On 2015 Dec 09, Donald Forsdyke commented:

PURINE LOADING AS A THERMAL ADAPTATION The proteins of thermophiles are generally more heat-stable than the corresponding proteins from mesophiles. This must be reflected in either, or both, of two major amino acid variables – composition and order. In the past the notion that amino acid composition might be reflective of the pressure in thermophiles to retain purine-rich codons (3) has been disparaged by Zeldovich et al. (6). In this elegant new paper (5), Venev and Zeldovich (2015) agree that the “multiple factors” not accounted for in their modelling “include the influence of the genetic code and guanine-cytosine (GC) content of the genomes on amino acid frequencies.” However, there is puzzlement that the “theory and simulations predict a strong increase of leucine content in the thermostable proteins, whereas it is only minimally increased in experimental data.” Perhaps it is of relevance that leucine is on the top left quadrant of the standard presentation of the genic code, its codons being extremely poor in purines.

My response to Zeldovich et al. (6) in 2007, and my follow-up references in 2012 (1, 4), are set out below. One of the coauthors of the 2007 paper has recently further contributed to this topic (2).

2007 Response

This paper draws conclusions tending to oppose those of myself and coworkers (cited). A "key question" is held to be: "Which factor - amino acid or nucleotide composition - is primary in thermal adaptation and which is derivative?" Previous evidence is considered "anecdotal." Now there is evidence for "an exact and conclusive" relationship, based on an "exhaustive study" that provides a "complete picture." A set of amino acids - IVYWREL - correlates well with growth temperature. It is noted:

"Signatures of thermal adaptation in protein sequences can be due to the specific biases in nucleotide sequences and vice versa. ... One has to explore whether a specific composition of nucleotide (amino acid) sequences shapes the content of amino acid (nucleotide) ones, or thermal adaptation of proteins and DNA (at the level of sequence compositions) are independent processes."

In other words, are primary adaptations at the nucleic acid level driving changes at the protein level, or vice- versa? To what extent are the two processes independent? Their conclusion:

"Resolving the old-standing controversy, we determined that the variation in nucleotide composition (increase of purine-load, or A + G content with temperature) is largely a consequence of thermal adaptation of proteins."

Thus, the superficial reader of the paper, while noting the purine-richness of some of the codons corresponding to the IVYWREL amino acids, will conclude that the "independent processes" alternative has been excluded. Reading the paper (e.g. Figure 7) one can question the validity of this conclusion. Many of the IVYWREL amino acids have purine-poor alternative codons (especially IYLV, which at best can only change one purine unit in their codons). One of the IVYWREL amino acids has relatively purine-rich alternative codons (R, which at best can change two purine units). Two (EW) are always purine-rich, and there are no alternatives.

Displaying more EW's as the temperature got hotter would satisfy a need both for more purines and for more tryptophan and glutamate, so here there is no discrimination as to whether one "shapes" the organism’s content of the other. Displaying more IYLVs gives only minimal flexibility in accommodating a purine-need. Most flexibility is provided by R codons.

The authors do not give statistics for the differences between the slopes of Figs. 7a (unshuffled codons) and 7b (shuffled codons), but they appear real, presumably reflecting the choice biologically of purine-rich codons, a choice the organisms might not have to make if there were no independent purine-loading pressure. Thus, the authors note, but only in parenthesis, that the slopes "are somewhat different suggesting that codon bias may be partly responsible for the overall purine composition of DNA."

2012 Response

As a follow up, it can be noted that Dehouck et al. (2008) report that relationship between a protein's thermostability and the optimum growth temperature of the organism containing it, is not so close as previously thought (1). Furthermore, Liu et al. (2012) now conclude from a study of xylanase purine-rich coding sequences that "The codons relating to enzyme thermal property are selected by thermophilic force at [the] nucleotide level," not at the protein level (4).

1.Dehouck Y, Folch B, Rooman M (2008) Revisiting the correlation between proteins' thermoresistance and organisms' thermophilicity. Protein Engineering, Design and Selection 21:275-278.Dehouck Y, 2008

2.Goncearenco A, Berezofsky IN (2014) The fundamental tradeoff in genomes and proteomes of prokaryotes established by the genetic code, codon entropy, and the physics of nucleic acids and proteins. Biology Direct 9:29 Goncearenco A, 2014

3.Lambros RJ, Mortimer JR, Forsdyke DR (2003) Optimum growth temperature and the base composition of open reading frames in prokaryotes. Extremophiles 7:443–450.Lambros RJ, 2003

4.Liu L, Wang L, Zhang Z, Wang S, Chen H (2012) Effect of codon message on xylanase thermal activity. J. Biol. Chem. 287:27183-27188 Liu L, 2012

5.Venev SV, Zeldovich KB (2015) Massive parallel sampling of lattice proteins reveals foundations of thermal adaptation. J. Chem. Phys. 143: 055101Venev SV, 2015

6.Zeldovich KB, Berezofsky IN, Shakhnovich EI (2007) Protein and DNA sequence determinants of thermophilic adaptation. PLOS Comput. Biol. 3(1), e5.Zeldovich KB, 2007

