Detection of Angiogenic Doping

Detecting Angiogenic Gene Doping

Shortly after it was banned in sports, WADA began funding research into detecting gene doping [299]. Yet, two decades later, there is no standardized detection method [300]. The tests that could be used to screen for gene doping include detection of the doping agent itself, specific biomarkers in blood samples, and atypical DNA signatures by real-time PCR and next-generation sequencing (NGS) [225, 301]. While methods have been proposed to detect atypical DNA signatures [205, 301–305], multiple methods exist to bypass them. For example, detecting exon–exon junctions, which are absent in genomic DNA, can be bypassed using intron grafting (see Fig. 5).

Fig. 5: Detection of angiogenic gene doping. A Even the most advanced and sensitive detection methods rely on some form of discriminatory PCR, e.g., using exon junction-spanning primers to detect cDNA or next-generation sequencing to detect missing exon–intron junctions or novel junctions [205, 302]. However, large-capacity vectors can accommodate full genes, while low- and medium-capacity vectors can use intron grafting to evade detection. B Unlike the unique EPO gene, there are five VEGF genes with significant homology. Due to the degeneracy of the genetic code, silent mutagenesis can create thousands of different nucleotide sequences for VEGF genes while maintaining their amino acid sequence. In the example shown, the VEGF-A165 cDNA sequence was maximally modified by silent mutagenesis to become almost as different from its wild-type sequence as it is from the paralogous VEGF-B167 cDNA sequence. Due to their close evolutionary relationship [306], this is possible for all VEGFs. DNA swarms, all coding for the same protein, can be used to increase the detection threshold. The numbering of the nucleotides is according to NM_001171626 (VEGF-A) and NM_001243733.2 (VEGF-B). Exons are not drawn to scale. The % identities refer to the cDNAs of the entire mature proteins, not only to the exon 3 sequence shown. A exon/intron VEGF-A exon/intron, B exon/intron VEGF-B exon/intron, bp base pairs, cDNA complementary DNA, NGS next-generation sequencing, pA polyadenylation sequence, PCR polymerase chain reaction, wt wild type

Notably, the human genomic VEGF-B sequence—including six introns—measures only 3.5 kb, from start-ATG to stop-TGA [307], and can fit into cargo-limited viral vectors. Cargo up to ~ 38 kb has been inserted into baculoviruses [308], long enough to accommodate most human genes, which have a mean length of ~ 28 kb [309] and a median of ~ 24 kb [310]. An upper limit on the insert length for baculovirus genomes has not been established, as the capsid size appears to adapt to genome size, potentially allowing up to 100-kb inserts [311], thereby eliminating the need for intron grafting for all but the largest genes. Silent mutagenesis is another effective countermeasure to PCR-based detection methods. However, it remains to be shown how maximized codon optimization would perform against hybridization-based methods such as those proposed by de Boer et al. [301].

Detecting Angiogenic Protein Doping

Crucially, in gene doping, the active pharmaceutical protein is produced by the athlete’s cells and is therefore potentially indistinguishable from the ‘normal’ protein if expressed orthotopically. Not only is the molecular identity of angiogenic proteins in question, their levels also might not reliably distinguish between endogenous and doping-induced angiogenic factors, because the levels of angiogenic proteins in the blood plasma of normal healthy human individuals vary substantially. For example, VEGF-A levels in blood plasma can range from undetectable to nearly 500 pg/mL [312, 313], making the use of cutoff values challenging. On the upside, despite significant interindividual variability, plasma and serum VEGF-A levels in any given individual seem to be relatively stable, at least up to 6 months, and show no circadian variation [314, 315].

The detection of the angiogenic protein itself presumes that the protein (a) is a secreted protein and (b) enters the circulation or urine. Neither can be assumed. The transcription factor HIF-1α (see Fig. 2) is highly angiogenic but not secreted from cells [242]. Furthermore, VEGF-A is a tissue hormone, that is, unlike erythropoietin, not systemically distributed by the blood. Instead, VEGF-A is secreted into the extracellular space and encounters the VEGF receptors on endothelial cells via the interstitium from the abluminal side [187, 316]. While the soluble VEGF-A121 isoform might leak into the circulation to some extent, the longer heparin-binding isoforms of VEGF-A are much less likely to show up in the blood as they are bound tightly to the extracellular matrix and cell surfaces (see Fig. 6).

