induced to differentiate in a cell-culture dish. In addition, removing P3 cells from the ASPC pool improved the ability of the other cells in the dish to differentiate into adipocytes.

These data suggest that the low-abundance P3 cells inhibit adipogenesis. Schwalie and col-leagues named these cells adipogenesis regula-tors (Aregs), and confirmed the cells’ function in vivo. First, they transplanted two mixtures of ASPCs — one lacking Aregs, the other con-taining the entire cell pool — into mice. Each mouse received both mixtures, one on each side of the body. Next, the researchers fed the mice a high-fat diet to induce adipogenesis. Over a few weeks, the implanted cell mixture lacking Aregs grew many more adipo cytes than the other mixture did, indicating that Aregs inhibit fat growth. The authors also showed that Aregs exist in human fat, imply-ing that fat-development mechanisms are conserved between mice and humans.

How do Aregs inhibit adipogenesis? Schwalie et al. found that the cells reside near the blood vessels of fat tissues in mice, a loca-tion that was previously proposed as the site of adipocyte precursor cells8 (Fig. 1). Next, the authors investigated whether Aregs signal to neighbouring cells through physical contact or by sending chemical (paracrine) signals to nearby cells. Co-culture experiments, in which the authors placed a barrier permeable to small molecules between the Aregs and their target cells, revealed that direct contact is not needed for Aregs to influence fat-cell formation, indicating that the signal is paracrine.

To identify candidate signalling molecules, the researchers inactivated genes that are highly expressed in Aregs. They found that the gene Rtp3 needed to be turned on to enable Aregs to send their inhibitory signals. Little is known about the Rtp3 protein, and it is not obvious how it works in this context. This is an area ripe for future study, because modu-lating the signals released by Aregs could have therapeutic potential for controlling fat growth.

Schwalie and colleagues’ findings are exciting for several reasons. First, although high variation between ASPC subpopula-tions had been predicted, this study fills a major gap by adding molecular details to our understanding of that variability. Second, the authors use state-of-the-art technology for single-cell gene-expression profiling, enabling them to identify a regulatory cell type that would have been difficult to predict on the basis of previous studies. It is to be hoped that this study will stimulate other work aimed at elucidating the organization of adipogenesis (the hierarchy of cells that regulate the forma-tion of fat), as has been achieved for blood-cell lineages9.

The current study adds to the mounting evidence that paracrine signals help to remodel stem- and progenitor-cell function10, and opens up several avenues for future research. For instance, what is the anti-adipogenic

signal, and how does Rtp3 help to stimulate Aregs to produce it? Genetic or age-related differences in Areg number or function might contribute to body-fat pattern ing or the propensity to become obese, and these possibilities should also be explored.

It will be interesting to determine whether the ASPC pool can be divided into further subpopulations with more-specific functions. Tracing the in vivo fates of these different subpopulations would be a powerful strategy for picking apart which cells become adipo-cytes and which become fat-supporting cells. Finally, perhaps one of the most interesting questions raised by the study is whether there is a true adult adipocyte stem cell, which, by definition, would be capable of producing both committed adipocyte progenitors and more adipocyte stem cells.

As the devastating human costs of obesity-related conditions rise, research and health-care professionals must meet the challenge with breakthroughs in medical manage ment and care. This will require a

better understanding of adipogenesis, and Schwalie and colleagues’ work has pointed to a new way of advancing knowledge in this important area. ■

David A. Guertin is at the University of Massachusetts Medical School, Worcester, Massachusetts 01605, USA. e-mail: david.guertin@umassmed.edu

1. Schwalie, P. C. et al. Nature 559, 103–108 (2018).2. Sanchez-Gurmaches, J., Hung, C.-M. & Guertin, D. A.

Trends Cell Biol. 26, 313–326 (2016).3. Lynes, M. D. & Tseng, Y.-H. Ann. NY Acad. Sci. 1411,

5–20 (2018).4. Schoettl, T., Fischer, I. P. & Ussar, S. J. Exp. Biol. 221,

jeb162958 (2018).5. Rosen, E. D. & Spiegelman, B. M. Cell 156, 20–44

(2014).6. Hepler, C., Vishvanath, L. & Gupta, R. K. Genes Dev.

31, 127–140 (2017).7. Berry, R., Jeffery, E. & Rodeheffer, M. S. Cell Metab.

19, 8–20 (2014).8. Tang, W. et al. Science 322, 583–586 (2008).9. Orkin, S. H. & Zon, L. I. Cell 132, 631–644 (2008).10. Kusuma, G. D., Carthew, J., Lim, R. & Frith, J. E. Stem

Cells Dev. 26, 617–631 (2017).

