The Molecular Basis of Memory

Abstract

We propose a tripartite biochemical mechanism for memory. Three physiologic components are involved, namely, the neuron (individual and circuit), the surrounding neural extracellular matrix, and the various trace metals distributed within the matrix. The binding of a metal cation affects a corresponding nanostructure (shrinking, twisting, expansion) and dielectric sensibility of the chelating node (address) within the matrix lattice, sensed by the neuron. The neural extracellular matrix serves as an electro-elastic lattice, wherein neurons manipulate multiple trace metals (n > 10) to encode, store, and decode coginive information. The proposed mechanism explains brains low energy requirements and high rates of storage capacity described in multiples of Avogadro number (N_A = 6 × 10²³). Supportive evidence correlates memory loss to trace metal toxicity or deficiency, or breakdown in the delivery/transport of metals to the matrix, or its degradation. Inherited diseases revolving around dysfunctional trace metal metabolism and memory dysfunction, include Alzheimer's disease (Al, Zn, Fe), Wilson’s disease (Cu), thalassemia (Fe), and autism (metallothionein). The tripartite mechanism points to the electro-elastic interactions of neurons with trace metals distributed within the neural extracellular matrix, as the molecular underpinning of “synaptic plasticity” affecting short-term memory, long-term memory, and forgetting.

Background

Biologic memory in the brain is a mystery. Various adjectives have been used to describe memory (i.e., active, declarative, passive, associative, short-term (STM), long-term (LTM), super memory), but none in molecular terms. No consensus exists for how cognitive information (cog-info) is encoded or stored in the brain.

Scientists from disparate disciplines have suggested various molecular mechanism, such as DNA/RNA-based processes, to describe memory. Neuroscientists proposed neural firing patterns, neurocircuits, neural-networks, neurotransmittors, and synaptic firing as the basis for encoding sensory perceptions as memory.¹⁻²² The “synaptic plasticity” model is unsatisfactory from the perspective of compactness, kinetics, energy requirements, and lack of an information theory.²¹ What is missing is a physiologically relevant, molecular mechanism, whereby cog-info derived from the senses can be encoded, stored, and recalled by the neural system.

In computer ionic memory chips,²³⁻³⁸ information is received, processed, and stored by manipulating the distribution of elemental cations (dopants) within the chip matrix, usually solid electrolytes (metal sulfides, Ge-based chalcogenides, or oxides such as TaO₃, WO₃, SiO₂, TiO₂). Ionic memory chips are doped with Ag, Cu, and Zn . The information theories and binary value (0, 1) algorithms developed by von Neumann, Turing, Weiner, and Shannon are used to encode digital information in the memory chip.³⁹⁻⁴⁹

Neural Memory Traits

The ionic memory chip is a compelling model for how one would like to describe biological memory in the brain. The characteristics and traits that one wishes to describe include the following:

• a)A credible mechanism for memory, based on generally accepted biochemical principles, with available physiologic components.
• b)Molecular-scale encoding/decoding process, faster than the rate of neural firing (<100 ms).
• c)Large storage capacity for physically encoding cog-info.
• d)Low energy requirements (<400 cal/day).
• e)Capacity for storing cog-info for short and long durations (recall: 1 min to >1 day).
• f)Capable of acquiring (learning), storing, and losing (forgetting) cog-info.
• g)Applicable to all animals with neural circuitry.

Hypothesis: The Tripartite Mechanism of Memory

We propose that “memory” emerges from the dynamic interaction of three physiologic compartments, consisting of the following:

• 1)Neurons.
• 2)Neural extracellular matrix (nECM) around the neuron.
• 3)Trace metal cations dispersed within the nECM (dopants).

The nECM constitutes a hydrated lattice wherein cog-info is encoded, processed, and stored. The neurons manipulate and electro-elastically sense the surrounding ensembles of [nECM/metal] complexes, to encode and recall memory. More specific descriptions of these compartments follow.

1. Neuron⁵⁰⁻⁸⁴

The neuron and neural circuits, the computational components of the brain, operate by electrical signaling within an aqueous environment. The cells are intimately connected to the external nECM by electrically sensitive surface features, notably integrin receptors, gap junctions, nodes of Ranvier, and synapses. The electro-elastic contact between the neural surface and its external environment (nECM) is the keystone to neural computation.

2. nECM⁸⁵⁻¹³⁵

The nECM surrounding the neurons comprises 20–25% of the total brain volume. It exhibits gross- and microanisotropies in terms of ultrastructure, composition, and dielectric properties. It comprises a block copolymer lattice around the neuron, composed of polymeric glycosaminoglycans (GAG) (hyaluronic acid (HA), lecticans, proteoglycans, chondroitin sulfates (Chon), heparan sulfate (Hep)), which present many anionic/Lewis base moieties for attaching metal cations. Interspersed within the GAG lattice are also many proteins (e.g., collagens, integrins, tenascins, phosphacan, various enzymes) and glycoproteins (e.g., nidogen, reelin, TNC, thrombospondin).

