“Metal Machine Music” and Learning the Loops of Life

We react to our environment as much as our environment reacts to us.

Exposome Perspectives Blog by Robert O. Wright, MD, MPH

As a teenager in the 1970s and early 80s, I would peruse used record stores in search of a bargain, especially since I didn’t have much money and there was no such thing as Spotify. I grew up on vinyl, and used record stores were like an antique road show. Over the course of hundreds of hours, I found quite a few gems: The Kinks – “Muswell Hillbillies,” Stevie Wonder’s “Innervisions,” and the Velvet Underground’s “White Light/White Heat” among others. Granted, the better the album the more likely it was scratched up, and crackly on the turntable, but the price—as low as 50 cents in some cases, made it worthwhile. As a Velvet Underground fan when my 18 year-old self found Lou Reed’s “Metal Machine Music” in a dusty milk crate in the back of the store—I was incredulous!  It was spotless, no scratches at all! “These fools don’t know what they have!” I picked up a copy of the Doobie Brothers’ greatest hits on the way to the register and hid the Lou Reed album behind it as a decoy just in case it was all a mistake. I scampered out to my 1978 Ford Pinto and dashed home. Ironically, I only listened to Metal Machine Music once—less than once, actually, as after about 2 minutes I couldn’t take it anymore. To my shock, the Doobie Brothers had made a much better album. Lou Reed’s Metal Machine Music consisted of guitar and audio feedback—loud, screeching and unbearable. Maybe he was in a bad mood that day.

But feedback isn’t always bad and feedback loops are the hallmark of biological systems. Like music, biology is rhythmic, we sleep every 24 hours, we secrete hormones at predictable levels depending on the time of day, even our body temperature rhythmically changes throughout the day.  Biology also uses feedback to maintain these rhythms, as all biological and behavioral functions self-regulate their rhythms at the macro and micro levels. If the temperature rises we sweat to cool our bodies, and if it lowers we shiver to bring our temperature up. Hormone receptor activity regulates the production of increased or decreased hormone in endocrinology through feedback hormones released in response to other hormones.

As a medical student, I struggled to interpret hormone blood tests. Exam questions would give values for thyroid hormone secreted by the thyroid gland and Thyroid Stimulating Hormone (TSH) – the hormone secreted by the brain’s pituitary gland to stimulate secretion of thyroid hormone. Students would have to figure out if a scenario represented hypothyroidism or hyperthyroidism. But because it was a test, we weren’t given all the available information.  So we had to try to piece the right answer together from what we were given.

One lesson learned from that experience is that elevated TSH can mean hyperthyroidism in some settings and hypothyroidism in others.  A pituitary tumor can over secrete TSH and cause hyperthyroidism, but in hypothyroidism, the thyroid gland secretes less thyroid hormone because its thyroid receptors are less active—most commonly because of autoimmune disease. High thyroid hormone in the blood inhibits the secretion of TSH, so the low blood levels of thyroid hormone in hypothyroidism are part of a feedback loop that tells the pituitary gland to secrete more TSH in compensation- i.e. to maintain homeostasis. TSH can be high in hypothyroidism as well.  Because elevated TSH could mean either hyper- or hypothyroidism. Often, there was a seemingly unrelated piece of information in the question that I ignored that if noted, would point me in the right direction. I learned not to interpret blood levels of a biomarker in isolation. I needed more information to understand what was happening because of the feedback loops.

Much of what we do in research is akin to searching in the dark—like the parable of the blind men and the elephant. In the Buddhist version of this story, a king asks a group of blind men to describe an elephant he has had brought to his palace. Each of the blind men feels a different part of the elephant—her ear, her tusk, her leg, or her tail. The king then asks “How do you describe this animal.” Based on where they had touched her, they described a fan (ear), a plowshare (tusk), a plow (trunk), a wall (body), a pillar (foot), or a rope (tail). This parable is a metaphor of data interpretation when we don’t know all the parts of a system. It depends in many ways on where we are in feedback loops and rhythmic cycles of our lives. DNA methylation is an epigenetic mark that regulates gene expression.  DNA methylation is a relatively recently discovered biological measure. We believe it changes over time and is part of the regulation of gene expression (i.e. turning a gene on or off). In fact, think of DNA like a slinky—when it is closed all the rings are next to each other. 

