miRNA, Trinucleotide Repeats, Glutathione Status and the Genetics of IQ

Volkmar Weiss


Until now, more than 1 000 000 SNPs have already been investigated for an association with IQ by Plomin, Deary et al. and others. No major effect has been discovered which could explain the high heritabiltiy of general cognitive ability and the pattern of Mendelian segregation of IQ in the normal range of variation. Plomin et al. found only very small and mostly non-replicable effects. Therefore, our mind should be open for a broader outlook and new hypotheses.

MicroRNAs as regulatory factors in gene expression renders them attractive candidates for harbouring genetic variants with effects on IQ. There is already ample evidence that miRNA-mediated gene regulation plays an important role in a number of neurodegenerative diseases. MiRNAs bind to complementary sequences in the three prime untranslated regions (3' UTRs) of target messenger RNA transcripts. MiRNA genes are found in intergenic regions or in anti-sense orientation to genes and contain their own miRNA gene promoter and regulatory units. As much as 40% of miRNA genes may lie in the near-gene introns of protein and non-protein coding genes or even in exons.

Our first example:

Harold, D. et al., Nature Genetics 41 (2009) 1088ff. have found out and replicated compelling evidence that rs3851179 of the PICALM gene is associated with alzheimer. The homozygotes and heterozygotes of the rare allele A of rs3851179 have a 0,86x decreased risk for alzheimer. - Since many years I am looking for such findings, because we should expect that probands of high general intelligence (high IQ) have a later onset of alzheimer and a decreased risk. Surprisingly, the allele frequencies not only of rs3851179, but even more of rs669556 and of at least 30 other SNPs in this near gene region exhibit the frequencies of a hypothetical major gene locus of general intelligence, see http://www.v-weiss.de/majgenes.html and http://knol.google.com/k/national-iq-means#

As it seems, at the moment, nobody has an explanation why the non-coding SNP rs3851179 is associated with alzheimer and why in a large chunk of DNA with copy number variation a high number of SNPs exhibits similar allele frequencies in all the populations of the HapMap project. In which way could such a phenomenon have been stabilized by natural selection? Could this region be coding for miRNA or be its binding target?

The second example:

By checking routinely the bibliographical details of paper published together with A. Payton et al. on "Investigation of a functional quinine oxireductase (NQO2) polymorphism and cognitive decline" in  Neurobiol. Aging 31 (2010) 351,  I became aware of a publication on "Genetic variant
of glutathione peroxidase 1 in autism" Brain Dev. 32 (2010) 105

Since 1982 I did collect evidence on a relationship between glutathione peroxidase activity, general cognitive ability (IQ) and social status (for example, measured by years of education as a good surrogate).  I quote in the following from an editorial published by me 1994 in the journal "Intelligence":

"In 1982 I became aware of a paper published by Sinet, Lejeune & Jerome (1979) in which a correlation of .58 between IQ and erythrocyte glutathione peroxidase activity (GSHPx, now GPX1) was reported for 50 trisomy 21 patients. None of the other enzymes studied correlated with IQ. Sinet et al. thought the correlation to be trisomy-specific, because an increase of about 50% in the superoxide dismutase activity (SOD-1) can be observed in cells from trisomy 21 patients. There is a feedback control of GSHPx concentration by the amount of superoxide, which explains the elevated activity of GSHPx in cells of trisomy 21 patients. However, Fraser and Sadovnick (1976) had found that the correlations of IQ between trisomy 21 probands with their fathers, mothers and sibs are about .50, consequently of the same size as with healthy  children despite the mean IQ of trisomy 21 probands being about 70 points lower. Therefore already Lenz (1978) had concluded that individual differences in trisomy-IQ have generally the same biochemical background as in normal persons. And Brugge et al. (1992) confirmed a correlation of .73 between erythrocyte GSHPxactivity anda short-term memory score. ...  By Gerli et al. (1984) GSHPx was assayed in families and the results support the existence of two Mendelian alleles.?  For the full text see http://www.v-weiss.de/intellig.html

In the following years I instigated a number of colleagues from all over the  world to discover the underlying genetic cause of the cited correlations, but nearly completely in vain. We investigated SNPs of GST transferases, NQO and many, many others. Therefore I came to the conclusion that the major contribution to the correlation between lipid peroxidation and IQ could or should be the effect of a gene with copy number variations and repeat polymorphisms for which data are still not available or incomplete.

