicate.

In vitro model of senescence in muscle progenitor cells

   For the in vitro experiments, we used the cell line C2C12 [American
   Type Culture Collection (ATCC) CRL-1772], a mouse myoblast established
   line. The culture medium used for the expansion and maintenance of the
   cells was high-glucose DMEM with 10% FBS and 1% P/S. Cells were
   maintained at 37°C, with 5% CO[2]. Cells were seeded at a density of
   10,000 cells/cm^2, and 24 hours after seeding, they were treated with 5
   μM palbociclib for 96 hours. Seventy-two hours after seeding, myoblasts
   were treated for 48 hours in an EV-depleted culture medium with
   ADSC-sEVs, 3T3-sEVs, or C2C12-sEVs (5 μg/ml) in PBS or PBS alone. After
   that, cells were collected for further analysis by flow cytometry and
   immunofluorescence.

Flow cytometry analysis of C2C12 cells

   To determine cellular levels of SABG, we used the FluoReporter lacZ
   Flow Cytometry Kit (Invitrogen) following the manufacturer’s
   instructions. In brief, we harvested cells from the cell culture plate
   and diluted them to 1000 cells/μl in a staining medium and introduced a
   1:1 volume of fluorescein di-β-d-galactopyranoside (FDG) 2 mM working
   solution; cells were stained for exactly 1 min at 37°C. FDG loading was
   interrupted by adding 1.8 ml of ice-cold staining medium containing 1.5
   μM propidium iodide. Fluorescence values were read by FACSVerse flow
   cytometry (BD Biosciences, San Diego, CA, USA) until 10,000 events were
   recorded.

   Apoptosis was assessed using an Annexin V-FITC Apoptosis detection kit
   (Immunostep, Salamanca, Spain), following the manufacturer’s
   instructions. ADSCs were harvested and diluted to 1000 cells/μl in
   annexin-binding buffer. Then, 100 μl of aliquots of resuspended cells
   were placed into an appropriate flow cytometer tube and stained with 5
   μl of annexin V–FITC and 5 μl of propidium iodide for exactly 15 min at
   37°C in darkness. After incubation, 400 μl of 1× annexin-binding buffer
   was added. The values were read by FACSVerse flow cytometry (BD
   Biosciences, San Diego, CA, USA) until 10,000 events were recorded.

Immunofluorescence analysis of C2C12 cells

   Cells were fixed with 2% PFA for 10 min at RT and permeabilized with 1%
   Triton X-100 in PBS for 10 min. Cells were blocked with 10% normal goat
   serum (Invitrogen) in PBS-t and incubated with anti-Ki67 antibody
   (anti-Ki67; Thermo Fisher Scientific, PA5-19462; 1:1000 dilution)
   overnight at 4°C. After incubation with primary antibody, sections were
   washed three times with PBS-t and incubated with secondary antibody
   [Alexa Fluor 488 anti-rabbit (Abcam, ab150079; 1:2000 dilution)] for 2
   hours at RT, washes were repeated, and cells were counterstained with
   Hoechst (Invitrogen; 1:1000 dilution) for 30 min at RT. Images were
   acquired with a LEICA DMI 3000-B fluorescence microscope. Images were
   processed in ImageJ, all levels were adjusted equally, and the ratios
   were not altered. A minimum of 200 nuclei were analyzed per sample.

sEV isolation from other cell types

   We used two additional cell lines to test the effect of in vitro sEV
   treatment in senescence. C2C12 and 3T3 (ATCC CRL-1658) cells were
   expanded until they reached 80% confluency in high-glucose DMEM with
   10% FBS; at that moment, medium was changed to high-glucose DMEM with
   2% exosome-depleted FBS (Gibco, A2720803) and 1% P/S. After 48 hours,
   conditioned medium was collected, and sEVs were isolated by
   differential ultracentrifugation using the same method described above
   for the ADSCs.

Epigenetic age estimation using epigenetic clocks

   DNA from different tissues was obtained using the Qiagen DNeasy Blood
   and Tissue Kit, following the manufacturer’s instructions. Raw files
   were processed using the R package SeSaMe version 1.3.0 ([189]72). The
   beta values were obtained using the default sesame procedure with
   nondetection.mask and quality.mask set to FALSE. Differentially
   methylated cytosines were detected using robust linear regression
   implemented in R package limma separately for each tissue. The
   treatment effect was estimated for each CpG by adjusting for potential
   confounding from age, sex, and sentrix array row. Obtained P values
   were corrected for multiple testing using the FDR procedure, and those
   with q < 0.05 were deemed significant. Pathway enrichment analysis was
   performed by annotating the loci to the nearest gene and ranking them
   by signed log P value. The ranked list was passed to g:profiler for
   enrichment analysis using R package gprofiler2 ([190]46).