This article was published online on 20 June 2018.

T I S S U E E N G I N E E R I N G

Living replacement heart valves remodelledBioengineered heart valves are a promising treatment for heart-valve disease, but often undergo mechanical failure when implanted. Computational modelling of the initial valve design has now improved their performance in sheep.

C R A I G A . S I M M O N S

Heart-valve disease has been described as an emerging epidemic1, owing to its worldwide prevalence, its potential to

kill, and the lack of therapies for its prevention or treatment. Damaged and defective valves can be replaced with prosthetic ones, but the inability of prostheses to grow or adapt to change makes them a poor solution for young patients2. An alternative is a living replace-ment made from bioengineered tissue. But, so far, tissue-engineered heart valves (TEHVs) have failed because detrimental valve-tissue remodel ling occurs in vivo, impairing normal function. Writing in Science Translational Med-icine, Emmert et al.3 address this problem using computational modelling to design a TEHV that remodels favourably after implantation.

A promising strategy for heart-valve tissue engineering is to grow tissues in the shape of a valve in the laboratory using cells and a degradable biomaterial, then remove the cells to leave an empty extracellular-matrix scaffold. After implantation in the heart, the tissue scaffold is populated by the recipient’s cells, presumably from the blood and the adjacent

blood-vessel wall. Some of these cells then transform into contractile cells that degrade the scaffold and replace it with new tissue, while pulling on the tissues to hold everything together, and speed up remodelling. In all ani-mal studies so far, however, this process has led to excessive tissue production, which thickens and stiffens the valve’s three leaflets to obstruct blood flow, or to excessive tissue contraction, which causes the leaflets to shorten and retract, preventing proper valve closure and allowing backflow of blood (Fig. 1a).

After valve repair in healthy hearts, the contractile cells typically enter a deactivated state called quiescence, or are cleared from the region by programmed cell death. But certain stimuli, such as biomechanical stresses4, can cause the cells to persist and remain activated. This can lead to excessive collagen production and tissue contraction — a process known as fibrosis4. In the native valve, fibrosis leads to leaflet thickening or retraction, which is similar to the problem seen in TEHVs. Bio-mechanical activation of fibrosis can be a particular problem in heart valves that are subjected to abnormalities in stretch, com-pression and pressure changes as they open

4 2 | N A T U R E | V O L 5 5 9 | 5 J U L Y 2 0 1 8

NEWS & VIEWSRESEARCH

© 2018

Springer

Nature

Limited.

All

rights

reserved. ©

2018

Springer

Nature

Limited.

All

rights

reserved.

and close with each heartbeat5. Using computational modelling, the

group behind the current study have previ-ously shown6 that the leaflets of conventional TEHVs are compressed in the radial direc-tion (towards the centre of the valve) by blood pressure when the valve is closed. This type of compression, which does not occur in native valves, acts to shorten the leaflets. It is also associated with contractile-cell activation and fibrotic remodelling.

One way to address this problem would be to directly inhibit fibrosis. But Emmert et al. took an alternative approach, attempting to limit leaflet shortening by minimizing radial compressions, and to guide the remodelling process to prevent persistent contractile-cell activation. The authors used computational modelling to design TEHVs that initially had a non-physiological geometry predicted to minimize radial compression, and a large area of contact between the leaflets when the valve was closed. The computational models predicted that, after implantation, these valves would remodel into a stable geometry that mimicked the shape of native valves (Fig. 1b).

The authors tested this hypothesis by replacing heart valves with computationally inspired TEHVs in ten sheep. They assessed the valves’ performance over one year. As the com-putational models predicted, the valves adopted a stable tissue architecture in vivo. The valves performed comparably to native valves in nine out of ten animals. The recipient cells that popu-lated the TEHVs produced new collagen with-out leaflet thickening. Collagen fibres aligned around the circumference of the valve, although not to the same extent as in native valves. The leaflets shortened as the valves remodelled, but shortening stabilized six months after implanta-tion and the leaflets remained in contact with one another, so valve function was not affected. Finally, although many recipient cells infiltrated the valves, few of these were activated contrac-tile cells after one year. Together, these results demonstrate that in vivo TEHV remodelling can be guided towards stable physiological structure and function.