In April 2010, at the present state of knowledge, GPX1 is such a gene for which at least 4 common frameshift polymorphisms are already known. The arguments in favor of a relationship between trinucleotide repeat polymorphisms of GPX1 and general cognitive ability are holding for GCLC and its GAG-repeat polymorphism, too. GCLC influences the glutathione status of an individual to a high degree. The GCLC repeat polymorphism with its known large population differences in allele frequences is located in the three prime untranslated region (3' UTR) and therefore a likely binding target of a miRNA! In view of the claimed relationship of GSH/GSSG redox status with schizophrenia and other neurodegenerative diseases, we should be eager to see whether a correlation between GCLC GAG-repeat status and general cognitive ability can be confirmed or not, see

As I see, new publications and reviews on the relationship between glutathione status and brain function are not aware of some older publications. In the appendix ?Memory as a Macroscopic Ordered State by Entrainment and Resonance in Energy Pathways?, pp. 201-221, of the monograph ?Psychogenetik der Intelligenz?. Dortmund. Verlag Modernes Lernen 1986, I wrote on p. 210ff:

?It would defy the most fundamental laws of thermodynamics, when individual differences in brain power would not find their counterpart in individual differences of brain energy metabolism. ? Reactions involving S-S or S-H groups of proteins may readily account for the apparently opposite effects of the same control mechanism.  ? . At this point we direct attention to the correlation (.58) between IQ and glutathione peroxidase (GSHPx, now GPX) activity (SINET et. 1979) ? In modulating the GSH/GSSG ratio, GSHPx not only contributes to the regulation of glycolysis (GILBERT 1984) but consequently also the adenylate energy charge (REHNCRONA et al. 1980) and the NADP/NADPH ratio (GRIMM 1978) are prefectly correlated (r = 1.00!) with glutathione status. Thus the fundamental chemical needs of a living cell, high-energy phosphate stores (ATP) and reducing power (NADPH) depend upon the cortical concentration of glutathione, and the dynamic behavior of a complex system can be reduced to the molecular properties of a master enzyme in an energy pathway.?

The combination of autozygosity mapping and microarray RNA expression analysis has led to the discovery of new genetic polymorphisms underlying nonsyndromic mental retardation with autosomal-recessive inheritance, see http://www.medicalnewstoday.com/articles/174204.php  and
We can be convinced that the application of similar methods to consanguineous families with several members in the high IQ range will lead to the discovery of gene polymorphisms underlying variability of IQ in the upper and normal range of the distribution.  Sites where miRNA is coded or binding and these are especially regions with trinucleotid repeats in the 3?UTRs as GCLC should be investigated as soon as possible.

For all such neurodegenerative diseases as schizophrenia, autism, alzheimer and so on general cognitive ability (IQ) has to be seen as a major confounding variable, and nobody will obtain clearcut results in the genetics of neurodegenerative diseases as long as the genetic background of IQ in the normal range of variation remains unknown.

I thank my eldest daughter, Dr. Cornelia Weiss-Haljiti, for helpful discussions, Easter 2010.

The Advent of a Molecular Genetics of General Intelligence


Published in: Intelligence 20 (1995) 115-124 (Editorial), see also the postscripts (2001 and 2003)

Volkmar Weiss

The approach of the IQ Quantitative Trait Loci Project by Plomin et al. (1994) is criticized. Raw scores of IQ test do not exhibit the bell curve, only the normalized scores. Even a bellshaped distribution does not need a large number of underlying genes. Family data are in agreement with Mendelian segregation at a major gene locus of IQ, responsible for the regulation of brain energy metabolism. Therefore, the background of the correlation between glutathione peroxidase activity and IQ should be disentangled, applying methods of molecular genetics.

