The Parabiosis Mouse Experiment

Mice Holding Hands

First Age-Reversal Signs

The causes and mechanisms of aging have been one of the most constant topics that people have studied. One technique that has shown promising results for studying aging is parabiosis, a process in which blood vessels in two animals are joined together. Both animals share a single circulatory system, allowing the study of many complex systemic effects. In 2005, researchers at Stanford University in California released a pioneering study on the effects of parabiosis on muscle tissue regeneration.

The study, “Rejuvenation of aged progenitor cells by exposure to a young systemic environment,” by Ingrid Conboy et al., created pairs of mice via parabiosis. For this study, two types of pairing were created, heterochronic (from hetero – different, and chronic – in time) pairs of young and old mice, and the isochronic (iso – same) control groups, of two aged or two young mice. Mice were paired for 5 weeks before being injured in the hind leg, and the proliferation of new muscle and liver cells were compared in different animals.

The study found that the aged mouse from heterochronic pairs experienced tissue regrowth on a level similar to that of the young mice in heterochronic or isochronic pairs. More specifically, they found increased amounts of eMHC, a protein that is expressed by nascent myotubes formed during muscle regeneration. Nascent myotubes are formed by the activation of satellite cells, a type of stem cell specific to muscle tissue. Pairs of aged mice didn’t demonstrate the same effect, so increased muscle regrowth wasn’t simply a result of parabiosis.

Now, it could be argued that regeneration was the result of new cells from the young mice being transferred through the bloodstream to the aged partner. To prove this wasn’t happening, the young mice used in the experiment were transgenic, and had a gene expressing GFP, or green fluorescent protein, in their cells. This allowed the researchers to determine the source of individual cells by exposing them to UV light. If the cells produced green fluorescence, they were originally from the younger mouse. However, fluorescence was found in less than 0.1% of cells in the aged mice, indicating that old satellite cells were being reactivated, and not just transferred from the younger partner. However, research didn’t stop there.

Next, satellite cells from both young and old mice were grown in the presence of blood serum – blood with all the cells and clotting factors removed. Their findings were twofold – first, that the presence of serum from young mice improved growth of old satellite cells, and secondly, that the presence of old serum could inhibit the growth of young satellite cells. This indicates that aging results in changes in signaling between cells, that “either fails to promote or actively inhibits successful tissue generation,” to use their words.

Finally, the researchers also investigated the effects of parabiosis on liver regeneration. While parabiosis does increase the growth of liver tissue, the rate of liver tissue regrowth in aged mice was significantly higher in heterochronic pairs than isochronic ones. Additionally, the scientists determined that the aged mice in isochronic pairs had increased amounts of a protein complex containing two proteins – Brm and cEBP-α - which has been found to inhibit liver tissue growth. They did this through a process called immunoprecipitation, in which specially generated antibodies are bound to a column. These antibodies bind to specific proteins in a solution and allow them to be isolated. Interestingly, they found that while the levels of the protein complex were increased for aged isochronic pairs, the total amount of Brm and cEBP-α were the same in all animals, regardless of age or the type of pairing they were involved in. This provides more evidence that aging related changes are at least partially due to signals from the organism itself, as some factor must be causing the complex to form.

The findings of Conboy et al, are foundational work in aging related treatment, and their work has been cited more than 1200 times in the last ten years. It demonstrated that aging organisms undergo changes in signaling that affect stem cell growth, and provided researchers with a new tool to determine what those signals were.

Conflicting Findings

Study of the effects of parabiosis on aging didn’t end with the Conboy et al. paper in 2005. One of the authors, Dr. Amy Wagers, left Stanford University to establish a lab at the Harvard Stem Cell Institute. Between 2013 and 2014, that lab published a trio of papers [1-3] about the use of parabiosis to treat symptoms of aging in mice. Dr. Wagers and her colleagues claimed that a decrease in the amount of protein called Growth Differentiation Factor 11, or GDF11, was responsible for a variety of age related effects - from skeletal muscle degeneration to nervous system decline. By joining the circulatory systems of young and old mice, the levels of GDF11 could be increased, reversing or reducing many of these sympto ms. Additionally, treating aged mice with GDF11 seemed to have similar effects as pairing them with younger ones.