The nECM lattice is an electroelastic hydrogel, characterized by viscoelastic traits, carrying currents on the order of 200 nAmp, with conduction velocities 2–10 m/s, accommodating voltage changes exerted at neuronal synapses and other anatomical sites (i.e., gap junctions, nodes of Ranvier).

We propose that the complexation of a metal cation to specific chelating groups (nodes) within the nECM concomitantly modulates the nanoscale structural, viscoelastic, and dielectric properties of the lattice, sensed by the neuron.

Thus, the nECM serves two functions:

• •As the structural scaffolding encasing the neurons, through which gases (oxygen and carbon dioxide), metals, and metabolites diffuse.
• •As an electro-elastic lattice used by the neuron to encode and decode cog-info, by modulating or sensing the pattern of metal cations bound to specific nodes (addresses).

3. Trace Metals in the Brain^136−153

Brain levels of more than 15 trace metals have been measured within the gross tissue as well as within the individual neurons. Metal levels were highest within the neuron, but were present in the nECM, at levels ranging from 10^–6 to 10^–9 M (Table 1). Table 1 shows the composition of total human brain tissue, presenting the 15 most prevalent elemental metals. Most (>90%) are sequestered within the neurons, with <10% found in the nECM. The distribution of trace metals is not homogeneous between and within anatomical regions of the brain.

Table 1. Brain Levels of Elements.

metal	∼value	unit	+ valence
K	3.4	M	1
Na	2.7	M	1
Mg	0.3	M	2
Ca	60	mM	2
Fe	10	mM	2/3
Co	6	mM	2/3
Zn	6	mM	2
Cu	3	mM	2/1
Rb	3	mM	1
Li	∼ 1	mM	1
Sc	370	uM	3
Mn	211	uM	2/3
Cr	153	uM	2/3
Al	14–20	uM	2/3
Cd	18	uM	2
Pb	3	uM	2
Hg	0.5	uM	1

nECM/Metal Complexes: Conformations, Kinetics, and Energetics^154−165

Parameters, such as metal chelate dissociation constant (K_d), solubility constants of salts (K_sp), or molar binding ratios, reflect the inherent binding proclivities of elemental cations for the various anionic moieties (such as carboxyl, sulfate, amine, hydroxyl, phosphate, and other electron-rich groups), within the nECM lattice encasing the neuron. The anionic and electron rich moieties anchor the metal to specific nodes on the lattice, serving as “addresses” for encoding the cog-info.

The inherent bonding geometries of the elemental metals range from square planar (d²sp³), tetrahedral (sp³), to octahedral (d²sp³), with individual bonding distances for each. Thus, each elemental oxidation state presents a unique bonding “signature”. The polyanionic nECM can flex to achieve the most energetically favorable metal complexes.

The disposition of different metal cations within the nECM is modified by iontophoretic and electro-elastic effects, exerted or sensed by the neurons. The binding/desorption of hydrated metal cations to/from hydrated anionic substrates are among the most rapid biologic reactions, requiring low energy of activation (E_act) and generate little heat (low ΔH). The low levels of various elements (millimolar to nanomolar (10^–3 to 10^–9 M) concentration) further minimize heat generation.

nECM as “Information Lattice”

Consider that the nECM around each neuron as a three-dimensional lattice capable of binding more than 15 different mobile components (elemental cations) as arrays/clusters/packets/stacks of almost limitless, combinatorial complexity.

The diversity of the nECM/Metal is enormous. It arises from the following:

• 1)Many elemental metals: various metals are present in the brain, each binding with a unique binding configuration, depending on its electron shell disposition. Some elements can exist in more than one oxidation state (i.e., Cu^+1/2, Fe^+2/3, Mn^+2/4/5).
• 2)The nECM is a complex lattice of polymeric components, each which present multiple anionic and electron rich moieties. These moieties can entrap metal cations; the lattice flexes to achieve the optimal disposition of groups of anionic moieties (variable size chelate rings) to accommodate the bonding preferences of the various elemental cations. Stable configurations encode cog-info.

Metal binding to a particular nECM locale (address) results in conformational changes (flexing, contraction) of the local lattice geometry, with resultant modifications in the local neural membrane polarizabiltity/resistivity/elasticity sensed by the neuron. The [nECM/metal] complexes comprise configurable molecular switches by which neurons encode/decode cog-info.^166−196

The computational possibilities are astronomic. The metal-binding capacity of the nECM around each neuron reflects the molar equivalents of anionic moieties, multiplied by the Avogadro number, N_A (6 × 10²³). Thus, the molar cog-info storage capacity is very large (multiples or exponentials of Avogadro’s number (N_A = 6 × 10²³). If only a fraction of the elemental cations in the nECM function in a combinatorial mode, there are enough to serve as mobile components for encoding/decoding and processing large amounts of cog-info, on which memory is based.