In 1980 Jensen was of the opinion: "The genotype is itself a theoretical construct. No one can look at a genotype for intelligence under a microscope or isolate it in a test tube" (p. 183). The rapid progress of molecular genetics during the last years has made it possible that the first research report on the molecular genetics of IQ has been published in 1994 by Plomin et al.. This editorial will criticize the approach taken, but with the only aim to make progress in this area of research even more rapid and more straightforward. Since an exhaustive critique of the very popular myth of the bell curve and its underlying genetics has already been published (Weiss, 1992a), we will restrict ourselves here to the most pertinent points.


Plomin et al. (1994) state: "General cognitive ability (intelligence, often indexed by IQ scores) is one of the most highy heritable behavioral dimensions. ... General cognitive ability ... is a quantitative trait with a roughly normal distribution." Of course, normalized by psychologists and definition, respectively. An IQ of 140 compared with an IQ of 70 suggests the double amount or the half of cognitive ability, respectively. However, a look of the raw scores, i.e. of the non-normalized scores, of IQ subtest shows that IQ 140 means an about fourfold amount of cognitive ability compared with IQ 70 (see Table 1). This relationship holds under the condition that all subtests are elementary cognitive tasks and the superior speed of the most intelligent is not clouded by a ceiling effect.

Table 1. Not-normalised raw scores in the subtests 3+4+7+8+9 of the mental power test LPS for some selected occupations

Mathematically gifted (tested by Weiss, 1979a)


Production engineers




















Data from Horn (1962)

Raw scores of IQ tests do not exhibit the wellknown bellshaped distribution curve, only the normalized scores.

About 1970 geneticists became aware that genotypes, who could be separated qualitatively by electrophoresis of human blood (Harris, 1966), could exhibit quantitatively a normal distribution of enzyme activities. The three alleles A, B and C underlying this distribution (see Fig. 1), fulfill the conditions of Mendelian segregation at one locus. The heritability for this distribution is 0.82 (Eze, Tweedie, Bullen, Wren & Evans, 1974). Now, in 1995, dozens of such cases of normal distributions with a small number of known underlying genotypes could be cited.

A normal distribution does not need a large number of underlying genes. High or low heritability of a variable is no indicator of the number of underlying genes.

A well-confirmed relationship holds between total values and variances of IQ scores: The higher the IQ, the less the variance. The correlation is such a strong one, that the IQ could even be measured only by the scaled variance of repeated tests. Intuitively, we would even predict the reverse relationship: Someone who is solving more elementary cognitive tasks per unit time, should make more errors. But this is not the case. We do not know any simple theory in all of psychology explaining this relationship. However, genetics offers such a theory: In such genetic polymorphisms as in Fig. 1 the mean m2 for the heterozygotes (for example, BA) is intermediate between the means m1 and m3 for the two homozygotes (AA and BB, respectively).

m2 = (m1 + m3)/2

In many genetic polymorphisms the standard deviations are directly proportional to the means (or si/mi, the coefficient of variation, is constant), which are simple manifolds: s1/m1= s2/m2 = s3/m3 = c

This discovery (by Spielman, Harris, Mellmann & Gershowitz, 1978) is one of the tracks leading to an underlying relationship between variance, speed, and frequency of biochemical switching, and on the more reductionist level of physics to a theory of capacity, energy and brain power (Weiss, 1992b).

Low IQ subjects need more units of time to solve a given elementary task, high IQ subjects less time. In such a way the intinuitive expectation, that higher values (more time) should correpond to higher variances, is again confirmed.

Supported by the relationship between EEG parameters and IQ (Weiss, 1992b), is also the conclusion: The molecular background of IQ differences must have something in common with individual differences in the biochemical regulation of brain energy metabolism.



If we would try to investigate the genetic background of a construct, called "physical ability for athletics" (for example, for decathlon) the appropriate approach would be clearly a polygenic one. Fundamental speed, long-distance endurance, body height and muscular power are relatively independent major factors of "physical achievement" among a larger number of other factors. However, if we would investigate only the background of the 100 m dash and eliminate influences of body height, the contractibility of muscle fibers underlying fundamental speed could turn out to have a relatively simple genetic background with a major locus. According to Plomin et al. (1994) the genetics of IQ is analogous to the case of "physical ability". But there are a number of reasons (Weiss, 1992a) that the genetics of IQ could be much simpler and a speed (and hence energy) factor with a major locus could play a decisive role. The existence of such a major locus does not mean that the overall number of factors, involving environmental as well as genetic sources of variance, is less than advocated by Plomin et al., but only that the role of one or two factors is greater.