Dr. Wagers and colleagues don’t claim that their results reverse or even slow the actual aging process – parabiosis won’t grant immortality. However, it does affect the general ‘frailty’ associated with aging. Known as sarcopenia, the decrease in muscle mass and strength is one of the major causes of disability, loss of independence, and decline in living standards for the elderly.

However, the story gets a little more complicated. Another group of researchers, led by Dr. David Glass at the University of California, who were also studying aging and tissue regeneration felt that Wagers’ research was significantly flawed. Their issue was that GDF11 is part of a group of proteins known as the Transforming Growth Factor Beta (TGF-β) superfamily. TGB-β proteins, such as myostatin, have been previously found to have the exact opposite effect during aging – they slow down damage repair and tissue regeneration. They do this by inhibiting muscle cell differentiation, the process by which stem cells create new muscle tissues.

The main issue was with several experiments similar to the immunoprecipitation experiment used in the Conboy study. Glass’ group thought that the antibodies and another type of protein targeting molecule called an aptamer were not specific enough to distinguish between other TGB-β proteins such as myostatin. Additionally, previous research suggested that GDF11 would have much the same effect on tissue regeneration as myostatin did.

So Glass and colleagues performed their own set of experiments on GDF11 levels in mice. They found that the antibodies and aptamers used by Wagers’ group couldn’t distinguish between myostatin and GDF11 effectively. Additionally, they found that instead of decreasing with age, the levels of GDF11 actually increased, and muscle cell differentiation decreased in the presence of GDF11 instead of increasing! While Dr. Glass’ group contradicted Dr. Wagers’ findings in regards to how GDF11 effects aging in mice, there are a couple of important things to mention. The first is that their experiments focused on the accuracy of measurements of GDF11 levels and the effectiveness of treating aging-related symptoms with GDF11 directly. However, they never attempted to replicate the parabiosis experiments carried out by Wagers’ team, or if they did, they never bothered to publish the results. And Wagers’ group is not the only team studying parabiosis and its effects on aging. It’s also important to note that studies of complex effects and interactions, such as those seen in aging, can be incredibly difficult and contradictory results sometimes occur. While the Glass paper does raise serious doubts about Wagers’ research, especially in regards to the accuracy of GDF11 expression, it shouldn’t be taken as proof of fraud or incompetence in Wagers’ team. Instead, it highlights the incredible number of factors that can affect research and the many ways in which small changes in methods can lead to vastly different results. In the next few weeks, we’ll take a more in depth look at the three papers from Wagers’ laboratory, as well as the replication experiment performed by Glass and his colleagues.

Promising Results From the Heart

As we mentioned previously, the Wagers’ Lab at the Harvard Stem Cell Institute published several papers using parabiosis to investigate the symptoms of aging. The first of those papers, entitled “Growth Differentiation Factor 11 is a Circulating Factor that Reverses Age-Related Cardiac Hypertrophy” by Loffredo et al. specifically investigates how parabiosis can counteract age-related hypertrophy.

Hypertrophy is a process in which an organ becomes larger through the growth of its component cells. More specifically, the individual cells that form the organ become larger, rather than increasing in number. While cardiac hypertrophy isn’t necessarily a bad thing – it’s often observed in response to exercise to increase blood flow through the heart – a specific form of hypertrophy of the left ventricle is common with age. The wall of the left ventricle thickens, reducing the ability of the heart to pump blood effectively; this condition is strongly correlated with risk of heart disease or death.

Dr. Wagers’ lab wanted to determine if cardiac hypertrophy was the result of specific factors in the blood. They performed parabiosis experiments using hetero- and isochronic pairs of female and male mice, much like by Conboy et al,. did to study liver and skeletal muscle proliferation [link to previous post]. They found that the hearts of old mice in heterochronic pairs were approximately 20% smaller than old mice in isochronic pairs, or that were unpaired, and had significantly smaller individual cells.

Dr. Wagers’ group also used genetically distinct mice to determine cell sources, much like the GFP expressing mice used by Conboy et al. Instead of GFP, however, this research used two strains of mice, called CD45.1 and CD45.2, with distinct, alternative forms of a specific gene called ptprc. These two alleles produce identifiably unique but similarly functioning copies of a protein, allowing researchers to determine which mouse in a pair any cell came from.

