Beta living through stem cells

Insulin-producing cells will be tested first in patients lacking a pancreas.


Diabetes is way complex. “But it’s a simple disease conceptually—your body doesn’t produce enough insulin,” notes Joslin Diabetes Center researcher Gordon Weir.

In type 1 diabetes, an autoimmune attack wipes out insulin-producing beta cells, which are found in clusters of pancreatic cells called islets. In type 2 diabetes, the beta cells are still there but not hauling all the freight. That disease can be treated with many other types of drugs, along with lifestyle changes. But over time, beta cells wear out. In fact, more people with type 2 take insulin than people with type 1.

And there’s no way to make insulin injections pleasant or easily controllable or as good as insulin production by beta cells.

Thus the huge interest in a long-term research project spearheaded by Harvard’s Doug Melton to create working beta cells by manipulating stem cells. An update on the ambitious project from Melton, Weir and other partners drew a crowd at Harvard on Monday.

Making insulin-producing cells good enough for clinical trials “turns out to be rather difficult; it took more than a decade,” Melton said. “We haven’t made it really perfect, but it’s at the goal line.”

Technology from Melton’s lab has been licensed exclusively to the startup Semma Therapeutics, which is joining with Joslin, Brigham & Women’s Hospital and Dana-Farber Cancer Institute to move toward clinical trials. Traveling under the ungainly title of the Boston Autologous Islet Replacement Therapy Program (BAIRT), the collaboration launched in June.

The first BAIRT studies, starting at least three years from now, will not be among people with type 1 diabetes. Instead, they will recruit people who have had their pancreases removed, usually because of uncontrollable pain after the organs are chronically inflamed by years of heavy drinking.

This approach bypasses the biggest problem in cell treatments for type 1 diabetes: the body renews its autoimmune attack and wipes out the newly introduced cells. “We decided to solve one problem at a time,” Melton explained.

Patients who have prostatectomies often now are given islet cells salvaged from their own pancreas, which helps to improve their diabetes control, but those cells may themselves be damaged or in short supply, said Brigham surgeon Sayeed Malek. Transplants of brand-new beta cells, made from the patients’ own blood, should help.

These reengineered cells will be injected in the arm, where they will be easy to monitor  and to remove if necessary, said Semma CEO Robert Millman. Decades of experience transplanting cells from cadavers has shown that “you can put beta cells just about anywhere,” Weir added.

Against autoimmunity. If all goes well, the project will continue into trials for type 1 diabetes with non-personalized beta cells, where the autoimmune attack will be blunted via encapsulating the cells. Seema is spending about half its budget on encapsulation technologies, Millman said.

Encapsulation is the near-term solution to fend off the autoimmune attack. “The long-term solution is to use the power of biology to understand why the immune system has made this mistake,” Melton remarked.

He briefly mentioned two promising research thrusts. One effort is to learn from the rapid advances in knowledge about how cancer cells dodge the immune system.

Another, led by Chad Cowan of Massachusetts General Hospital, aims to create a “universal donor pluripotent stem cell.” Missing all the billboard signs that alert immune enforcers, these cells could play a role like that of O-positive cells in blood transfusions.

Asked about his own take on the causes of type 1, Melton mentioned one theory that the autoimmune attack may be triggered by gut cells that naturally produce insulin or similar substances under certain conditions.

Slow and steady. Bringing beta cell therapies to the clinic will be a marathon march with not only many scientific steps but many regulatory steps. Millman emphasized, however, that “the FDA is working with us very early on the regulatory path.”

Among potential safety risks, all stem cell therapies must be carefully vetted to avoid the growth of teratomas—tumors with a jumbled mix of cells, usually benign. These cellular junk piles would be relatively easy to remove, but much better to avoid altogether, Millman said.

Another concern is that the cells will secrete insulin even when it’s not needed, dropping the recipient’s blood sugar levels to dangerously low levels.

There also is much cause for worry that the cells won’t last long, a major problem in transplants of cadaver beta cells. However, built-from-scratch cells function “for more than a year in mice, which bodes well for people,” Weir commented. And Millman pointed out that the cells resemble juvenile cells, which may help them withstand the high stresses of transplantation better than worn-out adult beta cells do. “We hope these almost pristine cells going into the patients will last a lot longer,” he said.

None of this will come cheap. Asked about pricing for cell therapies, way down the road when and if they hit the market, Millman was understandably wary. Initial costs for these treatments will be very high, accompanied by very close regulatory scrutiny. Semma has raised about $50 million, but “we need philanthropy and we need institutions to support this,” he said.

Melton suggested, though, that successful cell-based therapies will make complete  economic sense, given the soaring numbers of people with diabetes and the huge costs of diabetes care. Each year the world spends about $30 billion on insulin alone. “Diabetes is not an orphan disease,” he said. “The cost will come down very quickly.”