On 2015 Dec 09, Donald Forsdyke commented:

PURINE LOADING AS A THERMAL ADAPTATION The proteins of thermophiles are generally more heat-stable than the corresponding proteins from mesophiles. This must be reflected in either, or both, of two major amino acid variables – composition and order. In the past the notion that amino acid composition might be reflective of the pressure in thermophiles to retain purine-rich codons (3) has been disparaged by Zeldovich et al. (6). In this elegant new paper (5), Venev and Zeldovich (2015) agree that the “multiple factors” not accounted for in their modelling “include the influence of the genetic code and guanine-cytosine (GC) content of the genomes on amino acid frequencies.” However, there is puzzlement that the “theory and simulations predict a strong increase of leucine content in the thermostable proteins, whereas it is only minimally increased in experimental data.” Perhaps it is of relevance that leucine is on the top left quadrant of the standard presentation of the genic code, its codons being extremely poor in purines.

My response to Zeldovich et al. (6) in 2007, and my follow-up references in 2012 (1, 4), are set out below. One of the coauthors of the 2007 paper has recently further contributed to this topic (2).

2007 Response

This paper draws conclusions tending to oppose those of myself and coworkers (cited). A "key question" is held to be: "Which factor - amino acid or nucleotide composition - is primary in thermal adaptation and which is derivative?" Previous evidence is considered "anecdotal." Now there is evidence for "an exact and conclusive" relationship, based on an "exhaustive study" that provides a "complete picture." A set of amino acids - IVYWREL - correlates well with growth temperature. It is noted:

"Signatures of thermal adaptation in protein sequences can be due to the specific biases in nucleotide sequences and vice versa. ... One has to explore whether a specific composition of nucleotide (amino acid) sequences shapes the content of amino acid (nucleotide) ones, or thermal adaptation of proteins and DNA (at the level of sequence compositions) are independent processes."

In other words, are primary adaptations at the nucleic acid level driving changes at the protein level, or vice- versa? To what extent are the two processes independent? Their conclusion:

"Resolving the old-standing controversy, we determined that the variation in nucleotide composition (increase of purine-load, or A + G content with temperature) is largely a consequence of thermal adaptation of proteins."

Thus, the superficial reader of the paper, while noting the purine-richness of some of the codons corresponding to the IVYWREL amino acids, will conclude that the "independent processes" alternative has been excluded. Reading the paper (e.g. Figure 7) one can question the validity of this conclusion. Many of the IVYWREL amino acids have purine-poor alternative codons (especially IYLV, which at best can only change one purine unit in their codons). One of the IVYWREL amino acids has relatively purine-rich alternative codons (R, which at best can change two purine units). Two (EW) are always purine-rich, and there are no alternatives.

Displaying more EW's as the temperature got hotter would satisfy a need both for more purines and for more tryptophan and glutamate, so here there is no discrimination as to whether one "shapes" the organism’s content of the other. Displaying more IYLVs gives only minimal flexibility in accommodating a purine-need. Most flexibility is provided by R codons.

The authors do not give statistics for the differences between the slopes of Figs. 7a (unshuffled codons) and 7b (shuffled codons), but they appear real, presumably reflecting the choice biologically of purine-rich codons, a choice the organisms might not have to make if there were no independent purine-loading pressure. Thus, the authors note, but only in parenthesis, that the slopes "are somewhat different suggesting that codon bias may be partly responsible for the overall purine composition of DNA."

2012 Response

As a follow up, it can be noted that Dehouck et al. (2008) report that relationship between a protein's thermostability and the optimum growth temperature of the organism containing it, is not so close as previously thought (1). Furthermore, Liu et al. (2012) now conclude from a study of xylanase purine-rich coding sequences that "The codons relating to enzyme thermal property are selected by thermophilic force at [the] nucleotide level," not at the protein level (4).

1.Dehouck Y, Folch B, Rooman M (2008) Revisiting the correlation between proteins' thermoresistance and organisms' thermophilicity. Protein Engineering, Design and Selection 21:275-278.Dehouck Y, 2008

2.Goncearenco A, Berezofsky IN (2014) The fundamental tradeoff in genomes and proteomes of prokaryotes established by the genetic code, codon entropy, and the physics of nucleic acids and proteins. Biology Direct 9:29 Goncearenco A, 2014

3.Lambros RJ, Mortimer JR, Forsdyke DR (2003) Optimum growth temperature and the base composition of open reading frames in prokaryotes. Extremophiles 7:443–450.Lambros RJ, 2003

4.Liu L, Wang L, Zhang Z, Wang S, Chen H (2012) Effect of codon message on xylanase thermal activity. J. Biol. Chem. 287:27183-27188 Liu L, 2012

5.Venev SV, Zeldovich KB (2015) Massive parallel sampling of lattice proteins reveals foundations of thermal adaptation. J. Chem. Phys. 143: 055101Venev SV, 2015

6.Zeldovich KB, Berezofsky IN, Shakhnovich EI (2007) Protein and DNA sequence determinants of thermophilic adaptation. PLOS Comput. Biol. 3(1), e5.Zeldovich KB, 2007