Genomics and proteomics: of hares, tortoises and the complexity of tortoises

Keywords::

• albumin depletion
• antibodies
• glycosylation
• human proteome
• low abundance proteins
• membrane proteins
• phosphoproteins
• post-translational modifications
• protein biomarkers
• proteomics sample preparation

The draft human genome sequence was published in 2001 and took 13 years to complete. Necessity being the mother of invention, the accelerated development of DNA sequencing technologies led to the completion of the Human Genome Project nearly a year ahead of schedule [1]. In 1998, the total sequencing output from the entire Human Genome Project was about 200 million base pairs for the year. By January 2003, the Department of Energy Joint Genome Institute had reached the capacity to sequence 1.5 billion bases per month.

Today, the Bejing Genome Institute in China has the capacity to generate the equivalent of 10,000 human genomes per year. In 2010, the Bejing Genome Institute generated ten times the amount of data the National Center for Biotechnology Information had generated in the past 20 years. In early 2012, Oxford Nanopore Technologies announced that its second-generation GridIons would be able to sequence a human genome in 15 min [2]. Enter proteomics, the proverbial tortoise in the race with the hare.

The number of genes & the number of proteins

In 1953, Wilkins et al. [3], Franklin and Gosling [4] and Watson and Crick [5] simultaneously published the structure of DNA in the same issue of Nature. (Earlier that year, Pauling and Corey [6] had published an incorrect triple helix model of DNA). By 1964, Vogel [7] very accurately calculated the size of the human genome, but inaccurately predicted it might consist of over 6 million genes. By recent estimates, the human genome has only 22,000–23,000 genes, slightly more than the worm Caenorhabditis elegans, which has only 959 cells but 20,000 genes [8], and fewer than a grape plant, which has 30,434 genes [9].

There is a surprising lack of correspondence between genome size and complexity of the organism [10,11]. If the DNA from a single human cell were uncoiled and laid end to end, it would measure over two meters in length. By comparison, the genome of the flowering plant Paris japonica is 50 times larger than the human genome [12], and outstretched, the DNA from a single cell would span the length of a soccer field.

While transcriptomics attempts to monitor gene expression by quantifying mRNA transcripts, the poor correlation between mRNA levels and protein expression is well known [13]. Hence, neither genomics or transcriptomics can accurately predict the constituency of the proteomic complement, since an undeterminable number of possible post-translational modifications (PTMs) can potentially produce multiple isoforms of each protein. For instance, there are at least 3778 distinct genes encoding for plasma proteins, of which at least 51% of these genes encode for more than one protein isoform [14]. Therefore, the total number of proteins in the human proteome is expected to reach into the millions. Consider the immunoglobulins for which there are 70 genes encoding 320 possible light chain combinations and 10,530 possible heavy chain combinations, which collectively could contribute 3,369,600 possible antibody types [15].

The diversity of proteins

To emphasize the diversity of proteins, consider a single stranded DNA hexamer in which there are 4096 possible combinations of A, G, T and C. By comparison, there are 20 proteogenic amino acids, as such the number of possible sequences for a hexapeptide is 20⁶ or 64,000,000. Now consider only one of the most common PTMs, the glycosylation of Asn, Arg, Ser, Tyr and Thr residues, and the number of possible sequences for the same theoretical hexapeptide increases to 25⁶ or 244,140,625. To further add to this complexity, a calculation of all possible oligosaccharide isomers, both branched and linear, predicts 10¹² structures for a reducing hexasaccharide [16].

The median length of eukaryotic proteins is 360 amino acids [17,18], therefore the prediction of possible sequences for an ‘average’ protein, not including PTMs, reaches an astronomical 20³⁶⁰. Incidentally, titin is the largest known protein, 34,350 amino acids in length. Written out, its full chemical name would be the longest word in the English language consisting of 189,819 letters.

As of 2011, more than 87,000 post translational modifications had been identified experimentally from over 530,000 protein entries in the Swiss-Prot protein database, while another 234,938 are postulated [19]. While over 200 different protein PTMs have been identified [20], there remains much disparity over the frequency at which PTMs occur. For example, Apweiler et al. [21] predicted as many as 50% of all human proteins might be glycosylated, whereas more recent estimates by Khoury et al. [19] suggest the actual number might be fewer than 20%.

Carpe articulum (Seize the important moment)

It is estimated that 30% of all cellular proteins are phosphorylated on at least one residue [22], the relative frequency of serine, threonine and tyrosine phosphoylation sites being 54%, 31%, 14%, respectively. PhosphoNET presently holds over 657,000 known and putative phosphorylation sites in over 23,000 human proteins, only 15% of which have been experimentally validated [23].

