Human depression: a new approach in quantitative psychiatry

Human depression in the interpretation of an artificial neural network

Following the theory that a biomolecular involvement of the cell could be an expression of a psychiatric disorder, we have tried to understand and explain this phenomenon.

The intention was to study the platelet fatty acids composition in normal and depressed subjects [1], because of their similarity to neurons [2–10].

Membrane platelet fatty acids of subjects with a clinical diagnosis of major depression versus apparently normal subjects were assessed. The complexity of membrane dynamics has also suggested study by means of non-linear advanced analytical tools would be appropriate. In particular, it seemed more appropriate to use artificial neural networks: the self-organising map (SOM) Kohonen network [11–13]. This particular algorithm allows viewing of the result graphically, building a two-dimensional map that places the subjects in a continuous, not necessarily dichotomised way.

The values for fatty acids in the two populations were entered into the SOM, mixing normal and pathological subjects and hiding the information relating to their pathological condition. The SOM was then used to map the two populations using three specific fatty acids: palmitic acid (PA), linoleic acid (LA) and arachidonic acid (AA), which represent the majority of total membrane fatty acids, recognising as similar those belonging to the same population and then separating the normal cases from the pathological cases [1]. All the artificial neural networks (ANNs) tested gave essentially the same result. However, the SOM gave superior information by allowing the results to be described in a two-dimensional plane with potentially informative border areas. The central property of the SOM is that it forms a non-linear projection of a high-dimensional data manifold on a regular, low-dimensional (usually 2D) grid.

This experiment was performed outside of evidence-based medicine (EBM) rules. The direct task of finding biomarkers according to the EBM rules requires the elimination of selection bias, and in psychiatric illness the leads to the selection of a population that is often clinically unrealistic. The results are shown in Figure 1a, b.

The SOM has shown considerable correlation to the clinical diagnosis of major depression, and indeed, revealed the existence of differences within the same diagnosis. The literature shows that a diagnosis of major depression is very often misleading, and can be changed to a diagnosis of bipolar disorder [14].

Using the following equation (Equation 1), which relates each fatty acid percentage with the melting point and the molecular weight, we obtained a result that led us to understand that platelet membranes had different degrees of viscosity and/or unsaturation (B2 index).

Where A _i = percentage of i-th fatty acid, mp = melting point, mw = molecular weight, mw _i = molecular weight of i-th fatty acid, mp _i = melting point of i-th fatty acid, and i:

The result clearly showed that the platelet membranes of depressive subjects were characterised by a much higher degree of fatty acid unsaturation than normal subjects.

According to Donati et al. [15] rapid changes in membrane lipid composition or in the cytoskeleton could modify neuronal signalling. As this could have implications for a new understanding of some aspects of psychiatric disorders, a private meeting was organised in Bologna (Faculty of Veterinary Medicine) and in Treviso, University, October 2008) with some expert scientists in the field (Kary Mullis and Stuart Hameroff).

Three essential points constituted the crucial elements of the discussion at the meeting: (1) the viscosity of the platelet and neuronal membrane; (2) the protein Gsα; (3) the relationship between tubulin and consciousness.

With regard to the first point, Cocchi and Tonello observed that the platelet membrane was substantially differentiated from a chemical point of view with regard to the indexes of saturation between depressed and normal populations [1].

On the second point, the protein Gsα modifies its structure according to the degree of viscosity of the neuronal membrane, as seen in patients who commit suicide for psychiatric reasons in comparison to deaths from other causes [15].

With regard to the third point, Tubulin, because of its connection to Gsα and its position in the cellular cytoskeleton, determines those changes that have been assessed with quantum computation under conditions of wakefulness in comparison to the condition of anaesthesia [16].

Biomolecular depression hypothesis

A very suggestive hypothesis was built, as summarised in Figure 2.

Figure 2 describes the molecular depression hypothesis formed according to the experimental findings of Cocchi et al. [1], Donati et al. [15], Hameroff and Penrose [16, 17] and Hameroff [17]. Because of the possible similarity in behaviour of platelets and neurons, membrane viscosity may therefore modify the Gsα protein status. The Gsα protein is connected with tubulin. Tubulin, depending on local membrane lipid phase concentration, may serve as a positive or negative regulator of phosphatidylinositol bisphosphate (PIP2) hydrolysis, as G proteins do. Tubulin is known to form high-affinity complexes with certain G proteins. The formation of such complexes allows tubulin to activate Gα protein, which, in turn, can activate protein kinase C (PKC), and fosters a system whereby elements of the cytoskeleton can influence G-protein signalling. PKC activation (Figure 3) is preceded by a number of steps, originating from the binding of an extracellular ligand that activates G-protein on the cytosolic side of the plasma membrane [18].

The G-protein, using guanosine triphosphate (GTP) as an energy source, then activates PKC via the PIP2 intermediate, the diacylglycerol/inositol triphosphate (DAG/IP3) complex [15]. The schematic biomolecular mechanism of the Gsα protein is described in Figure 4.

The Gα subunit is activated and starts a cAMP signalling cascade, as shown in Figure 5.

The international scientific literature has reported abnormalities in the cAMP signalling cascade of the human brain in suicidal and depressive subjects for over two decades [19–25].

According to Donati et al. [15] there is a further possible condition: the position of Gα (Gsα in particular) within the lipid raft microdomain. Lipid rafts are specialised structures on the plasma membrane that have an altered lipid composition as well as links to the cytoskeleton (Figure 6). They are local lipid microdomains that float in the liquid-disordered lipid bilayer of cell membranes. The effect of lipid rafts on neurotransmitter signalling has also been implicated in neurological and psychiatric diseases [26].

Raft localisation of Gsα in human peripheral tissue (possibly platelets, see [15]) may thus serve as a biomarker for depression. Several studies using human platelets suggest that adenylyl cyclase may, in fact, serve as a biological marker for depression [27–34].

The membrane fatty acid-Gsα hypothesis

It is known that G proteins could be targeted to raft domains by several mechanisms. The most plausible mechanism is that Gα subunits are subject to palmitoylation. Palmitoylation is a process of covalent attachment of palmitic acid to cysteine residues of membrane proteins.

Palmitic acid is one of the three fatty acids (together with arachidonic acid and linoleic acid) used by SOM as marker of depression [1].

Is the critical composition of the membrane platelet fatty acids an indirect measure of the G protein status? (see Figure 7.)

Rapid changes in membrane lipid composition or in the cytoskeleton might modify neuronal signalling. Hameroff hypothesised that through this mechanism it is possible to modify the consciousness state [16, 17]. According to Hameroff [16, 17] the best measurable correlate of consciousness is a γ synchrony electroencephalogram (γ waves are a pattern of brain waves, with a frequency between 25 to 100 Hz, prototypical at 40 Hz), which indeed rapidly moves and redistributes throughout the brain. γ Synchrony derives not from neurocomputation, but from groups of neuronal dendrites (and glia) transiently fused by electrical synapses called gap junctions, more or less sideways to the flow of neurocomputation. The process could be mediated by tubulin and its correlates i.e. membrane viscosity and Gsα protein (see Figure 2).

Recent studies reported a model of the disconnection hypothesis of schizophrenia through the demonstration of abnormal stimulus induced γ phase synchrony [35].

The idea discussed by the authors with Hameroff and Mullis that platelets could represent the peripheral markers of the depressive disorder and that platelets are 'brain ambassadors', has become a more and more realistic proposal [36].