Lithium carbonate · Li₂CO₃ · MW 73.89 g/mol · CAS 554-13-2 · elemental Li⁺, MW 6.94 · ATC N05AN01
Lithium (as the carbonate or citrate salt). The lightest metal and the oldest psychiatric drug still first-line — a bare monovalent ion, not a receptor ligand. It has no binding pocket, no scaffold, no metabolism: it works by out-competing Mg²⁺ at the active sites of a family of magnesium-dependent phosphomonoesterases, above all GSK-3β and inositol monophosphatase. Brand names: Eskalith, Lithobid, Priadel, Camcolit, Carbolith.
Lithium is not a drug in the medicinal-chemistry sense. There is no ligand, no pharmacophore, no orthosteric pocket — just a hydrated Li⁺ cation with an ionic radius (~0.76 Å) close enough to Mg²⁺ (~0.72 Å) that it can slip into magnesium-binding sites. But Li⁺ carries only a single positive charge where Mg²⁺ carries two, so once seated it cannot support the catalysis that Mg²⁺ enables. The result is inhibition of a whole family of Mg²⁺-dependent phosphomonoesterases. Two of them dominate the psychiatric story.
Li⁺ is an uncompetitive inhibitor of IMPase (Ki ~0.8 mM), trapping the enzyme–substrate complex by displacing one of the two catalytic Mg²⁺ ions. IMPase is the terminal step recycling inositol; blocking it starves the PI cycle — the inositol-depletion hypothesis.
Li⁺ directly inhibits glycogen synthase kinase-3β (Ki ~2 mM) by competing with Mg²⁺ at the ATP-adjacent metal site, and indirectly by promoting inhibitory Ser9 phosphorylation. GSK-3β sits under Wnt/β-catenin, circadian, and neurotrophic signaling — the GSK-3 hypothesis.
With IMPase and inositol polyphosphate 1-phosphatase (IPPase, Ki ~0.3 mM) blocked, free myo-inositol falls preferentially in the most active neurons. PIP₂ resynthesis slows, damping overactive Gq/PLC signaling only where it is firing hardest — an activity-dependent brake.
GSK-3β inhibition disinhibits β-catenin and stabilizes the pro-survival program: ↑ BDNF, ↑ Bcl-2, ↓ tau phosphorylation, ↑ neurogenesis. Chronic lithium measurably increases gray-matter volume in humans — a structural signature no other mood drug reliably shows.
GSK-3β phosphorylates core clock proteins (REV-ERBα, PER2, CRY). By inhibiting it, lithium lengthens and stabilizes the circadian period — plausibly why it blunts the manic/depressive oscillation itself rather than just the poles.
The same displacement inhibits bisphosphate 3′-nucleotidase (BPNT1, Ki ~0.15 mM), phosphoglucomutase, and fructose-1,6-bisphosphatase. Most are hit only at supratherapeutic Li⁺ — but they explain the systemic toxicity that shadows the narrow window.
The unifying physics: every high-value lithium target is a magnesium enzyme, and lithium's therapeutic serum concentration (0.6–1.2 mM) sits precisely in the range of its inhibition constants for IMPase, IPPase and BPNT1. That is not a coincidence — it is why lithium works at all, and simultaneously why its therapeutic index is razor-thin. There is no way to raise the dose for more GSK-3β effect without dragging Li⁺ toward the concentrations that poison every other magnesium enzyme in the body.
Lithium's PK is unusually simple and unusually dangerous. Being an element, it is not metabolized, not protein-bound, and not conjugated. It distributes into total body water, and it is cleared entirely by the kidney, unchanged. Everything that governs lithium safety therefore reduces to a single variable: renal handling of the ion. Anything that shifts sodium balance or glomerular filtration shifts lithium — which is the root of every major interaction and every toxicity.
Renal handling — the whole game: lithium is freely filtered at the glomerulus, then ~60–80% is reabsorbed in the proximal tubule via the same paths as sodium. Lithium is invisible to the distal nephron, so it tracks sodium and volume status. This is why the clinically relevant "cascade" is not a metabolic one — it is a tubular one.
The dehydration / sodium coupling (★) is the single most important safety fact in the file. Because the proximal tubule reabsorbs lithium in proportion to sodium, anything that depletes sodium or volume — dehydration, a low-salt diet, vomiting, diarrhoea, heavy sweating, a heat wave, or a diuretic — makes the kidney claw back more lithium, and serum levels climb into the toxic range on an unchanged dose. Conversely, a sodium load speeds lithium excretion and can drop a patient subtherapeutic. Renal impairment (including lithium's own long-term nephrotoxicity) directly lowers clearance.
Steady state takes ~5 days; levels are drawn as a standardized 12-hour trough. The long half-life plus zero metabolic buffering means a change in renal function does not announce itself — it accumulates silently until a routine level, a tremor, or a confusional state reveals it.