In 1980 (p. 114) Jensen stated: "There are several critical threshholds within the total range of IQ, each having important educational and occupational consequences for the individual. ... The socially and personally most important threshhold regions on the IQ scale are those that differentiate with high probability between persons who because of their level of general mental ability ... can or cannot succeed in the academic or college preparatory curriculum through high school (about 105), and can or cannot graduate from an accredited four-year college with grades that would gratify for admission to a professional or graduate school (about IQ 115)."

In 1970 in former East Germany a family study, starting with 1329 mathematically gifted with an IQ higher than 130, came not only to very similar threshholds but also from a large body of empirical data (for details see Weiss, 1992) to the conclusion that the three overlapping types (M2M2, 68% of the total population, median IQ 94/ M1M2, 27%, median IQ 112 / M1M1, 5%, median IQ 130, mean IQ 139) are segregating in a Mendelian manner. Twenty three years later a follow-up (Weiss, 1994) of the gifted proved Mendelian segregation among their children (see Table 2) and among their nephews and nieces. Of course, in social reality, where cognitive ability is embedded into personality and chance plays its role in each biography, a 100% fit between a Mendelian theory of IQ and empirical results cannot be expected.

Table 2. Percentage Obtaining the Abitur (German High-School Leaving Examination) Among the Children of Highly Gifted M1M1

Children With Abitur

Children Without Abitur

Marriage Combination of Proband and Spouse

Percentage Obtained

Percentage Expected

Percentage Obtained

Percentage Expected

Both spouses with IQ of 124 and higher - M1M1xM1M1a





Gifted with spouse with IQ below 124 - M1M1xM1M2b





Note. From Weiss (1994). Percentages expexted under the assumption of Mendelian segregation and a cutoff IQ of 112.
an=242. bn=184.

Table 3. Percentage Obtaining the Abitur (German High-School Leaving Examination) Among the Nephews and Nieces of Highly Gifted M1M1

Children With Abitur

Children Without Abitur

Siblings of Probands and Respective Spouses

Percentage Obtained

Percentage Expected

Percentage Obtained

Percentage Expected

Both spouses with IQ of 124 and higher - M1M1xM1M1a





One spouse with IQ of 124 and higher, the other with IQ below 124 - M1M1xM1M2b





Both spouses with IQ between 104 and 124 - M1M2xM1M2c





One spouse with IQ of 124 and higher, the other with IQ below 124 - M1M2xM2M2d





Both spouses with IQ below 105 - M2M2xM2M2e





Note. From Weiss (1994). Percentages expexted under the assumption of Mendelian segregation and a cutoff IQ of 112.
an=70.   bn=130.   cn=107.   dn=29.  en=12.

Where polygenic theory predicts a normal distribution of ability among siblings of the same family (Jensen 1980, p. 80), major gene theory predicts that the IQ distribution of the offspring of homozygotes and their M1M2-spouses should be quite markedly skewed (compare Table 2, second row). The skewness should be in the opposite direction for M1M1-M1M2 offspring compared to M2M2-M1M2 offspring (compare Table 3, second and fourth row). M1M2-M1M2 offspring, who is segregating according to the Mendelian rules, should have a much greater variance than the offspring of marriages, where both partners are homozygotes and are "breeding true" (compare Table 3, first, third and fifth row), that means only with regression to the mean of the specific type. The major gene effects should even be more pronounced if we compare not-normalized raw scores instead of IQ values.

From the strong relationship between mean and variances of raw scores, an exact prediction of the variance for each of the cells of Table 2 and 3 is possible. Another prediction of major gene theory of intelligence that the means of raw scores of genotypes are simple manifolds, had already been confirmed by Frank (1985). Since 1959 Frank (see Lehrl, Gallwitz, Blaha & Fischer, 1991, for the last updating of Frank`s theory and for representative empirical results) is claiming that general cognitive ability is limited by the channel capacity of short-term memory (Kyllonen & Christal, 1990). Frank and and his school (Lehrl & Fischer, 1990; Weiss, 1992b) are arguing that the capacity C of short-term memory (measured in bits of information) is the product of the processing speed S of information flow (in bits/s) and the duration time D (in s) of information in short-term memory.