Fig. 6: Current model of VEGF-A action via hypoxia-generated growth factor gradients and negative feedback. A Cells that experience hypoxia (due to tissue growth, decreased oxygen delivery, or increased consumption) upregulate hypoxia-regulated genes in a cell-type–specific manner; for example, muscle cells begin to produce VEGF-A [171], and renal cells erythropoietin [317]. Many cell types can respond to hypoxia by increasing VEGF-A expression, which attempts to restore normoxia via increased vascularization. B It is generally thought that by their differential affinity to extracellular matrix and cell surface heparan sulfate proteoglycans, the various VEGF-A isoforms lay down a growth factor gradient, which is sensed by specialized endothelial cells at the growth front (tip cells), leading the angiogenic sprout to form a hierarchical vascular network to supply the hypoxic tissue with oxygen. Interestingly, there is surprisingly little direct evidence for VEGF-A gradients in vivo [318, 319], suggesting that other mechanisms may be involved

Indirect Detection Methods

Indirect methods attempt to detect the means or consequences of doping. For example, tests could screen for the gene-doping delivery vector or recent blood vessel growth. Viruses used to deliver the genetic cargo would result in a measurable immune response [225]. Also, other delivery systems, such as lipoprotein nanoparticles or engineered extracellular vesicles, would leave traces for at least a short time in the athlete’s body. Because the ABP has proven to be a valuable tool [320, 321], its use could be expanded to encompass additional indirect biomarkers that may indicate angiogenic doping. It is unclear what biomarkers these would be, especially since tissue biopsies are currently the only reliable method for quantifying parameters indicative of pro-angiogenic interventions at the capillary level, such as vascular and branching density, vessel diameter and length, or vascular volume. Currently, there is no non-invasive imaging technology that can quantitatively image the microvasculature at the required resolution (< 1 µm) and depth for large leg muscles [322]. However, microRNA in the blood might serve as a proxy, at least for HIF-upregulating agents [255, 323], as hypoxia-mediated regulation of angiogenesis involves several specific microRNAs [324].

Unlike in the pre-ABP era, with fixed cut-off values (e.g., hemoglobin 150 and 175 g/L for women and men, respectively), the ABP uses, among many other parameters, previous test results to predict individualized upper and lower limits for blood parameters [325]. While retrospective studies suggest that athletes have ‘taken the fast lane’ in the past [320], slow changes that do not result in an abnormal blood profile score may represent a potential ‘Achilles heel’ of the ABP. Doping, increased slowly over several years, could produce physiological changes that fall within normal biological variation or phenocopy improvements gained through a sustained training effort [325].

Gene Doping Versus Gene Editing

In the long run, the single most critical technology to monitor is CRISPR, as it can perform precise genetic modifications to existing genetic material in a live organism. While FDA-approved treatments are currently limited to a single ex vivo drug [226], there is little doubt that CRISPR and similar gene-editing technologies will be the future of biologicals [227]. If the delivered gene-editing agents are transient enough and the editing is limited to single base-pair mutations, the results could only be distinguished from hereditary mutations by genome sequencing. Even with great technological advances, it is unlikely that we will achieve editing efficiency in vivo of close to 100% in the near future. But since genetic mosaics do also occur naturally, the mere existence of a mixed genotype would not constitute proof of editing. A detailed genomic analysis, including single-cell sequencing, might be required. However, single-cell sequencing is not only expensive but also requires an invasive biopsy. Since CRISPR technology can also arbitrarily change gene expression without modifying DNA by targeting methylating enzymes or transcription factors, detecting the CRISPR delivery system—despite its transient nature—may be easier than detecting the CRISPR effect.

Science has already identified a few naturally occurring mutations that can increase athletic performance [326]. Feasibility studies of such modifications have been performed in mice [327]. Once an athlete is found to carry a known performance-enhancing mutation, what would be the consequences? If it was acceptable when the mutation originated from ‘natural’ inheritance or from a spontaneous mutation during embryonic development, why would an ‘artificially’ acquired mutation be treated differently? Assuming sufficiently advanced technology, there appears to be no way to determine whether a person is mutated by nature or genetic engineering. Even genotyping all alive and deceased athlete's relatives would leave doubt as humans naturally acquire surprisingly many mutations during embryonic development [328].