However, cardiac hypertrophy isn’t solely caused by age – it can also be the result of changes in blood pressure or other effects. This meant that the differences in cardiac hypertrophy could be the result of blood pressure differences after joining the two circulatory systems rather than factors circulating in the blood. Investigation of blood pressure found that young CD45.1 and old CD45.2 mice have similar blood pressures, but young CD45.2 mice have significantly higher blood pressure. To determine if the changes in heart size were the result of difference in blood pressure, measurements of two new groups of mice were prepared. First, the researchers paired young CD45.1 and old CD45.2 mice together and measured their blood pressure over ten weeks. They found the only old heterochronically paired mice showed any significant change in blood pressure in 10 weeks. Additionally, they performed an alternative blood pressure test, which measured mean arterial blood pressure, at the end of the ten weeks and found that there was no difference in blood pressure between any of the mice. However, the reduction in cardiac hypertrophy was still observed in old heterochronically paired mice. Next, they repeated the above experiments, but using pairs of two CD45.2 mice, instead of one CD45.1 and one CD45.2. Once again, they found that the heterochronic parabiosis resulted in reduced heart size in old mice.

After that, the researchers “considered the possibility that physical constraints of parabiotic pairing introduced behavioral changes that contributed to the observed reversal of cardiac hypertrophy.” [ed. note – I think this, and the blood pressure tests, were required by reviewers, which is why they weren’t tested at the same time as the original experiment.]

To test this, they performed pseudo-parabiosis, in which they physically joined two mice, but did not connect their circulatory systems. Unlike the previous experiment, there was no difference in heart or cell size in old isochronic and heterochronic pairs, demonstrating that the changes couldn’t be simply be explained by increased exercise or effort.

In order to investigate what factors were involved in causing cardiac hypertrophy, the lab undertook a large-scale investigation of proteins, metabolites and lipids present in the bloodstream. While the lipid and metabolite screen relied on a more traditional mass spectrometry based analysis to find differences in expression level, the protein analysis used a much newer technology based on aptamers. Aptamers are single stranded DNA or RNA molecules that can be produced to bind to specific targets. After binding to target proteins, unbound aptamers are washed out of a sample and the remaining aptamers are identified and counted to determine the type and quantity of proteins present. While the search for lipids and metabolites failed to find any significant difference between mice in young or heterochronic pairs compared to isochronic ones, they did find that a number of differentially expressed proteins in the serum, including GDF11.

The GDF stands for growth differentiation factor, and the GDF proteins are all members of transforming growth factor beta, or TGF-β, family of proteins. A change in GDF11 concentration is significant because these proteins are known to be involved in cell signaling - especially signaling related to cellular growth.

To confirm the change in GDF11 concentration, they used a Western Blot assay - a very common molecular biology technique that involves separating proteins into bands based on size, then testing for the presence of the protein of interest using antibodies specifically created to bind to that protein. Additionally, they tested for expression of the GDF11 mRNA in a wide variety of tissues, from the eye and brain to the bone marrow, as well as at different ages. They found that t was highly expressed in the spleen in particular, and that the young mice expressed it at much higher levels in the spleen than old mice.

To confirm the effects of GDF11, they injected doses of GDF11 and myostatin into the bodies of mice for 30 days. This resulted in a reduction in heart mass in mice injected with GDF11, much like the parabiotic experiments, that did not occur in mice that were injected with saline, suggesting that GDF11 was an important factor in cardiac hypertrophy. Additionally, after restricting the size of the aorta, a major artery leading from the heart, in young mice before providing a daily dose of GDF11, they found that the GDF11 did not reduce cardiac hypertrophy caused by increased arterial pressure.

These experiments provide solid evidence for the role of aging-related blood factors in the development of cardiac hypertrophy, much like how the Brm/cEBP-α complex described by Conboy et al. prevented proliferation of muscle and liver cells. The reduction of cardiac hypertrophy in heterochronic pairs of mice demonstrates that circulating factors are highly important in the development of aging-related effects.

Additionally, Loffredo et al. claimed to have confirmed the role of GDF11 as a factor preventing hypertrophy. However, as we will see when we discuss the work by the Glass research group, there are some problems with this analysis.

