Diagnosing Autism with a Single Strand of Hair

Our interactions with the environment can have unexpected effects on our genes and trigger a biologic response that leads to the onset of disease. These interactions can also leave a measurable record in what’s referred to as the exposome. Linus Bio, which emerged from the exposome laboratory at Mount Sinai Health System, has developed a test for autism that relies on analyzing a single strand of hair. The company said the test is capable of diagnosing autism at birth. We spoke to Manish Arora, founder and CEO of Linus Bio, about the exposome, how the company’s test for autism works, and how this opens the potential for early interventions.

Daniel Levine: Manish, thanks for joining us.

Manish Arora: Danny, it’s a pleasure. Hello to you and hello to your listeners.

Daniel Levine: We’re going to talk about autism, Linus Bio, and its test that uses a single strand of human hair to diagnose the condition. Let’s start with autism. What is autism and how does it manifest itself and progress?

Manish Arora: Sure. Autism spectrum disorder is a neurodevelopmental disorder, and it’s characterized, and at present even defined, by its symptoms. And the symptoms mainly are these disruptions in communication, in language, in interaction. Children with autism spectrum disorder will also sometimes show repetitive behaviors. And when you package it all together, this is the condition we currently call autism spectrum disorder.

Daniel Levine: How well understood is the condition and is it a single condition or does the spectrum suggest multiple related conditions being lumped together?

Manish Arora: That’s an excellent observation and a very good question. So, autism spectrum disorder, because it’s based on symptoms, the underlying reality [is] we are still at that stage where we are learning new things every day. For example, we know it’s highly heritable, some estimates are 50 percent, others are higher up to 80 percent, yet we can’t identify a single gene that is linked with autism. In fact, for a decade or so, we were looking for the autism gene, but then we realized sometimes there’s more than a hundred genes that are involved, all contributing a small amount to the heritability, and they’re inconsistently involved. Not every autistic child will have those genetic variations. So, the underlying biology is not that very well understood. The piece that has the most that we need to discover, the most that work has to be done, is on the non-genomic or the environmental component. So yes, there is a lot that has to be done. I’d like to say that we are at the beginning of this journey rather than closer to the end.

Daniel Levine: Is there any sense of the balance between genetics and the environment and how they each contribute to the condition?

Manish Arora: There is a growing acceptance that autism is one of those disorders that has both a genetic and environmental component. I, for one, believe that that’s true of every disease. There is no such disease that’s purely genetic, and there’s no such disease that’s purely environmental. All diseases have a genetic and environmental component. One of the big, you know, flashing red lights when it comes to autism is that its prevalence is increasing. Now, some of that is due to better diagnostics and more awareness, but there’s also a real increase. The past hundred years let’s say now, in the last hundred years a human genome hasn’t changed much. We are still humans. So, our genetic code hasn’t changed. So, all this increase in prevalence cannot all be due to genetics. There must be an environmental or a non-genetic component. And the next big discoveries in autism will come from this combined approach of looking at genes and the environment together.

Daniel Levine: How is autism treated today and what’s the benefit of early diagnosis?

Manish Arora: Currently autism is treated by providing therapy. So, there are many modules of therapy and there’s evidence that this therapy is successful if delivered early. This therapy involves pace around play, around interaction, around language. So, it’s very much what we call psychosocial therapy. There’s evidence around the benefits of early intervention. For example, one of my favorite studies was done in the U.K. They delivered a packet of therapy. Part of that was delivered by the parents at home in fact, and it lasted just a few weeks. And then they brought these children back for reevaluation after 10 years. And those kids who had received therapy on some aspects of their autism measures, they were still doing better. So, here is this wonderful piece of evidence, and it’s been shown in other ways. There are wonderful pieces of work coming out of the U.S. One of them is called the Early Start Denver model. There is another model called Jasper, and all of them have shown that, well, if you can find those kids who need this therapy, you can characterize them well and deliver a customized package of therapy, then they do improve, and that improvement can last a long time. However, here is the problem, and here’s a problem that is a real failure at a public health level—that the median age of diagnosis of autism in the U.S. is always stuck around four years of age, and it hasn’t budged for the last several years, even though our awareness of this condition has improved so much. If you live in remote areas where there are less services, that median age can be as high as five and a half or even six years. But when is our brain developing? When are we developing language and interaction and all these other skills that are so much part of the autism spectrum? It’s all in the first year of life. So we are just not able to identify children early enough who would benefit from therapy. And that’s why my work has really focused on developing an objective test that can find those kids and accurately identify kids who are at risk of developing the autism phenotype.

