The lithium-magnesium-zinc axis under stress: regulation of N-glycosylation-linked proteostasis and Alzheimer’s disease Authors: John V. Schloss, Teresa Nguyen Published: 2026 ABSTRACT Inflammation resulting from the unfolded protein response to glycosylation dysfunction may be the key driver in the progression of Alzheimer’s disease. Stress increases the risk of Alzheimer’s disease, in part, by depleting essential cofactors required by enzymes involved in N-glycosylation (the first and third, second, fourth, and fifth enzymes-essential cofactors required for dolichol diphosphate-sugar conjugate synthesis are, respectively: DHRSX-NAD/NADPH, SRD5A3-NADPH, DOLK-Zn2+, and DPAGT1-Mg2+). Lithium reduces the risk of Alzheimer’s disease by: (1) increasing transcription of DPAGT1, the gene for the first committed enzyme of N-glycosylation, stimulating levels of β-catenin through glycogen synthase kinase 3-β (GSK-3β) inhibition; (2) increasing transcription of SRD5A3, DHRSX, and DOLK, the genes for the enzymes of dolichol phosphate synthesis, stimulating levels of transcription factors through GSK-3β inhibition; and (3) inositol monophosphatase (IMPase) inhibition-increased autophagy clearance of amyloid produced by N-glycosylation pathway dysfunction. Together, lithium, magnesium, and zinc constitute a regulatory axis for protein N-glycosylation, in which lithium levels determine the threshold for pathology during magnesium and zinc depletion under stress. Magnesium is also required by the oligosaccharyltransferase complex, which catalyzes the conversion of dolichol-pyrophosphate-GlcNAc2-Mana9-Glc3 and a protein asparaginyl residue to an N-glycosylated protein and dolichol-pyrophosphate. Stress-induced Mg2+ deficiency impairs the oligosaccharyltransferase complex by affecting MAGT1 (magnesium transporter 1) and the STT3A component, resulting in N-glycan defects and hypoglycosylation. Endoplasmic reticulum stress, resulting from the unfolded protein response, stimulates N-glycan branching (bisected N-glycans) and hyperglycosylation due to the effect on TUSC3 (a regulator of Mg2+ influx) and the STT3B component of the oligosaccharyltransferase complex. Introduction Stress may be a proximal cause of dolichol pathway dysfunction resulting in N-glycosylation defects, the unfolded protein response in the endoplasmic reticulum, and inflammation leading to the progression of Alzheimer’s disease (Schloss 2025). Stress is a recognized risk factor for Alzheimer’s disease, one of the most common forms of dementia, characterized by dysregulation of protein folding in the endoplasmic reticulum and inflammation (Chan et al. 2026; Palacios et al. 2025). Lithium prevents the development of Alzheimer’s disease, as evidenced by clinical data and animal model studies (Aron et al. 2025). Stress causes a deficiency of two essential cofactors required by the dolichol pathway: niacin-derived pyridine nucleotides, essential for DHRSX and SRD5A3, and Zn2+, required by DOLK to form dolichol phosphate (Schloss 2025). The first committed step in protein N-glycosylation, DPAGT1, requires another cofactor, Mg2+, which is depleted under stress (Pickering et al. 2020), and dolichol phosphate (Schloss 2025). Small reductions in the rates of five sequential steps required for N-glycosylation, due to stress-induced dietary deficiencies, would collectively have a disproportionate effect on downstream protein N-glycosylation, proteostasis, and the pathology associated with the unfolded protein response in the endoplasmic reticulum, leading to inflammation and oxidative stress (Schloss 2025). It is reasonable to ask whether the effects of lithium, zinc, pyridine nucleotides, and magnesium are interrelated. We will present evidence that lithium can offset the consequences of stress-induced nutritional deficiencies in these three essential cofactors for N-glycosylation. Stress, Mg2+, Zn2+, and neural hyperexcitability Under stress, magnesium transport