It is interesting to compare the operation of computer ionic memory chips with the biological mnemonic system, as in Table 2.

Table 2. Comparing Information Processing.

item	computer	brain
unit component	ionic chip	neuron
matrix composition	solid state matrix	nECM hydrogel
dopant(s)	1 elemental cation	n elemental cations (n > 10)
construction	static hardwiring	synaptic neural network
computational format	digital	analogue
information unit	bit/word	cuinfo
# coding options	n = 2 (0,1)	n > 10
programming mode	serial2	paralleln
groupings	dedicated circuits	sparse neural ensembles
underlying physics	electro-optic/magnetic	electro-elastic/chemical
read/write driver	voltage differential	iontophoresis, chelation
signal speed	c (speed of light)	1–80 m/s
signal frequency	50–60 Hz	2–70 kHz
energy	external (∼225 W·h)	metabolism (22 W·h)

Discussion

The brain contains 10¹⁰ to 10¹¹ neurons, which can each have 10⁴ excitable synapses. The histologic tissue sections by Cajal¹⁹⁷ and Golgi¹⁹⁸ (Figure 1) more than 100 years ago, which revealed neurons with synaptic contacts, were based on a selective staining process to image the neuron, but ignored the surrounding matrix. Since then, the neuron has been generally represented as a cell suspended in an “invisible” context or “space”, like an interplanetary object. How can one explain the function of a ship without reference to water?

Figure 1.

(A) Drawing of a single neuron by Cajal, based on Golgi’s silver stain technique which ignored the nECM and (B) with superimposed image of the [nECM/metal] array, lightly stained for everything . With many histologic stains, one cannot see the neuron for the “trees” of the nECM.

Biochemists know that the “space” around the neuron is a reticulum composed of polyanionic biopolymers called the nECM.⁸⁵⁻¹³⁵ We point out that the neuron is encased in a complex, nonhomogeneous nECM which functions both as a structural environment for the neuron, as well as a computational matrix wherein it encodes and stores cog-info.

Stacks and arrays of [nECM/metal] engulf the neuron with a continuum of stoichiometries, constituting the molecular correlates of cog-info, on which memory is based. In support of the tripartite mechanism of memory, we cite the following observations:

• 1)The literature describes effects associated with metal toxicity or deficiency, in terms of memory loss, as well as associated behavioral perturbations (confusion, disorientation, poor learning, personality changes) (Table 3).^199−210

  These all indicate that perturbations of the optimum concentrations of elemental metals, either by excess or low levels, are manifest by clinically observed changes in behavioral processes, ascribed to dysfunctional memory.

• 2)
  Metabolic disorders related to metals and memory:

  • •
    Alzheimer disorder (aluminum, zinc).
  • •
    Wilson’s disease (copper).
  • •
    Thalassemia (iron).
  • •
    Treatments: Li salts, zinc salts.
• 3)
  Chelation drugs that effects memory (aspirin, EDTA, deferrioxamine, penacillamine):

  • •
    Binding of zinc by drug disrupts hippocampal-dependent spatial-working memory.
  • •
    Chelating treatment correlated with performance tests of abstract reasoning... memory.
  • •
    Iron chelators used to treatage-related memory dysfunction.
  • •
    Aspirin use associated with greater prospective cognitive decline on select measures.
    The ameliorative effects of chelation drugs or zinc salts suggest that optimal levels of free (diffusible) metals in the brain underlie the mechanism of memory.
• 4)
  Metallothioneins (MT; 4 isotypes) function to transport metals throughout the body, notably the brain:^211−224

  • a)
    Knockout mice (KO) with deletions of MT-1 and MT-2 showed poorer rates of learning, evidence of poor working memory. (224)
  • b)
    MT dysfunction has been correlated with the following diseases or problems:

    • •
      Autism (a disorder of neural development characterized by impaired social interaction and communication, and by restricted and repetitive behavior).
    • •
      Behavior control and development of memory and social skills.
    • •
      Obsessive compulsive disorders (from Web site: Herb-Discovery.com).
• 5)Zinc transporters (ZnT; 4 types):^225−231

  ZnT3 KO mice have complete deficits in contextual discrimination and spatial working memory. Such animals were used to demonstrate that ZnT3 is involved in associative fear memory and extinction.