Lithium is the best-validated mood stabilizer in psychiatry and, uniquely, the only psychotropic with robust evidence of an anti-suicidal effect that is at least partly independent of its mood effect. It remains first-line for bipolar maintenance more than seventy years after Cade's 1949 report — a survival almost no other drug of that era can claim.
Meta-analyses of randomized trials (Cipriani et al., 2013, BMJ) show lithium reduces completed suicide and self-harm versus placebo and versus active comparators in mood disorders — an effect not fully explained by relapse prevention. Population and observational data even associate trace lithium in drinking water with lower regional suicide rates. No other mood agent carries this evidence. This is the reason lithium is not replaced despite its inconvenience.
First-line for long-term prophylaxis in bipolar I, reducing both manic and depressive relapse; effective in acute mania (onset over 1–2 weeks, so often paired with an antipsychotic or benzodiazepine for rapid control). Maintenance targets ~0.6–0.8 mM; acute mania is pushed toward 0.8–1.2 mM. "Classic" euphoric mania with few prior episodes responds best.
Lithium is an evidence-based augmentation strategy for treatment-resistant major depression added to an antidepressant, and it potentiates ECT. The effect is modest but real, and — again — it pulls down suicidality, which is often the deciding factor in resistant, high-risk depression.
Anticonvulsant stabilizers (valproate, lamotrigine, carbamazepine) and atypical antipsychotics cover pieces of the bipolar spectrum, but none combines lithium's maintenance efficacy, anti-suicidal signal, and neurotrophic/gray-matter effects. The cost is the monitoring burden and the narrow index — a trade every prescriber and patient negotiates explicitly.
The mechanistic irony is that we still cannot say which of lithium's actions — inositol depletion, GSK-3β inhibition, circadian resetting, or neurotrophic signaling — is the therapeutic one. They are almost certainly not separable: all descend from the same act of an ion displacing magnesium from a set of phosphomonoesterases that happen to sit at the crossroads of neuronal signaling, survival, and timekeeping.
Lithium is the degenerate limit of a docking problem. A conventional ligand carries
rotatable bonds, a hydrophobic surface, and a conformational ensemble; docking it means estimating the
ΔS_conf penalty as a flexible molecule is pinned into one pose. Li⁺ has none of that
— it is a point charge with zero internal degrees of freedom. Its entire binding thermodynamics live in
its hydration shell and in the protein and solvent around it, not in the ligand.
What lithium actually does at the site shown in PDB 1PYX is substitute for a
coordinated metal. In FlexAID∆S terms this is not a lock-and-key entropy collapse of a flexible
ligand but a desolvation + metal-exchange event: Li⁺ sheds part of its tightly held
water (a large, favorable ΔS_solv), enters the octahedral Mg²⁺ coordination geometry, and
— because it brings only +1 where +2 is required — fails to organize the catalytic transition
state. The Shannon entropy of the active-site metal microstate does not collapse onto the
productive conformer; it is diverted to a catalytically dead one.
The instructive contrast with a cocaine- or DMT-type ligand: those achieve inhibition by
rigid-pocket trapping — a high-affinity molecule freezing a flexible protein. Lithium achieves
inhibition by charge mismatch inside an otherwise normal coordination sphere. The measured
inhibition constants (IMPase Ki ~0.8 mM, GSK-3β Ki ~2 mM) correspond to weak,
millimolar binding — a ΔG of only about −4 kcal/mol — which is exactly why
therapeutic concentrations must be held in the millimolar band, and exactly why the therapeutic window is
so unforgiving. There is no affinity headroom to spend.
Clinical, non-moralistic. Lithium is highly effective and highly demanding: the gap between the therapeutic level and the toxic level is smaller than for almost any drug in routine use. Respect the number, monitor the number.
| Target | Li⁺ potency | Rel. | Mode |
|---|---|---|---|
|
BPNT1
Bisphosphate 3′-nucleotidase
|
Ki ≈ 0.15 mM
most Li-sensitive
|
Mg²⁺-comp. | |
|
IPPase
Inositol polyphosphate 1-phosphatase
|
Ki ≈ 0.3 mM
|
Uncompet. | |
|
IMPase
Inositol monophosphatase (IMPA1)
|
Ki ≈ 0.8 mM
inositol-depletion
|
Uncompet. | |
|
GSK-3β
Glycogen synthase kinase-3β
|
Ki ≈ 2 mM
+ indirect Ser9-P
|
Mg²⁺-comp. | |
|
FBPase
Fructose-1,6-bisphosphatase
|
Ki ~ mM
|
Mg²⁺-comp. | |
|
Serum (Rx)
Therapeutic window
|
0.6–1.2 mM
toxic >1.5 mM
|
in-range |