out of the cell increases, along with a concomitant increase in its urinary excretion rate (Pickering et al. 2020). Higher urinary magnesium levels are a hallmark of stress. Stress also causes zinc to be sequestered by metallothionein in the liver (Tian et al. 2014). Lower magnesium and zinc levels reduce their blockade of N-methyl-D-aspartate (NMDA) receptors and increase associated calcium influx, thereby increasing neuronal hyperexcitability (Guo et al. 2017). An influx of calcium into the cell increases glutamate-induced excitotoxicity and oxidative stress, which in turn leads to reduced levels of pyridine nucleotides (NAD(P) and NAD(P)H) via poly(ADP-ribose)polymerase (PARP)-dependent DNA repair (Høyland et al. 2024). The proximal effect of stress is to decrease the availability of intracellular magnesium and zinc. Chronic stress further reduces magnesium, zinc, and pyridine nucleotide levels. Because magnesium levels respond more strongly than calcium to stress, the serum Mg2+/Ca2+ ratio is likely to fall below 0.40 mol/mole, which is a clinical risk marker for a wide range of diseases (Nelson et al. 2025; Rosanoff et al. 2025). During times of stress, dietary insufficiencies in zinc, magnesium, calcium, and niacin can also result from reduced absorption in the gastrointestinal tract (Tao et al. 2013; Seelig 1994). The Li+-Mg2+-Zn2+ regulatory axis, GSK-3β, IMPase, and DOLK The pharmacology of lithium is largely defined by inhibition of two magnesium-dependent enzymes, glycogen synthase kinase−3β (GSK-3β) and inositol monophosphatase (IMPase), as shown in Fig. 1. Both enzymes require more than one Mg2+ per active site, two for GSK-3β and three for IMPase, with sigmoidal, second-order dependence on Mg2+ at low concentrations (Ryves & Harwood 2001; Leech et al. 1993). By contrast, DPAGT1, the first committed enzyme in N-glycosylation, requires only one Mg2+ per active site and shows a hyperbolic, first-order dependence on Mg2+ at low concentrations (Yoo et al. 2018). Under conditions of stress and magnesium depletion, GSK-3β and IMPase activation by Mg2+ would be expected to be disproportionately affected compared with DPAGT1, given their respective magnesium requirements and saturation kinetics. A quantitative evaluation of their relative activation states is complicated by the fact that although the total intracellular magnesium concentration is quite high (17–20 mM), only a small fraction (less than 3%) is free (0.4–0.5 mM), with most magnesium bound and partitioning between various chelators (Iotti and Malucelli 2011). Inhibition of GSK-3β and IMPase by lithium would further ensure that reduced magnesium availability reduces their activity more than DPAGT1-dependent N-glycosylation, since lithium competes with magnesium for activation of these enzymes but not DPAGT1. Activation of DOLK by Zn2+ presents an even greater challenge, since virtually all intracellular zinc is tightly bound, resulting in free zinc concentrations in the picomolar to nanomolar range. Although Zn2+ is optimal for DOLK activation, the enzyme shows some activity, at substantially lower levels, with other divalent metals, such as Mg2+ or Ca2+ (Sakakihara and Volpe 1985). So, if DOLK is activated by Zn2+ under normal conditions, it may operate at reduced levels under stress by switching to an alternate metal or by being less strongly activated by Zn2+. Even under normal conditions, a metallochaperone system may activate DOLK (Weiss et al. 2022). The activation of DOLK by available zinc may not depend on free zinc but be conducted at the expense of other intracellular zinc chelates.Fig. 1Increased levels of SRD5A3, DHRSX, DOLK, and DPAGT1, produced by lithium-stimulated transcription, and the NAD, NADPH, NADH, zinc, and magnesium cofactors required for catalytic activity Increased levels of SRD5A3, DHRSX, DOLK, and DPAGT1, produced by lithium-stimulated transcription, and the NAD, NADPH, NADH, zinc, and magnesium cofactors