Daniel Levine: How is autism typically diagnosed today?

Manish Arora: Today, autism is diagnosed based primarily on your appearance, your symptomatic and your behavior presentation to a clinician. So, it’s very much diagnosed and its core definition even based around deficits in social interaction, in language, in behavior, for example, repetitive behavior. So, to describe it to your listeners and to put a picture in their mind, these are questionnaires that are filled in by the parents or the caretakers. Some are filled by the clinician who observes them. And so, we are now using these kinds of tools. There’s some newer technologies, for example, that will monitor where the child’s eyes are looking. They’re called eye tracking software and very much symptom based. Now, as soon as you look at a medical condition and say, we will define it, we will detect it, and we will finalize the diagnosis entirely based on symptoms, we know that that can be problematic. Let me give you an example of where things went very wrong, and it’s sort of a lesson from history that I have learned. Back in the day, we taught [that] your gastric ulcers are caused by stress. So, you go to your family physician and you present with stomach ache and all the other symptoms, and they say, well, you have a gastric ulcer, we think, and it’s caused by stress. Now this took a very sinister turnaround in the thirties and forties when physicians were actually promoting certain kinds of cigarettes, and you can search it online and you’ll find lots of advertisements of physicians and physician organizations actually promoting certain brands of cigarettes, including as a treatment for stress and therefore a treatment for gastric ulcer. Fast forward a few decades, these two Australian doctors showed that, well, it’s caused by a bacteria. So, how different a condition is when it is purely detected and defined on symptoms versus when you bring some real biology into it. And that’s the sea change I want to bring about in autism. Yes, the symptoms will always be important. They’re the beginning of the story, but they cannot be the middle and the end of the story. There has to be a biological definition and a biological tool to detect and characterize autism.

Daniel Levine: Linus Bio has developed a test. Before you describe what it is, I think it would be useful to explain a concept. What is the exposome?

Manish Arora: The exposome is a fairly recently developed concept that focuses on the non-genomic components of our physiology and how those components play a part in disease. To be more specific, it looks at your lifelong environmental inputs. And when I say the environment, I mean that in the broadest possible sense as the environment outside your body. It’s also the environment inside your body. For example, the various metabolites that our body produces when we eat food and even the microbiome, the bugs that live in our gut and our mouth. So, it’s a very holistic view of our physiology. Now I’m a professor and vice chair of environmental medicine at Mount Sinai and we set up one of the first institutes in the country for exposomic science. And the technology that was developed and later became Linus Biotechnology and the company developed out of Mount Sinai’s academic department. And here’s the core idea, and here was the core academic innovation, which has now gone to the marketplace to actually help people. If you think of your genes, almost all of your listeners would know that your genes are fairly static. At conception, your base genomic sequence is coded. We get some from our fathers, some from our mothers, and then that is your genetic makeup for the rest of your life. However, all the non-genomic components change all the time. What you eat for breakfast is different than what you eat for dinner. When you’re taking a public commute to work, the exposures you experience are very different to what you are experiencing, what you’re inhaling when you’re asleep in the middle of the night. So, unlike our genetic makeup, our non-genomic components is always humming—there’s almost like a song like nature to it. Now, contrast that with how we experience medical care. We go to our physician once a year and we take these snapshot measures, a blood test, a urine test perhaps. So it’s almost like there’s a disconnect between the practice of medicine, of viewing our physiology almost in static terms, whereas our physiology is never static. So, that is the core idea, or the question that Linus Biotechnology was asking. Its answer to the innovation, the inventions that address this question, that’s what led to the autism biomarker.

Daniel Levine: Linus Bio has developed the StrandDx ASD test. What is it, what does it look for, and how does it work?