C (bits) = S (bits/s) x D (s)

According to Frank (1985) the mean of M1M1 is 140 bits, and of M2M2 70 bits, that means the contribution of a single M1 allele to short-term memory storage capacity C is about 70 bits, of a M2 allele about 35 bits. For a heterozygote M1M2 hence 70 bits + 35 bits = 105 bits. By Lehrl et al. testing of processing speed S was operationalized by reading rates, duration of information D by memory span. - see: The Basic Period of Individual Mental Speed (BIP)

In 1993 in the follow-up sample (Weiss, 1994) 97% (n = 357) of the highly gifted M1M1- males were in professions typically associated with an IQ above 123, compared with 55% (n =77) of the sons, 49% (n =220) of the brothers, 40% (n =346) of the fathers, 18% (n =570) of the male cousins, 22% (n =76) of the nephews, 14% (n=615) of the uncles (these data from 1971), 11% (n =2250) of the male cousins of the parents (with no evidence for any excess on the paternal or maternal side and hence no evidence for X-chromosome linked inheritance), 9% (n = 681) of the grandfathers, 5% (n =1996) of the uncles of parents and 4% (n =1290) of the greatgrandfathers. None of the major psychoses has such a high and regular "recurrence risk" among relatives as giftedness.

Drawing our results (Weiss, 1992a, 1994) into consideration, the research design by Plomin et al. (1994) was not optimal. The samples representing the top 5% (average IQ of 130) and top 1% (average IQ of 142) are quite right, as it is the idea of the comparison of extreme groups. However, the group of children with scores near the mean (average IQ of 105) is a mixture of M1M2 and M2M2 individuals, an average IQ of about 112 would have been more appropriate. Results from the bottom 5% and even 1% of the normal IQ distribution can be misleading, far better would be a representative sample from about the 40% lower tail of the IQ distribution (avoiding in such a way overlapping with M1M2 individuals). From the point of major gene theory the bottom 5% should even be avoided, because in this range of scores the effects of a large number of rare alleles can expected, which drastically disturb cognitive development, but do not play an important role beyond this range of scores.


Since in November 1970 I had calculated the first table, suggesting Mendelian segregation of IQ at a major locus (see Weiss, 1992a), it was my declaired aim to promote the discovery of the underlying polymorphism. All over the world IQ and intellectual achievement are more or less correlated with income and social status. Therefore, the wanted polymorphism should also be correlated with social status. However, the poor and the rich are not only differentiated by their average IQ, but also more or less dramatically by their living conditions and hence their conditions for Darwinian selection. The polygenic approach should run into this trap and produce a number of "polygenes of IQ" (the new HLA marker for a gene unique to the human brain could be the first one; see Plomin et. al. 1994), not connected with cognition at all, but reflecting different selection pressures (for example, for immunological factors) among the social strata. Association does not necessarily imply causation. Assortment of some genes (causing myopia, for example) could last over generations and accumulate different alleles in different social strata.

The first report of the IQ Quantiative Trait Loci (QTL) Project (Plomin et al., 1994) presents allelic association results for 60 markers. For two-allele markers allelic frequency differences between the low- and high-IQ groups of about .20 are significant. The only difference near significance, which from my point of view is exciting, is the small association found between IQ and mitochondrial superoxide dismutase 2. - From the point of major gene theory for the wanted locus an allelic frequency of 1.OO for the top IQ group would be predicted and of about .02 for the bottom group. For a discovery with such clearcut contours we should strive for.