required for catalytic activity Lithium and N-glycosylation Lithium inhibition of GSK-3β reduces phosphorylation of β-catenin (a transcription factor in the Wnt pathway) and SREBP-1 (a Sterol Regulatory Element Binding Protein, transcription factor), thereby preventing their ubiquitination and proteasomal degradation (Fig. 1; Kim et al. 2004; Sengupta et al. 2010; 2013). GSK-3β is not the only magnesium-dependent kinase that induces proteolytic degradation of β-catenin and SREBP-1, since casein kinase can also phosphorylate both and increase the rate of their elimination, but is not sensitive to inhibition by lithium. By inhibiting GSK-3β, lithium increases steady-state DPAGT1 transcription and the rate of DPAGT1 protein production, offsetting the loss of activity due to stress-induced or dietary Mg2+ deficiencies. GSK-3β inhibition also likely elevates the transcription of DHRSX, SRD5A3, and possibly DOLK. Independent of the mechanism by which the elevated expression levels of these enzymes are obtained, lithium is known to increase the expression of all four genes (SRD5A3, DHRSX, DOLK, and DPAGT1: 24, 17, 16, and 11%, respectively) potentially overcoming activity losses due to stress-induced reduction in Mg2+, Zn2+, and NAD(P)/H (Sengupta et al. 2013; Sung et al. 2019: esbl-nhlbl.nih.gov). Lithium inhibition of IMPase would complement the increased rate of N-glycosylation provided by GSK-3β inhibition by eliminating misfolded amyloid proteins through an mTOR-independent enhanced autophagy pathway (Sarkar et al. 2005). Stress and N-glycosylation In addition to N-glycosylation of proteins essential to brain health and dementia prevention, which depends on proper dolichol pathway function (Schloss 2025), several other health-related issues depend on the lithium-magnesium-zinc axis for N-glycosylation regulation. The beneficial effect of lithium in preventing delayed wound healing (Huang et al. 2025), osteoporosis (Köhler-Forsberg et al. 2022), and immunodeficiency (Nagy et al. 2013) can be explained, in part, due to lithium decreasing the risk of stress- and diet-induced deficiencies in necessary cofactors for N-glycosylation, especially Mg2+ and Zn2+. A major pleiotropic reduction in the risk of all collagen-related diseases would share in the benefits of the higher-level function of the lithium-magnesium-zinc axis regulation of N-glycosylation. Collagen has N-glycosylation sites at its C-terminus that are not important under non-stress conditions but become essential for correct transport and localization of mature collagen under conditions of endoplasmic reticulum stress (Li et al. 2021). Also, the two ascorbic acid-dependent posttranslational modifications essential to proper collagen folding, hydroxylation of proline and lysine residues, depend on N-glycosylation of prolyl 4-hydroxylase and lysyl oxidase (Shi et al. 2019; Pirskanen et al. 1996). Without these essential N-glycosylation modifications, hydroxylations essential to the proper folding of collagen and the prevention of the unfolded protein response due to misfolded collagen in the endoplasmic reticulum are likely to occur. In addition to the correlation of key proteins subject to N-glycosylation dysfunction in Alzheimer’s disease (Schloss 2025), there are several examples of abnormal N-glycosylation linked to stress. Post-traumatic stress disorder accelerates aging by 15 years, as measured by N-glycosylation compared to control subjects (Moreno-Villanueva et al. 2013). A direct causal link is established between specific IgG N-glycan traits and severe psychiatric disorders, such as post-traumatic stress disorder (Lv et al. 2024). Acute and chronic stress deeply alter the hippocampal glycoproteome in animal models, promoting depressive behavior (Ren et al. 2025). Stress clearly contributes to N-glycosylation deficits, the progression of Alzheimer’s disease, and other disorders linked to induced nutritional deficiencies. Lithium can