• 6)Tenascins C and R^{114−129,133,135,232−242} are protein components of the nECM which express “fibrinogen-knobs” including haptide epitopes. KO mice, which were incapable of synthesizing tenascins, exhibited behavioral abnormalities associated with memory dysfunction.
• 7)Chondroitinase (an enzyme which selectively degrades chondroitins) was injected into the mouse brain, resulting in the loss of fear-driven responses. This demonstrated that learned traumatic fear memory, which usually lasts a lifetime, is located in the degradable nECM. (109)
• 8)Histology revealed that human brain tissue comprises significantly more nECM between neurons than chimpanzees.¹⁰¹ We interpret this as indicating increased memory capacity for superior cognitive ability (e.g., language, memory).

Table 3. Metal Correlations with Memory.

metal	levels	behavioral changes
aluminum (Al)	toxic	memory loss, altered behavior, confusion, disorientation (see Alzheimer's disorder)
calcium (Ca)	deficiency	severe intellectul changes, mental retardation, poor memory
copper (Cu)	tissue overload (inherited)	Wilson’s disease; psychiatric manifestations
iron (Fe)	deficiency	poor memory; dietary iron supplemetntation correlated with improvement of memory
iron (Fe)	tissue overload (inherited)	thalassemia; anxiety, depression, psychiatric disfunctions
lead (Pb)	toxic	mental deterioration, aggressive, poor memory, lower IQ
lithium (Li)	therapy (high dose)	memory improvement, mental slowness (variable reports)
magnesium (Mg)	toxic	mental confusion, impaired memory
mercury (Hg)	toxic	loss of memory, behavioral changes
thorium (Th)	toxic	mental confusion
zinc (Zn)	deficiency	loss of memory

Conclusion

We propose that the nECM, in combination with diffusible metal cations, is the locus wherein the neurons encode basic, molecular correlates of cog-info. The minimal cognitive unit of information (cuinfo) corresponds to the formation of a single metal-complex, with one or more metal cations trapped at specific chelating nodes of the nECM (presented in Figure 2).

Figure 2.

Schematic representation of cuinfo* formed with M = 1, 2, or 4 metal cations per unit. Such catenary metal complex units are formed within the chelating node of the nECM, as molecular correlates of cog-info. At least one metal atom is required, though more could be involved in forming the minimal cuinfo ensemble.

The cuinfo is equivalent to the “bit” of the computer chip. Instead of representing information linearly with only two parameters (0 or 1), the neuron/[ECM/metal] complex operates with many ionic mobile components (n > 10) constrained within a flexible lattice, providing very large information storage and parallel processing capabilities.

Short-term storage of cog-info can employ trace monovalent elemental cations Li⁺, Rb⁺, and Cs⁺ (excluding Na⁺ and K⁺ which are present in much higher levels to generate the intraneural high voltage potentials). The monovalent [nECM/M⁺¹] complexes are not very stable, and could be expected to decay rapidly, manifest as short-term memory (STM).

Longer term storage of cognitive information units (derivative cuinfo) would employ polyvalent cations [nECM/M^+2/3/4/5] to form more stable complexes. Cross-linking, by enzymes (transglutaminases) or free radical reactions, stabilizes the cuinfo (metal complexes), appropriate for long-term memory (LTM).

When the [nECM/cation] arrays become degraded, the cog-info encoded therein also decays, manifesting as memory loss (storage failure). Of course, breakdowns at any critical point of the neural network chain (circuit failure) are also manifest as memory loss.

To conclude, we posit that:

• •Memory is based on tripartite interaction of neurons, nECM, and trace elements.
• •The tripartite mechanism involves low energetics with high speed/computational capabilities.
• •Cog-info is encoded by the neuron as cuinfo, like bits in memory chips
• •Degradation of nECM or metals excess/deficiency correlates with memory loss.
• •Cited experimental observations support the proposed tripartite mechanism.

Just like other metabolic processes, man and animals share the biochemical basis of memory. The tripartite mechanism operates in all animals with brains, albeit at increasing degrees of complexity, coincident with the increasing complexity of the anatomical subunits arising from evolutionary development.

Of course, much clarification is required to discern the workings of this tripartite mechanism. A formalism is lacking which elaborates on how sensory input (cog-info) is encoded via the distribution of n-metals within the nECM enveloping the neurons. Detailed metabolic, viscoelastic, and dielectric characterizations of the nECM with various elemental cations would clarify the nanoscale modifications employed by neurons to encode cog-info. Nonetheless, we identify the key physiologic compartments and suggest a credible biochemical mechanism for the phenomenon of recall.

The authors declare no competing financial interest.

This paper was published on the web on August 15, 2012, with the August 2012 issue. The blank reference 189 has been deleted (and subsequent references renumbered) in this online version published August 31, 2012, and an Addition and Correction appears on the Web on August 31, 2012, and in the September 2012 issue.