Plasma VEGF-A: An Unreliable Epiphenomenon

There are several methods for measuring VEGF levels in biological samples, such as bioassays and enzyme-linked immunosorbent assays (ELISAs). Bioassays use growth factor–sensitive cell lines, whereas ELISAs are based on VEGF-A-specific antibodies. ELISA tests have been the method of choice in most preclinical studies [329–331]. The best current ELISAs detect VEGF-A levels down to the single-digit pg/mL range [312], and typical VEGF-A concentrations in blood plasma are in the double-digit pg/mL range, below what is considered necessary for a biological effect.

Currently, none of these tests has FDA clearance, and thus they are not widely available. However, unlike erythropoietin, whose producer cells are in close contact with permeable capillary networks [332], hypoxia-induced VEGF-A is, by definition, produced far distant from capillaries, and there is no reason to believe that it serves—the platelet VEGF-A pool excluded [333]—any purpose in the systemic blood; its presence in plasma is likely an epiphenomenon.

Since VEGF-A can be produced by most cell types in the human body that are exposed to hypoxia, the glycosylation pattern of endogenous human VEGF can be heterogeneous [334]. Detecting recombinant CHO- or HEK-293–produced VEGFs might therefore be less straightforward than detecting recombinant erythropoietin, whose production is almost exclusively limited to a single cell type in the kidneys [335], and any size aberration of erythropoietin is likely to represent exogenous protein. Unlike erythropoietin, VEGF-A also exists in a variety of different isoforms. Only the smallest isoforms (VEGF-A110, VEGF-A121, VEGF-A145, VEGF-A165) are thought to be soluble and to leak into the vasculature, because the longer VEGF-A isoforms are strongly ‘heparin-binding’, meaning they bind to cell surfaces and extracellular matrix due to their interaction with heparan sulfate proteoglycans (HSPGs) [55]. However, these longer VEGF-A forms are essential for angiogenesis [336, 337]. The major isoform in humans (VEGF-A165) is partly soluble and partly HSPG-bound, but in some mammals, isoforms corresponding to the human VEGF-A165 do not exist, and the major isoform (VEGF-A188/189) is strongly HSPG-binding [306], indicating that VEGF-A189 might also work well in humans (see Fig. 6).

Once produced by hypoxic cells, VEGF-A distributes in the interstitial space, and its interactions with VEGF receptor-1 or -2 are thought to happen on the abluminal side of the endothelial cell [338, 339]. Thus, the VEGF-A measured in blood plasma represents leakage, transendothelial transport, or is introduced into the blood via lymphatic drainage. This view is supported by the typically low VEGF-A plasma concentrations, which are often a fraction of what is needed to elicit an angiogenic response. In line with this view is the fact that the correlation of plasma or serum VEGF-A with angiogenesis is not very strong. While an increase in interstitial VEGF-A will also increase plasma VEGF-A, individual differences in permeability (how much VEGF-A leaks into the blood) and lymphatic function (how much VEGF-A is carried into the blood via interstitial fluid drainage) easily explain the variation between individuals. In cancer patients, plasma VEGF-A levels are significantly higher, most likely due to VEGF-A entering the blood via the tumor’s leaky vessels [340]. High VEGF-A plasma levels have consistently been associated with disease severity (as a prognostic marker), but their use as a biomarker to predict the response to antiangiogenic cancer treatment remains elusive [341, 342], and may even depend on the sampling method [330].

The distribution of receptors on the luminal and abluminal sides of the cell has been modelled. While the model was relatively simple (i.e., it included only two VEGF-A isoforms, two receptors, and one co-receptor), an increase in VEGF-A production in this model, first and foremost, increased the internalization of VEGFR-1, which primarily functions as a non-angiogenic decoy receptor [316]. Not much wetlab research has compared luminal and abluminal signaling; however, based on the available literature, it is very likely that the spatial distribution of signaling is a crucial factor in the response to VEGF-A [187, 188, 343].