Manish Arora: Sure. Like I said, the exposome is constantly humming or moving. It’s not static. So think of it like this. You might need a hundred blood tests to characterize the exposome, but nobody has a hundred blood samples that they’ve collected and stored somewhere. So, we came up with a technology that can give you the same amount of data, but all from a single strand of hair. In fact, we generate data that is more than 500 blood samples, sometimes up to 2000 blood samples, all from one strand of hair that you can collect at home. You never have to go into the clinic, just put it in the envelope and ship it to us. From this one strand of hair, we use a laser. The laser tracks along the growth rings of your hair strand, very much like the growth ring that you would see in a tree. And as it’s moving along, there’s growth rings. We are going back in time and we do this at a two hourly resolution. So, we have a measurement that’s every two hours and within that we measure thousands of molecules. Then we use machine learning and artificial intelligence. And what we were able to identify was a specific molecular pattern of autism. When we first discovered it, we discovered it in Sweden amongst twins. Then we actually said, this is an interesting finding. How do we make sure that it actually works? So, we went to Japan and we studied newborn Japanese babies. Some of them went on to develop autism. And then we replicated a third time in our own autism center here at Mount Sinai in New York. And again, the patterns were consistent, we put it all together and we went to the FDA for diligence and for their advice. And the FDA says, yes, this is a breakthrough. They’re labeled, they described this as a breakthrough technology. And so currently the StrandDx ASD biomarker has what we call an FDA breakthrough designation.

Daniel Levine: What’s the result that’s returned to a physician? Is it a positive negative? Does it provide some kind of score? And given that this is a spectrum disorder, does it give any reflection of the intensity?

Manish Arora: Excellent questions and absolutely that we start off with a positive and negative answer. We describe this tool not as a diagnostic, but as a diagnostic aid. So, what we do is we provide the clinician with a yes and no answer, but we leave it for the clinician to make the final judgment call. Think of it very much like a cholesterol test where you say, oh, it’s high or low, but the clinician still has to use their expert judgment to decide whether you have heart disease. And that is one aspect of StrandDx. What it also does is it can subtype what type of autism you have, as you have said before, and very rightly that it’s a spectrum disorder. It’s many different conditions under one umbrella term. One thing that we have published is to show how some kids can have both autism and ADHD. That is the attention deficit hyperactive disorder. Other kids will only have autism and some kids will be misdiagnosed and they actually don’t have autism. They have ADHD and some kids will be neurotypical and don’t have either. We cannot do that with this test. From the same strand of hair, we can make a prediction of whether you fall on the ADHD spectrum of things or on the autism side of things. But beyond that, because we are measuring molecular signatures, we often find signatures that are readily modifiable. I’ll give you one example. In one child, we found that there was a certain drop in calcium and magnesium, because remember with hair, we are going back in time. This is not a snapshot technology. This is like a little time machine that we build out of a single strand of hair. We spoke to the pediatrician and the child’s parent. They said, yes, my child decided developing this gastric symptom and certainly stopped consuming all dairy. In fact, they had developed this kind of nutritional deficiency because they stopped consuming dairy and straight away the pediatrician could provide some supplements and say, look, we have got to build this part of your nutrition back up. It just can’t be a dairy source. So, we provide much more than just a yes and no. We provide what we like to call actionable information.

Daniel Levine: How accurate is the test and what’s been done to validate it?

Manish Arora: Overall? It’s more than 80 percent accurate. What we have done to validate it is that first of all, like I mentioned before, we studied it in twins in Sweden. The reason we studied it in twins is because there is a genetic component and identical twins obviously are genetically identical. Some of our patients, although they were identical twins, one went on to develop autism, the other one did not. So, we could really focus on the non-genetic component because they were genetically identical. But I doubt if most of your listeners or any of your listeners are Swedish twins. So, how does it apply to them? Well, we said let’s replicate this result in a completely different country. And we did a national study in Japan where we had patients from every prefecture of Japan. Now, in that study, we had excellent data, but again the Japanese population is not as diverse as let’s say a New York population. So, we replicated it a third time, which is very rare in studies like this. And we replicated that in studies in New York, in Mount Sinai. And when put all the data together, we see this very clear signature of what distinguishes a child with autism spectrum disorder versus those who do not have autism. We went a couple of steps further in a subset, we actually did MRIs of these children’s brains and we could show that all of those exposomic signatures, those biomarkers that we see in the human hair of autistic children actually correlate to how the brain is functioning. So, we’ve done many deep analyses where we look at this new technology versus known measures that have been studied in autistic children and found to show abnormalities.

Daniel Levine: Your tests can detect autism at birth prior to the presence of recognizable symptoms. Does that suggest anything about the condition?

Manish Arora: We know autism has a big heritability component. Based on that simple fact, we know that hot autism would exist at conception, more clear risk factors for autism would exist at conception. I would argue as a public health person to go even one step further and say autism risk can exist in either parent even before they decided to have the child. And they might be sub-threshold for the autistic phenotype, but when they have a biological child together, that child is above the threshold in terms of risk. So we know autism exists early in life. Many others, my own work, but work of others, for example, on air pollution has shown that the risk of autism, there is a critical window of susceptibility somewhere in the third trimester that are hit at that time, can really exaggerate your risk of ending up on the autism trajectory. So, autism exists at birth for sure, but in my view, based on the heritability data, based on many other folks’ work, it can actually exist from conception onwards.

