Lost in Post-Translation

The treatment of bipolar disorder (BPD) is a huge unmet medical need. As the sixth leading cause of disability in the world, BPD is a widespread and lethal condition with no cure. Although the stem cell field has become quite adept at modeling monogenic diseases, unraveling the molecular pathogenic mechanisms underlying polygenic, multifactorial diseases remains a challenge – and psychiatric disorders could be considered the “poster children” for complexity. But of all the psychiatric disorders, BPD is the only one with a molecular handle. A third of bipolar patients respond to lithium – but the mechanism of action remains a complete mystery.

I reasoned with my colleagues that if we were able to map the molecular lithium response pathway that is specific to BPD, it would give us a clue as to the underlying pathogenic roots and pathophysiology of the disorder – potentially bringing more effective treatments one step closer.

Taking advantage of a happy accident

If we don’t know why lithium works in some BPD patients, how did it become the standard treatment? Back in 1871, some clinicians in Denmark believed that mood disorders were associated with gout and high uric acid levels in the blood, and so lithium was prescribed to enhance renal excretion of uric acid in manic patients. It turned out that the drug did nothing for gout or uric acid levels, but some patients nevertheless appeared to improve, and lithium use continued. Such anecdotal observations ultimately led the Australian psychiatrist John Cade to publish – in 1949 – the first paper specifically on the use of lithium in the treatment of acute mania, without knowing its true mechanism of action. Lithium deficiency, for example, is not the cause of BPD. Such a knowledge gap is common in the history of medical therapeutics for oft-used treatments. However, in the case of lithium, the safety index is extremely narrow, and the adverse off-target effects, including nausea, irregular heartbeat and birth defects, are intolerable to some patients. Thus, finding a treatment that targeted the actual cause or pathophysiology of BPD might allow lithium to be replaced by better and more specific pharmacotherapies.

Much is known about the actions of lithium. In fact, it affects every organ and cell type in the body of every kind of organism, making that knowledge unhelpful for understanding how it can impact such a complex higher-order human malady. To unveil the key action of lithium specifically in BPD, we used human induced pluripotent stem cells (hiPSCs). (Indeed, this may represent the first use of hiPSCs to seek out the specific molecular underpinnings of a complex polygenic disorder rather than simply describing its phenotype and phenomenology.) We used what we dubbed a “molecular can-opener” strategy to identify the underlying molecular pathogenic mechanism of BPD. In this case, lithium was our “can-opener”, which we used to “pry” into the pathophysiology of the disorder. We reasoned that, if we could find lithium’s target specifically in BPD neurons, we could then define the “lithium response pathway” – which would likely be uniquely abnormal in BPD. We also reasoned that examining proteomics may be more informative than looking at genes and transcriptomics, because the proteome is the ultimate integrator for the cell of multiple inputs at the genetic and epigenetic level, and determines the neuron’s actual behavior.

Targeting CRMP2

Using unbiased proteomic techniques to study the protein interactions of BDP-hiPSC-derived neurons we identified the target: collapsin response mediator protein-2 (CRMP2). We then mapped upstream and downstream from that node to complete a picture of the lithium response molecular pathway, and validated it in human neurons, mouse models (both histologically and behaviorally), and in actual bipolar human patient brain specimens.

We found that lithium alters the activity state of CRMP2, which is a central modulator of cytoskeleton (1). When CRMP2 is active, it binds cytoskeletal elements; and when it is inactive – by being phosphorylated – it releases cytoskeletal elements. This toggling back and forth between an active and inactive state is a normal adaptive mechanism, and there is a certain ratio of inactive-to-active CRMP2 (~0.5 in hiPSC-derived neurons, but we are still in the process of defining the normal ratio in patients). In particular, the molecule appears to dictate the form and function of dendritic spines – the key to neural network assembly and function. When CRMP2 is active, dendritic spines are active; when it is inactive (or, in the extreme, missing), dendritic spines are inactive and even diminished in number. The set-point for the ratio of inactive:active CRMP2 is simply set too high in lithium-responsive bipolar patients; lithium “resets the thermostat” to a normal level, and neural networks function more normally.

It came as a surprise that an inherited and developmentally-based disease like BPD was not caused by an abnormal gene, but rather by the aberrant post-translational regulation of a normal gene. We have now described the first disease caused by an abnormal post-translational modification (PTM) in psychiatry (and perhaps in any non-neoplastic medical condition). It was also surprising that the pathogenesis came down to something as structurally fundamental as cytoskeleton.

Better diagnoses – and new drugs?

Assessing the CRMP2 ratio for a given patient could confirm the diagnosis of BPD, indicate whether or not the patient should be started on lithium, and allow the progress of the patient to be followed. We envisage a test that determines the phosphorylated:unphosphorylated CRMP2 ratio in blood samples from patients. What we now need to find out is whether the ratio can be seen in lymphocytes alone or whether we need to convert the lymphocytes to neural progenitor cells or neurons first and then determine the ratio. Even though the conversion process could take up to 8–12 weeks at this stage, it would still be faster and cheaper than the typical management of bipolar patients, which entails a trial-and-error prescribing of drugs (and often more than one) over years before the right treatment is found. Intriguingly, we would be using a highly accessible means (a blood draw) for predicting and characterizing the complex circuitry of a highly inaccessible organ (the brain).

Ultimately, finding compounds that alter this pathway would allow us to discover drugs that are much more effective, selective, and less toxic than lithium. We also believe that the pathway plays a role in other neurological conditions (including some neurodegenerative diseases); therefore, effective drugs might be even more widely applicable. Another take-home message is a reinforcement of the critical nature of cytoskeleton to the function of neurons. Furthermore, although the neural stem cell and regenerative medicine fields have tended to focus on neuron replacement and neurogenesis, it is probably more accurate – and fruitful – to address neural network preservation and reconstruction as the key to the reversal of disease symptomatology.

The technique we used could be thought of as “reverse drug discovery” – rather than first discovering a mechanism, pathway, or target and devising a drug against it (the traditional route of drug discovery), we started with an agent that is already known to be bioactive, and then worked backwards to find a pathogenic mechanism or pathway. With that knowledge in hand, we can now go forward and seek better pharmacotherapeutics, which could potentially benefit many more patients.