In 1982 I became aware (see Weiss, 1984) of a paper published by Sinet, Lejeune & Jerome (1979) in which a correlation of .58 between IQ and erythrocyte glutathione peroxidase activity (GSHPx) was reported for 50 trisomy 21 patients. None of the other enzymes studied correlated with IQ. Sinet et al. thought the correlation to be trisomy-specific, because an increase of about 50% in the superoxide dismutase activity (SOD-1) can be observed in cells from trisomy 21 patients. There is a feedback control of GSHPx concentration by the amount of superoxide, which explains (Chan, Yu, Chen & Epstein 1989) the elevated activity of GSHPx in cells of trisomy 21 patients. However, Fraser and Sadovnick (1976) had found that the correlations of IQ between trisomy 21 probands with their fathers, mothers and sibs are about .50, consequently of the same size as with healthy children despite the mean IQ of trisomy 21 probands being about 70 points lower. Therefore already Lenz (1978) had concluded that individual differences in trisomy-IQ have generally the same biochemical background as in normal persons. And Brugge, Nichols, Delis, Saitoh & Truaner (1992) confirmed a correlation of .73 between erythrocyte GSHPx activity and a short-term memory score.

Because of the political situtation in former East Germany, since 1982 the author of this editorial had lost all possibilities (see Weiss, 1991) to promote investigation into these correlations with a laboratory background of his own, and he could accumulate only indirect evidence (Weiss, 1984, 1992a).

We all know that an occupational group with higher education, whether black, white or yellow, has a mean IQ of at least 30 points higher than the social stratum of unskilled workers. Up to World War One for historical reasons Germans in the Baltic States were nearly all members of the former ruling upper stratum; and now the about 50 000 Baltic Germans, resettled in Germany, are a cultural community far above average. In 1913, a representative comparison of IQ sampled in Estonia, would have had the false result that Germans are about one standard deviation of IQ higher than Estonians. Therefore, I conclude that nations or races could only be compared if all social positions are filled by one race (as in Nigeria and Sweden). In all other cases, IQ-differences are mixed up with the consequences of social inequality or are themselves the cause of inequality. With this in mind, we look at the following data: Golan, Ben Ezzer & Szeinberg (1980) published findings on a quantitative genetic polymorphism of erythrocyte GSHPx actitivity. The population mean of Jews in Israel was 23,7 + 7,0 U GSHpx/Hb. By Gerli, Mongiat, Gualandri, Orsini & Porta (1984) GSHPx was assayed in families of Mediterranean origin (population mean of about 20 + 5 U) and the results support the existence of two Mendelian alleles. Mean GSHPx activity among Australian aborigines is 12,6 +3 U (Agar, Gupta, Gruca & Welch, 1980). 100 healthy university student had a mean of 40,5 + 11,9 U (Lane, Dudrick & Warren, 1981). However, there are differences between the methods used, and the results can only be only tentative.

Also glutathione S-transferases (GST) possess GSH peroxidase properties (see, for example, Saneto, Awasthi & Srivastava, 1982). GST is a whole family of genetic polymorphisms, and at least one new polymorphism is discovered each year. However, until now we do not know anything about the contribution of any GST-polymorphism to the correlation between GSHPx and IQ. The molecular genetics of GSHPx and GST polymorphisms is in rapid development, and many locations on chromosomes are well known. A research group, who wants to contribute to this area, will have to refer to the most actual electronic data base.

At the beginning (Weiss, 1984) GSHPx seemed only to be involved with lipid peroxidation. However, the disvovery of the regulation of the NMDA receptor-channel complex (Sucher & Lipton, 1991) by oxidized glutathione has opened a new research front. The involvement of the NMDA complex in short-term memory and long-term potentiation is one of the hottest fields of neurochemical research. In view of this direct and indirect evidence, it is to complain that Plomin et al. did not investigate GSHPx or the family of GST polymorphisms until now.

Among the polygenes of IQ, one will turn out to have the largest effect. By definition it will be the major gene locus. Whether approach will be taken, either a search for polygenes or an approach, led by major gene theory, the result of replicable science must finally be the same. One of the minor genes of IQ (at least for the eldery), already discovered, is apolipoprotein E (Goedert, Strittmater & Roses, 1994). Considering the public opinion on this subject (Weiss, 1991) in general, nobody, however, would have got the money for a major gene approach as early as 1990. From this point of view the IQ QTL project (Plomin et al., 1994) has always to be understood as a breaktrough.