The Role of GDF11

For the second part of the trio, Dr. Wagers’ lab published work they did investigating the role of parabiosis and GDF11 in restoring skeletal muscle.

Work investigating skeletal muscle growth had already been done in the Conboy et al paper, so the paper primarily focuses on the role of GDF11 in muscle skeleton growth rather than parabiosis. The researchers found that mice that underwent heterochronic parabiosis or injection with GDF11 showed fewer breaks in satellite cell DNA, and had satellite cells that were more likely to differentiate into new muscle tissue.

Unlike previous work by Conboy, the researchers had a more direct tool for assessing satellite cells. Satellite cells were marked fluorescently using a specific combination of antibodies. The cells were then sorted using fluorescence activated cell sorting (or FACS), which determines the fluorescence of individual cells. During FACS, each cell is moved past the detector in an individual droplet. If a cell has a particular fluorescence, a charge is applied to it, and a charged deflector plate is used to sort the cells into different samples.

This meant that satellite cells could be sort and counted, allowing scientists to determine how many were present in a particular muscle sample after treatment of the mouse with GDF11 or heterochronic parabiosis. Additionally, the integrity of the DNA of satellite cells could also be confirmed, by visualizing the number of breaks present in it. Two methods were used to determine the integrity of the DNA. The number of breaks in the DNA strand was first visualized by subjecting fixed cells to an electric field, which causes DNA to move.

Smaller strands of DNA move faster, and thus the amount of spread is an indicator of the number of breaks in the DNA in a cell. Additionally, an antibody was used to stain cells for a specific type of protein, the pH2AX histone, which indicates the presence of breaks in DNA.

The paper also investigates the effects of GDF11 injections on formation of new muscle tissue, and the ability of the mice to heal. They viewed tissue samples to determine how many new myofibers formed, and what size they were. Additionally, they transplanted satellite cells expressing GFP into mice being inject with GDF11, which allowed them to determine how many new myofibres were being incorporated with and without the injects of GDF11. The physical abilities of aged mice were also tested, by measuring their grip strength, the length of exercise they could perform, and the amount of glucose being used during a given period of exercise. All of the tests turned out positive for mice treated with GDF11, suggesting that it was significantly improving the physical condition of aged mice.

All this evidence taken together strongly suggests that GDF11 plays a significant role in the regulation of muscle growth and the symptoms of aging. However, as we will see soon, not everyone agreed with the findings of Wagers’ Lab.

Replication Issues

In 2015, a paper was published that attempted to replicate many of the Wagers’ labs finding. Called “GDF11 Increases with Age and Inhibits Skeletal Muscle Regeneration” by Egerman et al, the work was primarily performed in the laboratory of David Glass. Egermen et al. did not think that the findings of the Wagers’ lab were credible, mainly because it strongly contradicted what was known about the TGF-β family. Previous work all suggested that GDF11, like the other members of the TGF-β family, was actually an inhibitor of muscle differentiation. Additionally, GDF11 is highly similar to myostatin, which had been well studied, and both proteins acted the same pathways and cellular receptors. Mice which did not have a functional myostatin gene nearly doubled in muscle mass, and inhibiting the rest of TGF-β family of proteins increased observed muscle growth even further.

While the Egerman paper attempted to replicate several experiments by the Wagers’ lab, the most important, and perhaps most damning, was their investigation into levels of GDF11 found in the blood over time.

First, the Glass group investigated the aptamer technology initially used to detect GDF11. An aptamer for GDF11 was synthesized and exposed to different concentrations of pure myostatin and GDF11. They found that the aptamer bound both proteins, not just GDF11, though it was less specific for myostatin. This suggested that the changes in GDF11 concentration Loffredo et al [2] found may be incorrect, as they would have been detecting both myostatin and GDF11.

However, the attempt to replicate the aptamer test has some issues. While Glasses team used an individually synthesized aptamer, the Loffredo paper had performed their test with a SOMAscan assay, which uses ~1000 aptamers to test for concentrations of different proteins, including myostatin and GDF11, simultaneously. There is no indication that the identical aptamer was used to test for GDF11 in both cases, so the specificity test may not be comparable to the Wagers’ group test. Additionally, the presence of the myostatin aptamer might have nullified any possible lack of specificity in the GDF11 aptamer. While their findings do throw doubt on the Loffredo paper finding, it would have been more compelling to add known concentrations of GDF11 and myostation to a sample and determined if that had any effect on the detected concentration of GDF11 using a SOMAscan assay.