Daniel Levine: Linus Bio is offering the test today under CLIA lab model, but you’re seeking to get FDA approval. What’s the thinking there? Is that driven by discussions with payers?

Manish Arora: There are two streams of thought there. One is that we will have a clear LDT test ready to be offered by the end of this year. And of course, that is how a majority of diagnostic testing in the U.S. works. Not every diagnostic test has to be cleared by the FDA and that’s why this LDT model, which stands for laboratory developed test, is so popular. However, all the co-founders of Linus Bio are academic professors. We invite diligence. The FDA is the highest office in the land that provides scientific oversight when it comes to providing care and services to patients. So, we are inviting them for diligence because all they will do is help us improve our work. And yes, you said it very correctly that once we have an FDA approved product, it makes it so much more accessible to a majority of patients in the country. We live in a prioritized healthcare system and your insurance company will decide where to invest their funds. Once a product has FDA approval, it removes some of that risk. It removes some of those barriers. And at the end of the day, as a public health professional, I want this technology to be accessible to as many people as possible.

Daniel Levine: Have there been discussions with payers yet today? Do you anticipate reimbursement as a CLIA lab product?

Manish Arora: Yes, we are talking to private insurance providers, also to IDNs and employer-based networks. So, in the U.S. large employers will often provide access to services which are not necessarily reimbursed by insurance, but are part of their incentives to their employers. You know, often they might take the form of reproductive assistance, as an example that quickly comes to mind. They might take, for example, smaller things, you know, yoga classes or help with maintaining your mental and physical health. Why not this? So, we are already having some conversations with large employment groups saying, hey, can we then provide this service to your employees even before it is fully FDA approved or even before it has private insurance coverage.

Daniel Levine: The same testing approach I imagine can be applied to other conditions. Are you looking at other conditions and if so, why start with autism?

Manish Arora: Yes, we are definitely looking at other conditions. In fact, we’ve already submitted one packet of information on ALS—that stands for amyotrophic lateral sclerosis, most commonly known in the U.S. as Lou Gehrig’s disease. And we have submitted initial data to the FDA and we will now continue that conversation. I’ve already mentioned autism, ADHD. We’re also working on psychosis and schizophrenia. We started with neuroscience as a broad area because think of it, most neurological disorders still do not have accurate testing and that snowballs into very big problems. You’re diagnosed with schizophrenia because you had a schizophrenic episode. You’re diagnosed with ADHD because you’re showing attention deficit. This is very different to, let’s say, how diabetes is diagnosed, which is based on a blood test. This also creates a problem of the inability to develop proper drugs. How do you develop a drug when your endpoint is based on questionnaire measures? Whereas if you have an accurate biomarker, a molecular marker, you can start developing drugs. In fact, we are doing two phase 2 trials of interventions developed by other companies using the Linus Bio platform. So, we are not just a diagnostic company, we are a true platform of innovation that can help develop drugs to treat autism. The reason we picked autism to begin with in the CNS field is because now between 2 and 3 percent of American children and children in other countries are impacted by autism. So, it’s a major public health problem. This is not a small issue facing us as a society. It is a major issue and I firmly believe that this new science can have an impact here.

Daniel Levine: The company completed a $16 million financing in January. How far will existing funding take you?

Manish Arora: This funding will take us to our CLIA lab, which is now fully operational, to LDT designation and FDA designation and beyond. So, we are financially quite sound in a position where we will have the ability to roll out this autism testing in high volume. And we are laying down the groundwork for that, using this capital. For example, we just engaged a Japanese robotics company to automate all our lab platform. You know, just taking a piece of hair or strand of hair out of a tube and laying it down straight so that my laser can analyze it, well you should be able to build an automated platform. This is a task that robots were meant to do, not a human. So, we are now automating everything. By the end of this year, we’ll have enough funding to roll out tens of thousands of tests every year to those people who need it and to provide the clinical data to the clinicians who can then use that to improve the care to their patients.

Daniel Levine: Manish Arora, co-founder and CEO of Linus Biotechnology. Manish, thanks so much for your time today.

Manish Arora: Danny, it was a pleasure. Thank you for the opportunity.

This transcript has been edited for clarity and readability.