mitigate the consequences of stress-induced nutritional deficiencies by upregulating the synthesis of enzymes that depend on these nutrients. Lithium lowers the threshold below which the concentrations of Mg2+, Zn2+, and pyridine nucleotides (NAD(P)/H) contribute to disease progression. Collectively, this constitutes the Li+-Mg2+-Zn2+ regulatory axis for N-glycosylation and disease risk management. Hyperglycosylation and dolichol pathway dysfunction Reduction in the rate of N-glycosylation appears inconsistent with the observed hyperglycosylation, specifically N-glycosylation, as O-glycosylation and hyaluronidation levels are decreased, associated with Alzheimer’s disease and mild cognitive impairment (Hawkinson et al. 2026). However, a counterintuitive explanation for this inconsistency is possible. Overproduction of N-glycosylated proteins is consistent with a potential glycosylation rate that does not meet or exceed the rate of protein production determined by transcription and translation, so long as regulatory mechanisms to prevent this fail. Oligosaccharyltransferase (OST) complex fidelity also depends on Mg2+ levels in the endoplasmic reticulum, which are maintained by MAGT1 (magnesium transporter 1), resulting in glycosylation defects at low Mg2+ (Matsuda-Lennikov et al. 2019; Blommaert et al. 2019; Park et al. 2025). Mutations in the MAGT1 magnesium transporter cause hypoglycosylation of most proteins but also hyperglycosylation of some proteins (Matsuda-Lennikov et al. 2019). Under conditions conducive to Alzheimer’s disease, the OST complex will N-glycosylate proteins not normally used as substrates, such as tau (Wang et al. 1996). By contrast, stress-induced Mg2+ deficiency can cause global hyperglycosylation due to endoplasmic reticulum stress from the unfolded protein response and its likely effect on TUSC3 (a regulator of Mg2+ influx) and the STT3B component of the OST complex (Horak et al. 2014). The transcription and translation rates of a given protein respond to specific metabolic needs. If the rate of N-glycosylation of a protein in the endoplasmic reticulum is insufficient, then a partially N-glycosylated protein may be produced that lacks the functionality, localization, and/or the solubility necessary to meet the metabolic need for that protein’s activity to reduce the rate of its transcription. This can lead to an overproduction of partially N-glycosylated proteins that lack the requisite functionality to meet the metabolic need driving transcription. Protein glycosylation is segregated temporally and compartmentally, with N-glycosylation taking place coincident with translation in the endoplasmic reticulum and O-glycosylation occurring later in the Golgi, nucleocytoplasm, and mitochondria (Very and El Yazidi-Belkoura 2026). Defective N-glycosylation can lead to protein misfolding and aggregation of amyloid β, tau, and α-synuclein (Very and El Yazidi-Belkoura 2026). Downstream O-glycosylation of tau prevents aggregation and hyperphosphorylation (Liu et al. 2004). The ratio of potential N-glycosylation rate to the protein transcription/translation rate likely determines the fraction of N-glycosylation errors. If the N-glycosylation rate is only marginally below the rate of protein synthesis, the total error rate should be highest due to retention of incompletely or improperly N-glycosylated and misfolded proteins. Partially N-glycosylated proteins can also trigger the unfolded protein response (Mahl and Meyer 2026; Kozutsumi et al. 1988; Gething and Sambrook 1992), which induces transcription of inflammatory proteins (Gargalovic et al. 2006), many of which are themselves N- and O-glycosylated (Scapin et al. 2024; Goodall et al. 2010; Kohrs et al. 2025). Hyperglycosylation of GRP94 in response to partial inhibition of DPAGT1 by low doses of tunicamycin has been observed (Cherepanova et al. 2019; Dersh et al. 2014). The many hypotheses for the underlying cause(s) of Alzheimer’s