Man petabytes dog


One of the earliest stories I wrote about genomics past the gee-whiz aspects of the Human Genome Project covered the first whole-genome sequencing of a dog. Kerstin Lindblad-Toh of the Broad Institute patiently explained the project to me, and scientists who used dog models to study inherited blindness told me why they were more than excited about the prospects.

More than a decade later as I’m putting together a special report on big data for Nature, the genomic revolution has marched ahead, well, much as predicted.

The cost of genomic sequencing has dropped arguably faster than any other technology in human history. Research initiatives that most of us haven’t heard about are gathering genomic data on hundreds of thousands of people. This flood of data is multiplied with data from proteomics and other omics now scaling up to the genomic scale. We talk casually about petabytes (millions of gigabytes). Data scientists, many of them coming in from fields outside biology, are integrating these data and making some astonishingly good predictions about what drugs might work for a given condition, without needing any new wet-lab work. We’ve seen wonderful progress in stem cells and cellular models and genetic engineering tools. And this revolution is on television, also websites, social media and an entirely sufficient plenitude of TED talks.

Not so much in the clinic, though.

Of course omic research on many diseases is starting to pay off for actual patients—for example, The Cancer Genome Atlas has spun off clues for real advances in many cancers—and its grand march points straight ahead through enormous but movable objects.

But clinical steps are slow. Part of the reason is the sheer complexity of disease, for instance the ways cancers duck and weave to dodge treatments. And, of course, clinical trials can’t be rushed.

Last week I asked one neuroscientist why we still lack drugs that treat the causes of neurodegenerative diseases, as opposed to their symptoms. She responded, reasonably enough, that it takes years to build better lab models of the disease and push findings from those models into the long tunnel of pre-clinical work toward trials. She expected that some of the compounds coming from her work will help. She didn’t predict home runs.

But we haven’t lost the gee-whiz discoveries and our faith that they’ll end up in the clinic in our lifetimes. My favorite: Scientists can take a human skin cell, bombard it with select small molecules until it morphs into a reasonable facsimile of an insulin–producing cell (a notoriously fickle beast) and produce such cells in the millions. Maybe those cells will arrive in the next decade, bringing actual cures. And although I don’t follow discoveries in dog proteomics, I see that University of California/Berkeley researchers have restored vision to blind dogs via genetic therapy. Progress, yes. Dogged research!


Capping off

sphere_scarringAn MIT alginate microcapsule holding islet cells (in green) and being covered by immune cells (in blue and magenta). Image credit: Omid Veiseh, Joshua Doloff, Minglin Ma and Arturo Vegas.

There’s a worldwide deficit in insulin-producing beta cells, for people with either type 1 or type 2 diabetes, Harvard’s Doug Melton told a session at the ADA annual scientific conference on Friday.

“It’s a completely non-trivial thing that you can now make billions of human beta cells,” he said. “We spent more than a decade trying to march these cells through this procedure.”

Currently, it takes his lab about 40 days to produce the cells at a cost of about $6,000 per flask, but Melton is confident that these numbers can be chopped down.

The achievement required not only brilliant scientific detective legwork, especially on the last steps of differentiation, but lab drudgework on a dramatic scale.

Picking apart the steps that drive cells into beta shape, “we had to sort out three or four factors at a time,” he noted. The lab ran screens of small molecules to find what combinations were effective. Testing eight small molecules, in three concentrations, for different periods, in triplicate meant about 65,000 combinations to examine per screen.

The lab of MIT bioengineer Dan Anderson, collaborating with Melton to build microcapsules for the beta cells, took high-volume testing to a much higher level for various capsule designs.

Over the decades, many groups have tried to encapsulate beta cells in tiny spheres of alginate. Historically, “all these capsules end up covered in scar tissue,” Anderson told the ADA session.

But after endlessly tweaking the properties of these spheres, “we have a growing list of materials we could use,” he said.

One capsule material seems to work well in mice with strong immune systems—and in very early testing in macaques. Details on the material aren’t yet public, but the secret isn’t in the material’s permeability but in how the immune system reacts to it, Anderson said.

His group’s exhaustive testing also gave clues to how capsule size affects immune scarring. Last month, he and colleagues reported in Nature Materials that 1.5-mm-diameter capsules do better than 0.5-mm structures. Was that a surprise? “It was for us,” Anderson replied. “We thought smaller would be better.”

Beta than the real thing?

Melton_ beta_cellsHuman stem cells implanted successfully in a mouse. Image courtesy Doug Melton.