However, the Loffredo paper had also used a western blot to confirm the changes in GDF11 over time. The Glass group investigated the antibody used for the Loffredo paper, to see if they found any different results. This time, the antibodies used were explicitly the same as those used by the Wagers’ lab team. Not only did they find that the antibody lacked specificity and could not distinguish between myostatin and GDF11, they also found that GDF11 and myostatin were present in dimer form on the western blot. To understand the significance of this, we must first discuss how a western blot is performed.

The Role of the Western Blot

A western blot is a two-part process. The first is the gel electrophoresis, in which proteins are moved through a polyacrylamide gel matrix by an electric field. Proteins and nucleic acids are pulled toward the negatively charged side of the field. The polyacrylamide gel slows the movement of the protein or nucleic acid based on size, as smaller molecules move more easily through the pores in the gel. Over time, similarly sized molecules separate into distinct bands based on size, with the smallest molecules moving the farthest. Once the proteins have been separated by size, they are transferred to a cellulose sheet and stained with an antibody. The antibody binds to one specific protein (ideally, though as Dr. Glass’ lab proved, not always) allowing researchers to identify the presence of the protein. Additionally, by running different samples though the gel at the same time and comparing the amount of antibody that binds to each, the relative amounts of a protein can be compared.

However, proteins can form dimers, in which two copies of a protein bond together effectively doubling their size. While the Loffredo paper looked at the amount of GDF11 present at the band expected for a single copy, or monomer, of the protein, Egerman et al. found that significant amounts of GDF11 could also be found in the dimer band weight and above. This means that, in addition to confounding effect of myostatin binding to the antibody, a significant portion of the protein present in the samples might have been missed by Wagers’ group.

To confirm this, the researchers replicated the western blot using the same antibody as Wagers’ group, but looked for antibody staining over the whole range of band sizes. They found that while the amount of GDF11/myostatin did decrease with age in monomer form, the amount of protein present in the dimer band range increased with age, and at a far greater rate.

Finally, they used a new antibody, that did not bind myostatin, to detect the levels of GDF11 in human and rat blood. The antibody was not sensitive enough to detect GDF11 levels in mice blood, but they did confirm that levels increased with age in humans and rats. While testing in mice would be preferred for a direct evidence of GDF11 increase with age, rat and human systems are a good proxy, especially in light of the other evidence the Glass lab accumulated.

The Glass laboratory replication attempt covered far more than just GDF11 concentrations over time. The next section will look at some of the experiments they performed in an attempt to replicate the rest of the Wagers’ lab results.

Egerman's Refutation

For the second part of the Egerman paper, we’ll be looking at the effects of putative effects of GDF11 on muscle cells and differentiation as investigated by the Glass lab.

After considering the concentration of GDF11 over time, the Egerman paper moves on to the signaling networks invoked by myostatin and GDF11. They demonstrated that myostatin and GF11 both invoked cellular signaling through the SMAD2/3 pathways. Human skeletal muscle-derived cells (hSkMDCs) showed an increase, via western blot, in SMAD2 and SMAD3 proteins after exposure to both myostatin and GDF11. Additionally, a segment of DNA was inserted into the cells that was designed to produce a fluorescent protein if the SMAD2/3 pathway was activated. Protein fluorescence was observed in cells exposed to both GDF11 and myostatin.

Additionally, gene regulation changes were produced by exposure to GDF11 or myostatin. A device called a microarray was used to determine how much RNA was produced by human muscle tissue. Analysis of the microarray data showed that GDF11 and myostatin produced almost identical gene expression changes.

However, while the Glass work on signaling pathways and gene regulation is useful, there are a couple caveats. The first is that it looks at the effects of GDF11 on isolated cell cultures, or as it’s referred to in scientific literature, in vitro, not the effects in an organism as a whole, in vivo. Studying cells in their normal biological context, inside an organism in this case, can cause significant differences in the results of a study. In this case, the whole purpose of the Wagers’ lab work was to study circulating systemic factors and their effect of the organism. The Glass lab choice to focus mostly on cell culture, while simpler, does cloud the issue.