disease In explaining protein hyperglycosylation; the cognitive effects of stress and stress-induced nutritional deficiencies; and the effect of lithium in reducing or reversing the risk of Alzheimer’s disease, dysfunction of dolichol phosphate-dependent N-glycosylation provides an integrated, non-siloed explanation for the underlying mechanism of Alzheimer’s pathology resulting from stress. Hyperglycosylation observed in Alzheimer’s disease (Hawkinson et al. 2026) and hypoglycosylation associated with congenital disorders of glycosylation (Dupré et al. 2000), such as mutations in PMM2 or SRD5A3, also result in cognitive dysfunction (Matthijs et al. 1997; Jain et al. 2026). A wide range of observations and explanations exist for the underlying pathology of Alzheimer’s disease. Some of these observations and proposed mechanisms for the pathology include: aluminum toxicity (Perl and Brody 1980); the cholinergic hypothesis (Davies et al. 1976; Bartus et al. 1982); altered dolichol levels in the postmortem brains of Alzheimer’s patients (DOLK deficiency) (Wolfe et al. 1985); the Tau/NFT hypothesis (Kosik et al. 1986); the glutamate excitotoxicity hypothesis (Greenamyre et al. 1987); zinc deficiency (Wallwork 1987); aberrant posttranslational modification: proteolysis, phosphorylation, glycosylation, or transglutamination (Caputo and Salama 1989); magnesium deficiency (Glick 1990); magnesium deficiency and stress (Kirov and Tsachev 1990); the synapse loss hypothesis (DeKosky and Scheff 1990); the amyloid cascade hypothesis (Hardy et al. 1992); the APOE and lipid dysregulation hypothesis (Strittmatter et al. 1993); early-onset familial Alzheimer’s disease (PSEN1) (Sherrington et al. 1995); early-onset familial Alzheimer’s disease (PSEN2) (Levy-Lahad et al. 1995); aberrant N-glycosylation of tau linked to Alzheimer’s disease (Wang et al. 1996); the oxidative stress hypothesis (Markesbery 1997); the metallobiology hypothesis (Bush 2003); the myelin breakdown hypothesis (Bartzokis 2004); the mitochondrial cascade hypothesis (Swerdlow and Khan 2004); the antimicrobial protection hypothesis (Soscia et al. 2010); the microglial/neuroinflammation hypothesis (Guerreiro et al. 2013; Jonsson et al. 2013); the neurovascular/blood–brain barrier breakdown hypothesis (Montagne et al. 2015); the gut-brain axis hypothesis (Cryan et al. 2019); the PANTHOS/autophagy hypothesis (Lee et al. 2022); dolichol pathway dysfunction (Schloss 2025); and hexosamine-induced pathology and hyper N-glycosylation (Hawkinson et al. 2026). Summary Stress-induced Alzheimer’s disease can best be explained by catecholamine-directed export of magnesium from the cell; cortisol upregulation of liver metallothioneins, reducing intracellular Zn2+; and cortisol and aldosterone inhibition of Mg2+ and Ca2+ reuptake in the kidneys. Reduced Mg2+ and Zn2+ suppression of the NMDA receptor cause pyridine nucleotide depletion through oxidative stress. Synthesis of dolichol phosphate and its N-acetylglucosamine conjugate are inhibited due to deficiency of the required cofactors: NAD, NADPH, Zn2+, and Mg2+. Lithium offsets the cofactor-limited inhibition of the dolichol phosphate pathway and dolichol pyrophosphate-sugar conjugate synthesis by increasing the transcription rate for genes of the required enzymes, through inhibition of GSK-3β. Mg2+ deficiency causes hypoglycosylation and mis-glycosylation of proteins primarily through inhibition of the MAGT1 and STT3A components of the OST complex, while endoplasmic reticulum stress, resulting from the unfolded protein response, stimulates global hyperglycosylation through its effect on the TUSC3 and STT3B components of the OST complex. Lithium mitigates the effects of the unfolded protein response by stimulating autophagy through inhibition of IMPase. It is tempting to speculate that lithium may also have additional, yet to be discovered, effects on the MAGT1, TUSC3, STT3A, and STT3B components of the OST complex.