We know a true cure for type 1 diabetes will require both a new supply of insulin-producing beta cells and a new way to stop the autoimmune attack that wipes out the original cells. We’ve seen great progress in the past decade on the first challenge, as researchers have learned to morph embryonic stem cells and then normal skin cells into beta cells that now might be very much like the real thing. But despite all we’ve learned about the autoimmune attack and all the clever ideas that have emerged to stop it, so far there’s no clear way to do so.

Today the best bet for autoimmune defense is to embed engineered beta cells in intricately designed porous capsules. While most attempts use spherical capsules on a millimeter scale, last fall Viacyte launched a clinical trial for a device the size of a credit card. We’re all rooting for this trial’s success. But even if it works well, the encapsulated cells won’t function quite normally or last forever.

“I want a forever solution,” says Harvard’s Doug Melton, who has led much of the stem cell engineering work.

At a JDRF session in Boston back in March, Melton suggested two approaches to further modify those engineered cells to dodge the autoimmune bullets. “These are two of my favorite ideas, but I’ll remind you that most of my ideas turn out to be wrong,” he said wryly.

One idea follows the playbook of the hot new class of cancer immunotherapies known as “checkpoint blockaders” —or rather turns it on its head.

As cancer researchers began to discover two decades ago, T cells that charge in to wipe out tumor cells are stopped in their tracks if the tumor cells express certain proteins on their surfaces. Well, how about engineering beta cells to defend themselves by expressing these proteins on their surfaces? (Work in mice by other labs indeed has demonstrated protective effects.)

Melton prefers a second concept, which taps into one of the large questions of immunology: Why doesn’t a mother’s immune system attack the fetus she carries?

Cells called trophoblasts that initially wrap the embryo help to provide the immune shield, he notes. So why not express key trophoblast surface proteins in beta cells, so that the beta cells look like fetal cells to the immune system?

While autoimmune researchers have been kicking around both of these ideas for years, it’s still very early days for bringing beta cells with self-protective surfaces toward the clinic.

But some year, Melton told the JDRF crowd, “my dream is to tell you, not only can we make billions of beta cells but we can transplant them into any person and they won’t be rejected.”

I’m looking forward to hearing about more and better betas from him and from other leading researchers this Friday afternoon, at a session during the American Diabetes Association’s annual scientific conference in Boston.

Smarter insulin?

insulinInsulin was the first hormone to be genetically engineered for human use, and synthetic insulins now control blood glucose levels for almost everyone with type 1 diabetes and millions of those with type 2 diabetes. Many variants of the molecule have been designed to act quickly or slowly or in between, but the Holy Grail is smart insulin—which would not only work over many hours but automatically adjust its own release to keep blood glucose levels in a good range.

Many labs have taken a stab at smart insulin since the first attempt in 1979. New approaches keep cropping up, and a few show particular promise in animal tests. A “nanoparticle network” reported in 2013 is one of the more interesting. These nanoparticles combine insulin, dextran (a complex sugar often employed to slow down glucose effects) and enzymes that target glucose. Given opposite electrical charges, the nanoparticles are thought to clump together in the body rather than wander off in the bloodstream, doing their duty like a tiny pancreas.

The current commercial champion for smart insulin, though, began back in 1999 with research by Todd Zion, then an MIT graduate student in chemical engineering. Zion came up with a design that combined modified insulin with a sugar gel, and worked dramatically well in rats. He and his colleagues spun out a startup firm called SmartCells in 2003 and honed the technology well enough to get the attention of Merck, which snapped up SmartCells in 2010. Last spring, Merck remarked that it would bring a L-490 smart insulin based on the company’s technology into a clinical trial this year.

But there’s been no news from Merck since, and the National Institutes of Health’s site doesn’t mention a trial for L-490.

So we’re still in the animal labs.

But I’m still encouraged by last month’s paper in the journal PNAS on a different approach to smart insulin, developed by a team led by MIT’s Daniel Anderson and Robert Langer. This group found that an engineered insulin with the sprightly name of Ins-PBA-F performed very well, maintaining good glucose levels in the blood of mice without functioning pancreases over more than 12 hours and also behaving itself in normal mice.

Which sounds like L-490. But the researchers suggest that the approach Ins-PBA-F spearheads may offer better control and safety in the long run than L-490 because it’s more like normal insulin.

Ins-PBA-F starts with a molecular structure of a long-acting synthetic insulin, and adds a chemical group called “phenylboronic acid” (PBA) that binds to glucose and other sugars. When the surrounding glucose levels climb high enough to occupy the PBA group, the insulin itself is released for action. (Curiously, though PBA is often used to sense sugars and other carbohydrates, the PNAS paper acknowledged that the exact mechanism by which Ins-PBA-F responds to higher glucose levels in the blood isn’t yet clear.)