   Multiple mouse age predictors (epigenetic clocks) were used to estimate
   the age of the animals. The mouse clocks were trained/developed in a
   separate dataset ([191]73). The age predictors were grouped into
   general and intervention clocks. In the general group, the clocks were
   trained on the following mouse tissues across a wide spectrum of
   animals: blood, liver, brain, cerebellum, cortex, fibroblast, heart,
   kidney, muscle, skin, striatum, tail, and a pan tissue clock spanning
   all different tissues. In the intervention group, the clocks were
   trained across the same tissues and animals, but the loci used for
   training were those known to change methylation levels upon known
   age-affecting interventions. The effect of ADSC-sEV treatment was
   estimated by fitting a mixed-effects model for each tissue and group of
   clocks where animal age, sex, and treatment were fixed variables and
   clock name was a random variable. Age predicted by the clock was the
   response variable for the model. The statistical significance of each
   fixed variable was evaluated using Satterthwaite’s method implemented
   in R package lmerTest. Array normalization, array manifest file, array
   annotation files, and epigenetic clock coefficients were obtained from
   S.H.’s laboratory. The methylation data were generated using the
   mammalian array platform (HorvathMammalMethylChip40) ([192]74).

Plasma metabolites by proton NMR spectroscopy

   For each sample, a mixture of 20 μl of blood plasma and 2 μl of
   phosphate buffer with trimethylsilyl propanoic acid and deuterated
   water was transferred into a 1-mm high-quality nuclear magnetic
   resonance (NMR) individual tube. Proton NMR spectra for all samples
   were recorded in a Bruker Avance DRX 600 spectrometer, equipped with a
   triple resonance 1H/13C/31P probe. The nominal temperature of the
   sample was kept at 310 K. A single-pulse presaturation experiment was
   acquired in all samples. The number of transients was 256 collected
   into 65 k data points for all experiments. Spectral chemical shift
   referencing on the alanine CH3 doublet signal at 1.475 parts per
   million (ppm) was performed in all spectra. Spectra were processed
   using MestReNova 8.1 (Mestrelab Research S.L., Spain) and transferred
   to MATLAB (MathWorks, 2012) using in-house scripts for data analysis.
   The chemical shift region including resonances 0.50 to 4.70 ppm (the
   aliphatic region) and 5.20 to 10.00 ppm (the aromatic region) was
   investigated. Metabolite spin systems and resonances were identified by
   literature data and the Chenomx resonances database (Chenomx NMR 7.6).
   Spectra were normalized to the total aliphatic spectral area, lipid
   excluded, to eliminate differences in metabolite total concentration.
   NMR peaks were integrated and quantified using semiautomated in-house
   MATLAB peak-fitting routines. Final metabolite levels were calculated
   in arbitrary units as peak area normalized to the total spectral area.
   Chemometric analysis was performed with PLS ToolBox 8.0 (Eigenvector
   Inc.) in MATLAB. To maximize the separation between samples and to
   identify discriminant patterns, PLS-DA was applied. The permutation
   test was performed to check the overfitting of the PLS-DA models. The
   multivariate chemometric models were cross-validated with 10-fold
   Venetian blind cross-validation. In each run, 10% of the data were left
   out of the training and used to test the model. Spectral regions with
   high VIP coefficients obtained during PLS-DA are more important in
   providing class separation during analysis, while those with very small
   VIP coefficients provide little contribution to classification. An MSEA
   over metabolites with VIP scores higher than 0.5 and P values below
   0.05 was performed with MetaboAnalyst and the SMPDB. MSEA is
   conceptually like gene set enrichment analysis and uses a collection of
   predefined metabolites sets to rank the lists of metabolites obtained
   from metabolomics studies. By using this prior knowledge about
   metabolite sets, we could identify significant and coordinated changes
   in metabolic networks and obtain biological insight.

Total RNA extraction from sEVs and small noncoding RNA expression profiling

   sEVs obtained from different sources were stored in PBS at −80°C; for
   the isolation of RNA, we used the Total Exosome RNA Protein Isolation
   Kit (Thermo Fisher Scientific), following the manufacturer’s
   instructions.