Additionally, while Glass’s investigation of cell signaling and gene changes does suggest that GDF11 and myostatin have similar functions, we know for certain they don’t have identical functions. This is simply due to the fact the mice that myostatin null mice, i.e. mice that can’t produce myostatin, are known to be physically different from GDF11 null mice. So regardless of the similarities in their signaling pathway, there must be some difference in the way they interact with the organism as a whole.

Finally, the researchers looked at the effects of GDF11 and myostatin on muscle cell differentiation. They did this using a variety of techniques. The first was to using antibody staining against myoblasts formed from hSkMDCs, to determine how myotubes were forming, much like Conboy et al did in 2005. They found that GDF11 has inhibitory effects on differentiation, though not to the same degree as myostatin, and it doesn’t seem to increase with dosage, as with myostatin.

After that, they injected both old (23 month) and young (6 month) mice with GDF11 in a similar regime to the Sinha et al paper [3], though the younger mice received a much higher dosage. These in vivo test found nearly the opposite effect as Dr. Wagers’ research, with no change in muscle regeneration in older mice, and a reduction in healing rate in young mice, when compared to the control group not receiving injections.

Finally, they used flow cytometry to isolate satellite cells and single muscle fibres from adult and aged muscle tissue in mice, and grew them in culture containing GDF11 for several days. They found that exposure to GDF11 significantly reduced the growth of satellite cells, and prevented the formation of new muscle tissue. However, unlike the study be Sinha et al, they exposed isolated cells to GDF11 after sorting, rather that sorting them from animals that had been exposed to a regime of GDF11. This again brings up the issue of of in vitro and in vivo testing. Were the differences in the satellite cell response observed by the two groups due to incorrect methods or error, or simply due to the differences in the experiment?

The findings of Egerman et al. on the effect of GDF11 exposure on muscle cell growth aren’t as conclusive as the ones on GDF11 concentration we discussed previously. Even on its own, however, the fact that GDF11 doesn’t decrease with age is a significant issue for much of the Wagers lab research. If GDF11 levels increase with age, then simple injections shouldn’t be an effective treatment for cardiac hypertrophy and reduced muscle regeneration. And while the Glass laboratory didn’t seek to replicate the parabiosis results, the issues identified with the rest of the papers do raise issue with the validity of the results from the Wagers’ lab.


Conboy, I. M., Conboy, M. J., Wagers, A. J., Girma, E. R., Weissman, I. L., & Rando, T. A. (2005).

Rejuvenation of aged progenitor cells by exposure to a young systemic environment. Nature, 433(7027), 760–764.

Katsimpardi, L., Litterman, N. K., Schein, P. A., Miller, C. M., Loffredo, F. S., Wojtkiewicz, G. R., … Rubin, L. L. (2014).

Vascular and Neurogenic Rejuvenation of the Aging Mouse Brain by Young Systemic Factors. Science, 344(6184), 630–634.

Loffredo, F. S., Steinhauser, M. L., Jay, S. M., Gannon, J., Pancoast, J. R., Yalamanchi, P., … Lee, R. T. (2013).

Growth Differentiation Factor 11 Is a Circulating Factor that Reverses Age-Related Cardiac Hypertrophy. Cell, 153(4), 828–839.

Sinha, M., Jang, Y. C., Oh, J., Khong, D., Wu, E. Y., Manohar, R., … Wagers, A. J. (2014).

Restoring systemic GDF11 levels reverses age-related dysfunction in mouse skeletal muscle. Science (New York, N.Y.), 344(6184), 649–652.

Egerman, M. A., Cadena, S. M., Gilbert, J. A., Meyer, A., Nelson, H. N., Swalley, S. E., … Glass, D. J. (2015).

GDF11 Increases with Age and Inhibits Skeletal Muscle Regeneration. Cell Metabolism, 22(1), 164–174.

Levy, D., Garrison, R. J., Savage, D. D., Kannel, W. B., & Castelli, W. P. (1990).

Prognostic implications of echocardiographically determined left ventricular mass in the Framingham Heart Study. The New England Journal of Medicine, 322(22), 1561–1566.