If those rings were coiled DNA they would look like the slinky on the top—how could a protein like RNA transcriptase get to a gene to turn it on?  If it were to open like the slinky on the bottom, there is now some space for RNA transcriptase to get in there and transcribe the DNA turning the gene on. Conventionally, we believe higher methylation means a tighter coil and lower gene expression while less methylation means a looser more open coil of DNA.  This is part of the ways that environmental signals may affect gene expression by altering the amount of DNA methylation near a gene. This is part of the way we turn on and off genes in response to our environment. If all we measure is DNA methylation in a gene’s promoter region following a chemical exposure and then interpret the results we are not looking at the whole picture or how feedback loops regulate biological functions like gene expression. Like computers we have a tendency to think linearly.  Increased DNA methylation is associated with tightly coiled DNA and decreased gene expression, and therefore reduced levels of the protein that gene encodes. But that would only apply if the exposure directly caused the DNA methylation to increase. If we are looking at a gene that codes for thyroid hormone synthesis and we find higher DNA methylation in that gene we would predict lower levels of thyroid hormone in the blood.  What if DNA methylation is part of a complex feedback loop that involves gene expression, hormone secretion, and serum hormone levels?  Remember the example above about interpreting thyroid hormone levels with incomplete information? If all we measure is DNA methylation and not gene/protein expression/metabolites as well—which are all part of the feedback loop we actually don’t know what is happening.

Let’s think about how this could work in a feedback loop. Is it possible that we could conduct a study and find that higher DNA methylation for the gene that codes for thyroid hormone synthesis is associated with higher levels of thyroid hormone? We think that higher methylation turns a gene off, but we didn’t measure all the components of the feedback loop. We are the blind man touching the elephant’s leg and thinking it is a pillar. Let’s say the chemical exposure we are studying inhibited the breakdown of thyroid hormone. That would lead to higher serum levels of thyroid hormone.  In response the body may try to reduce thyroid hormone synthesis, in order to maintain homeostasis. Part of this process could be to reduce the expression of genes involved in synthesizing thyroid hormone. This would lead to increased DNA methylation around the gene that is responsible for synthesizing thyroid hormone. This is all part of the feedback loop that maintains homeostatic levels of thyroid hormone. So yes, we could see higher DNA methylation of a gene involved in the synthesis of thyroid hormone associated with higher serum thyroid hormone.

This brings me to directed acyclic graphs (DAG). DAGs were developed in computer science as a method to illustrate a collection of events that occur in a sequence to produce an outcome. Today they are also used in epidemiology research to determine how/whether to use a variable in a statistical analysis—they are often used to help us decide if it is appropriate or inappropriate to statistically adjust for a potential “confounder” variable. Because every component of a DAG happens in a sequence requiring time to pass that make us consider the role of time. DAGs assume that all causal relationships have a direction, and therefore no variable can be both a cause and an effect (i.e. they are directed). For this reason, DAGs assume that no feedback loops exist (hence the term “acyclic”). Because DAGs forbid feedback loops they might create interpretive bias in how we explain study results- particularly when we try to bring in molecular mechanisms like DNA methylation.  We know that biology has feedback loops and DAGs conceptually force feedback loops to exist only via time travel—which means there are no feedback loops—a conundrum worthy of a Star Trek episode. Using DAGs also inadvertently imposes a second restriction—that individuals cannot influence their environmental exposures or that behavior and biology don’t happen in a self-reinforcing loop. There is ample evidence that we do influence our environment and it is not implausible that an external exposure can reinforce behavior that encourages additional exposure. Someone addicted to smoking gets reinforcement to continue smoking with each cigarette. Feedback loops don’t actually imply time travel—they imply that much of life fits within Newton’s concept of actions causing equal but opposite reactions.  

What DAGs get right is that they force us to think- a priori- about how a variable may impact the statistical analysis and therefore how to treat it appropriately, so I am not advocating that we abandon DAGs, but ignoring feedback loops is still a DAG flaw, because all of biology, behavior, economics and life consists of feedback loops. 

We react to our environment as much as our environment reacts to us. Without appreciating feedback loops within a system, we tend to interpret biomarker data in a linear, deterministic, manner—like a computer. When I bought my Lou Reed album, I didn’t have all the information, and didn’t understand that by releasing this album of cacophonic noise, he was freed from a contentious record contract. Like a researcher excited by a novel finding, but ultimately groping in the dark and misinterpreting, I bought a record I didn’t understand. But that’s ok, we just have to remember that possibilities are never set in stone when we interpret data. We should acknowledge that we have limited information and understanding of how biology works. Just to show that all of life is a series of actions and reactions (i.e. feedback), “Metal Machine Music” actually lives on—the album was the inspiration for the static language of the Alien species the “Breen” in the Star Trek canon of television series. So Lou Reed, as usual, got the last laugh.

“Creativity — like human life itself — begins in darkness.”  Julia Cameron