   Small noncoding RNA expression profiling was performed using a GeneChip
   miRNA 4.0 Array (Affymetrix, Santa Clara, CA, USA). Microarray
   experiments were conducted according to the manufacturer’s protocol.
   Briefly, 130 ng of total RNA was labeled with a FlashTag Biotin HSR RNA
   labeling kit from Genisphere. The labeling reaction was hybridized onto
   the miRNA array in an Affymetrix hybridization oven 645 at 48°C for 18
   hours. The arrays were stained using a Fluidics Station 450 with the
   fluidics script FS450_0002 (Affymetrix) and then scanned on a GeneChip
   Scanner 3000 7G (Affymetrix, Santa Clara, CA, USA), using the GeneChip
   Command Console Software supplied by Affymetrix to perform gene
   expression analysis. MiRNA probe outliers were defined and further
   analyzed as per the manufacturer’s instructions (Affymetrix), and
   quality control, as well as data summarization and normalization, was
   carried out using the web-based miRNA QC Tool
   ([193]www.affymetrix.com).

   miRNA expression levels were analyzed with Transcriptome Analysis Suite
   (Thermo Fisher Scientific). Data (.CEL files) were analyzed and
   statistically filtered using Transcriptome Analysis Suite (Thermo
   Fisher Scientific), and input files were normalized with the robust
   multi-array analysis (RMA) for miRNAs. A one-way analysis of variance
   (ANOVA) was performed with the Transcriptome Analysis Suite (Thermo
   Fisher Scientific) on all samples. Statistically significant small
   noncoding RNAs between the different groups and treatments studied were
   identified using an analysis of the variance model with a P value of
   0.01 or less. The imported data were analyzed by PCA to determine the
   significant sources of variability in the data.

miRNA pathway and biological process analysis

   From the up-regulated miRNAs comparing sEVs from young ADSCs to old
   mice’s plasmatic sEVs, the six miRNAs conserved across species were
   analyzed using mirDIP to determine miRNA target genes with very high
   confidence (top 1% in confidence). A total of 5379 unique genes were
   further analyzed by PANTHER to identify enriched biological pathways
   and biological processes with an overrepresentation test; in PANTHER,
   Fisher’s exact test was used with FDR correction. P value < 0.01 was
   considered statistically significant. For the biological process
   analysis, we considered only processes with twofold enrichment or more.

In vitro miRNA mimics treatment

   To test the effect of our selected miRNAs in vitro, we treated
   senescent C2C12 cells using the same culture method as the one
   described above for the in vitro model of senescence. Cells were seeded
   at a density of 10,000 cells/cm^2, and 24 hours after seeding, they
   were treated with 5 μM palbociclib for 96 hours. Seventy-two hours
   after seeding, myoblasts were transfected for 48 hours in an
   EV-depleted culture medium with 1 μM miRNA mimic or small interfering
   RNA control (Qiagen, 1027280) using HIPerFect Transfection Reagent
   (Qiagen, 301704). After that, cells were collected for further analysis
   by flow cytometry. Mimics used: Syn-mmu-miR-125b-5p miScript miRNA
   Mimic (Qiagen, MSY0000136/S0), Syn-mmu-miR-let7c-5p miScript miRNA
   Mimic (Qiagen, MSY0000523/S0), Syn-mmu-miR-214-3p miScript miRNA Mimic
   (Qiagen, MSY0000661/S0), Syn-mmu-miR-145a-5p miScript miRNA Mimic
   (Qiagen, MSY0000157/S0), Syn-mmu-miR-135a-1-3p miScript miRNA Mimic
   (Qiagen, MSY0004531/S0), and Syn-mmu-miR-143-3p miScript miRNA Mimic
   (Qiagen, MSY0000247/S0).

Statistical analysis

   Ratios comparing physical test and plasma values after treatment
   compared to baseline were determined and plotted as % over baseline.
   The baseline was defined as 0%. All groups were tested for the presence
   of outliers with the ROUT method (Q = 2%), and just one data point was
   excluded from the analysis; these data came from a mouse in the PBS
   group that showed about 80% decline in the Treadmill test at day 30.
   Saphiro-Wilk test was conducted in each comparison to test the
   normality of each group. Unpaired Student’s t test or Mann-Whitney test
   was used to calculate the P value for pairwise comparisons. For the in
   vitro experiments with multiple comparisons, ANOVA was used with
   Tukey’s multiple comparisons as a post hoc test; in case of
   nonparametric data, Kruskal-Wallis test was used with Dunn’s multiple
   comparisons as a post hoc test. Each data point presented throughout
   the manuscript refers to biological replicates; technical replicates
   are described in the Materials and Methods. GraphPad Prism 9.0 software
   was used for the analysis and graphical design.

Acknowledgments