In theory, such a directly modified insulin molecule may be safer from immune reactions and other side effects than smart insulins that add gels or other types of protein barriers for glucose, as do L-490 and most other approaches. If so, that benefit will appeal mightily to the FDA, which will give extremely tight scrutiny to a radical new drug that could be used around the clock by many millions of people.

A successful smart insulin will be very far from a niche product. Anderson and Langer (who may well hold the world record for co-founding biomedical startups) have the attention of the venture capital community. I hope that successors to Ins-PBA-F will indeed move toward clinical trials, and eventually the clinic. That might be a very smart bet.

Update: Merck actually but very quietly moved its smart insulin into a two-part clinical trial in fall 2014.

Encapsulating answers to type 1 diabetes

People with type 1 diabetes are understandably excited about progress toward an “artificial pancreas” but they never lose the hope for a true cure, in which they can live like everyone else, without juggling synthetic hormones and hardware clomped on their skin that pierces their skin and will never work perfectly.

A true cure is a blue-sky goal built on two sets of major medical advances, and we have no idea what year those advances might arrive.

One set is to understand the autoimmune onslaught that brings on type 1 and then find a way to stop it. Serious and sometimes brilliant research keeps charging ahead, but autoimmune diseases hold extremely devious secrets and guard them very well.

The second set is to create a cells that replace those wiped out by the autoimmune attack and can generate insulin (and maybe related pancreatic hormones) at appropriate levels. Stem cell research aimed to do so is going gangbusters but is generally a long way from clinical trials.

With one big exception:

Yesterday Viacyte filed with the FDA for permission to run a trial for its VC-01 device, which encapsulates human progenitor cells—human embryonic stem cells that in this case have gone partly down the development path to hormone-producing cells. (The company, then known as Novocell, began work with embryonic stem cells years before the 2006 discovery of ways to create induced pluripotent stem cells, which possess very similar abilities to differentiate into almost any kind of cell but can be created from adult cells.)

Viacyte’s encapsulation container is a Teflonish cartridge about the size of a band-aid and thickness of a credit card, with holes too small for immune cells to enter but big enough to allow oxygen, glucose and other key ingredients to flow in and to allow insulin and other hormones to flow out.

The theory is that the capsule is inserted via an outpatient procedure, the immune system mostly ignores it, blood vessels build up to feed the cells, the cells are driven by signals within their fairly normal local human microenvironment to differentiate into a range of hormone-producing cells, the cells churn out insulin and its hormone cousins, and normal blood glucose levels and related metabolism are maintained. A functional cure, in short, for a year or two or three while the device functions properly.

This all works nicely in mice, but mice are not always man’s best friend in diabetes research. Investigators have struggled with encapsulation techniques for many years, and stem-cell-derived cells are unproven. The list of what could go wrong in the Viacyte trial is very long. Patients might reject the capsule. The cells might die quickly or slowly or never gather suitable blood vessels or fail in other ways. They might generate side effects that no one has imagined.

But Viacyte seems to have the science on its side and its head on straight and a good step-by-step plan. Assuming the FDA agrees, I don’t expect a home run in the first trial, but simply getting on base would be huge. And we should know within a year.

Viacyte Encaptra

Smart insulin readies for trial

If you have type 1 diabetes, your body produces little or no insulin, and you survive on injections of synthetic insulin. You always have a little too much or too little insulin running through your blood, except for times when you have a lot too much or too little. Too much, and you may start to slide rather quickly toward serious wooziness and maybe unconsciousness. Too little, and you increase your risk of serious complications down the road.

Thus the appeal of the concept of “smart insulin”, an insulin derivative designed to automatically react to the level of blood glucose and adjust the amount of insulin released so that glucose levels remain in a healthy range.

Back in 1979, Michael Brownlee and Anthony Cerami presented one smart insulin approach in Science. There’s been plenty of research on the concept since then, but it has remained a concept.

In 2010, though, Merck announced plans to buy the MIT spinoff SmartCells for a purchase price that may eventually exceed $500 million. And last week at a Merck investor briefing, Roger Perlmutter, executive vice president and president of Merck Research Laboratories, noted that the company plans to move a smart insulin candidate based on SmartCells work forward into clinical trials.

At the briefing, Perlmutter displayed just one slide, showing that injections of a drug candidate called L-490 could maintain blood glucose levels at normal levels in dogs using smaller dosages than needed with regular synthesized human insulin, thus suggesting that L-490 indeed was releasing insulin only as needed.

Merck’s plan for a trial drew little attention outside the type 1 community, as the markets concentrated on other drugs headed for bigger markets sooner. But if smart insulin works it might radically improve treatment not only for the millions with type 1 diabetes but for the substantial group of people with type 2 diabetes who already use insulin among their other therapies.