Perhaps the most striking aspect of this study is that the authors’ predictive-modelling approach was successful even though they could not control the cell types that infiltrated the TEHVs. But although this approach worked well in healthy sheep, the cells that repopulate the tissue scaffold in people who have valve diseases or defects might respond differently to mechanical stimuli. Indeed, the authors’ computational model predicted that the TEHVs would work best when cell con-tractility was low, which may not always be the case in humans. For example, contractile-cell activation is driven by inflammation and immune responses7 that were largely absent in this study, but are often elevated in people with valve disease. In this context, excessive contractile-cell activation could lead to detri-mental fibrosis and TEHV failure.

Moving forward, it will therefore be important to determine the robustness of a strategy that involves computational mod-elling based solely on mechanical influences, given the diversity of pro-fibrotic stimuli that can derail remodelling in patients. Successful TEHV solutions will probably need to com-bine computationally guided remodelling with complementary strategies to control detri-mental fibrosis. For example, native valve cells express anti-fibrotic factors8 that could poten-tially be delivered alongside a TEHV to help suppress adverse remodelling, particularly in the early post-implantation phase, when inflammation might be heightened.

It is also notable that the remodelled valves showed little development of the trilayered microstructure of native valves, in which the top, middle and underlying sections of each leaflet have different compositions and mechanical properties. This structure is thought to be essential to native-valve mechan-ics5. Longer-term follow-up will be necessary to determine whether Emmert and colleagues’ TEHVs undergo further remodelling after 12 months to produce this microstructure, as has been observed in other TEHVs9, or if normal function can be maintained without it. Longer-term follow-up will also reveal whether the recipient cells in the TEHVs — which had quiesced and stopped making tissue after one year — can be reactivated to make new tissue and allow the valve to grow, as is required for children.

Ultimately, TEHV cells must be activated to grow and repair as necessary, but then quiesce to prevent fibrosis. Emmert and colleagues’ study clearly demonstrates the potential of a computational strategy to design TEHVs that can achieve this delicate balance on the basis of predicted mechanical and biological outcomes. The work also argues for further development of this approach to account for other factors in remodelling that could be predictably guided. ■

Craig A. Simmons is in the Translational Biology & Engineering Program at the Ted Rogers Centre for Heart Research, and in the Department of Mechanical & Industrial Engineering and Institute of Biomaterials & Biomedical Engineering, University of Toronto, Toronto M5G 1M1, Canada.e-mail: c.simmons@utoronto.ca

1. d’Arcy, J. L. et al. Eur. Heart J. 37, 3515–3522 (2016).2. Parvin Nejad, S., Blaser, M. C., Santerre, J. P.,

Caldarone, C. A. & Simmons, C. A. Adv. Drug Deliv. Rev. 96, 161–175 (2016).

3. Emmert, M. Y. et al. Sci. Transl. Med. 10, eaan4587 (2018).

4. Tomasek, J. J., Gabbiani, G., Hinz, B., Chaponnier, C. & Brown, R. A. Nature Rev. Mol. Cell Biol. 3, 349–363 (2002).

5. Sacks, M. S., Merryman, W. D. & Schmidt, D. E. J. Biomech. 42, 1804–1824 (2009).

6. Loerakker, S., Argento, G., Oomens, C. W. J. & Baaijens, F. P. T. J. Biomech. 46, 1792–1800 (2013).

7. Van Linthout, S., Miteva, K. & Tschöpe, C. Cardiovasc. Res. 102, 258–269 (2014).

8. Blaser, M. C. et al. Circ. Res. 122, 405–416 (2018).9. Hoerstrup, S. P. et al. Circulation 102 (Suppl. III),

III44–III49 (2000).

Figure 1 | Improving tissue-engineered heart valves using computational modelling. Heart valves can be replaced with alternatives made of living tissue. a, In conventional tissue-engineered heart valves (TEHVs), leaflets make small contacts with one another. But when these TEHVs are implanted into sheep hearts to replace the native pulmonary valve, they undergo adverse remodelling changes after 12 months. The leaflets become thicker and retract, leaving a hole through which blood leaks. b, Emmert et al.3 have used computational modelling to improve the initial valve design. Their modelling predicted that a valve shape that minimized shortening of the leaflets under pressure (not depicted), and had large initial contact areas between leaflets, would remodel more favourably after implantation. Indeed, 12 months later, valve geometry and function were stable and comparable to that of native valves.

Pulmonary valve

Lea�et

Implant

Sheep heart

12 months

Conventional TEHV Adverseremodelling

Functionalremodelling

Computationally inspired TEHV

a

b

5 J U L Y 2 0 1 8 | V O L 5 5 9 | N A T U R E | 4 3

NEWS & VIEWS RESEARCH

© 2018

Springer

Nature

Limited.

All

rights

reserved. ©

2018

Springer

Nature

Limited.

All

rights

reserved.