There are 518 known kinases [24] and possibly over 1000 phosphatases [25,26] that regulate their activity. The addition or elimination of a phosphate at one or multiple sites on a protein alters its net charge, conformation, and regulates functionality by rapidly oscillating between functional and nonfunctional states. Isoelectric focusing is particularly sensitive for simultaneously detecting changes in the phosphorylation state of multiple proteins [27].

Protein phosphorylation state, at any given moment, is the net result of the opposing activities of kinases and phosphatases and can change very rapidly. As examples, one molecule of tissue nonspecific alkaline phosphatase can convert 971 molecules of substrate per second [28], and a turnover rate of over 2000 molecules per second has been reported for human intestinal phosphatase [29]. Hence, it is extremely difficult to obtain a phosphoprotein profile that accurately reflects the in vivo state. Juhl showed that protein phosphorylation states were altered in surgical specimens, even prior to resection, where the ligation of arteries induces ischemic response [30].

Membrane proteins

It is estimated that membrane proteins constitute, roughly 20% of the human proteome, and more recently, Almen et al. [31] predicted this number might be as high as 27%. As of early 2012, the Protein Data Bank of Transmembrane Proteins held over 83,700 structures of which 1690 transmembrane proteins have been predicted so far [32,33]. Membrane proteins are of particular importance, since over 60% of current drugs are based on targeting this group of proteins [34].

The dynamic range of protein concentrations

The complexity of proteomes is exacerbated by the broad range of protein concentrations, spanning at least 12 orders of magnitude as exemplified by human plasma where the mass of albumin is nearly ten billion times greater than that of clinically important proteins like the interleukins [35,36]. Moreover, it is most likely that trace proteins which are not supposed to be in plasma, those which have leaked from damaged tissues, will be of the most diagnostic and prognostic importance.

Every tissue in the body is supplied by over 60,000 miles of veins, arteries and capillaries that comprise the human circulatory system, so conversely, tissues will have a background of plasma proteins. Perfusion of tissues to remove blood components is not a viable option, since the in vitro state of proteins is unlikely to be preserved under these conditions. For example, Quissell et al. demonstrated a half-life of less than 6 min for the phoshorylated state of an integral membrane protein regulating exocytosis [37].

Strategies for the depletion of high abundance structural and housekeeping proteins and the enrichment of extremely low abundance proteins will be necessary to effectively mine the proteome. Like peeling away the layers of an onion, the systematic depletion of the most abundant proteins reveals the next innermost layer of proteome, but as with chefs, this has not happened without sometimes bringing tears to the eyes of some scientists.

The most development in this area has been for plasma and serum, from which it is hoped a ‘biopsy’ of the human condition might be derived. In human plasma, only six proteins constitute more than 99% of the total protein mass. Immunoaffinity chromatography has been used to deplete the 20 most abundant proteins from plasma, however, this risks the depletion of other proteins, peptides and potential biomarkers that may be associated with the molecules targeted for depletion. The study of proteins and peptides bound to albumin has been termed ‘albuminomics’ and Gundry et al. [38] showed at least 35 other proteins of both high and low abundance that were retained in the albumin fraction. In terms of mining for biomarkers, these approaches seem to come with some risk of ‘throwing the baby away with the wash water’.

Conversely, immunoaffinity can be used in enrichment schemes designed to isolate and quantify very low abundance proteins, as exemplified by sandwich ELISAs, which utilize capture antibodies to selectively immobilize specific antigens. However, it is interesting to note that mouse monoclonal antibodies will be ineffective in many cases, since heterophilic antibodies such as mouse antihuman antibodies are expressed in 10–40% of the human population [39,40].

Alternatively, the synthetic peptide ligand libraries first described by Lam et al. [41] and later extensively developed by Righetti and Boshetti [42,43] have been shown to be highly effective for isolating extremely low abundance proteins and potential biomarkers from complex samples such serum [44], urine [45], saliva [46] and cerebrospinal fluid [47]. Immobilized on a single column, a hexapeptide library constructed from 17 amino acids would represent over 24 million possible ligands [48].

Challenges facing the proteomics generation

Nature’s antibodypedia antibody database lists over 186,000 reviewed antibodies, covering approximately 84% of all human genes [49]. However, the ability to quantify potentially valuable protein biomarkers is frequently hindered by the inability to reliably isolate these proteins. The Human Proteome Organization’s Human Antibody Initiative is aimed toward compiling comprehensive tissue profiles for normal and diseased tissues [50]. This will be particularly challenging, since the most stringent conditions are often required to isolate the total protein constituency from some tissues. A random survey of 327 validated antibodies listed on antibodypedia showed that 17% of ELISA compatible antibodies were not compatible with western blotting, indicating that conformational epitopes needed to be preserved for antibody recognition.

While sequencing the human genome in 13 years has been heralded as one of humankind’s greatest scientific achievements, compared with walking on the moon and the splitting of the atom, decrypting the enormous complexity of the human proteome and its ‘interactome’ will be no less daunting a task, requiring many more decades to complete.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.