JPROT-01738; No of Pages 7 JOURNAL OF P ROTEOM IC S XX ( 2014) X XX–X XX
Available online at www.sciencedirect.com
ScienceDirect www.elsevier.com/locate/jprot
Review
2
Six decades searching for meaning in the proteome☆
3Q1
Leigh Anderson⁎
4
SISCAPA Assay Technologies, Inc., P.O. Box 53309, Washington, DC 20009, USA
R O
O
F
1
5 87
AR TIC LE I NFO
ABSTR ACT
91Q6 16 0
Keywords:
This review describes one thread in a fabric of developments leading to the present state
11 17
Proteomics
of proteomics, stretching over 60 years and ending with a prediction for 2024. While
12 18
HPI
composed largely of personal reminiscences, the story offers some instructive success and
13 19
SISCAPA
failures, and appears to be nearing the long-sought goal of deep insights into real biology.
14 20
MRM
This article is part of a Special Issue entitled: 20 years of Proteomics.
15 21
MALDI
22
Biomarker
23
Plasma
35 36 37 38 39 40 41 42 43 44 45
P
D
E T E
R
34
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1964: molecular anatomy . . . . . . . . . . . . . . . . . . 1974: 2-D gels . . . . . . . . . . . . . . . . . . . . . . . . 1984: large scale biology . . . . . . . . . . . . . . . . . . 1994: big 2-D . . . . . . . . . . . . . . . . . . . . . . . . . 2004: mass spec . . . . . . . . . . . . . . . . . . . . . . . 2014: a biomarker era . . . . . . . . . . . . . . . . . . . . 2024: personalized, patient-centric medicine . . . . . . . 8.1. Technology . . . . . . . . . . . . . . . . . . . . . . 8.2. Menu . . . . . . . . . . . . . . . . . . . . . . . . . 8.3. Dried blood spots . . . . . . . . . . . . . . . . . . 8.4. Personal baselines . . . . . . . . . . . . . . . . . . 8.5. Patient-managed testing and the “Quantified Self” 9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
R
33
1. 2. 3. 4. 5. 6. 7. 8.
N C O
32
Contents
U
31
© 2014 Published by Elsevier B.V.
C
24 27 26 25 28 30 29
Q2
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
46
☆ This article is part of a Special Issue entitled: 20 years of Proteomics. ⁎ Tel.: + 1 301 728 1451. E-mail address: [email protected].
http://dx.doi.org/10.1016/j.jprot.2014.03.005 1874-3919/© 2014 Published by Elsevier B.V.
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
JOUR NAL OF P ROTEOM ICS XX ( 2014) X XX–XX X
58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83
2. 1964: molecular anatomy
86
In 1964, when I was 15 years old, my father (Norman G. Anderson, who has attended several of the Siena meetings and, at 95, is still contributing important ideas to science) was formulating what he called the Molecular Anatomy (MAN) Program at Oak Ridge National Lab. In summarizing the idea, he wrote in Science [1]:
90 91 92 93 94 95 96
N
89
U
88
C
85 84
87
Molecular anatomy is concerned with the description, at the molecular level, of the structure and organization of cells and tissues. It is the logical extension of microscopic anatomy, and it will ultimately be the basis of the molecular pathology of human cells.
97 98 99 100 101 102
3. 1974: 2-D gels
121 120
F
57
O
56
R O
55
C
54
E
53
R
52
R
51
O
50
P
September 2014 is an interesting point from which to look back at the history of what has been called proteomics, and I am grateful to the organizers of the 10th Siena meeting for the suggestion to do so — especially given that it will be 20 years since the first of these seminal conferences. I have been lucky enough to be present for all of them and in the process fallen in love with the city and its many treasures, along with the amazing, unique and wonderful circle of enthusiasts who organize and attend every other year. In the following I will look at the last 20 years of proteomics embedded in a personal reminiscence covering a larger swath of time — the five decades from 1964 to 2014, plus a peak 10 years into the future. Today we are in a period of real optimism regarding the part of proteomics that interests me most — proteins as diagnostic biomarkers. This is a huge improvement on the recent history of the biomarker field, in which enormous effort was expended with disappointing results. While genomics made major strides and produced significant clinical results, particularly in cancer (a genetic disease), proteomics did not produce any bona fide (i.e., clinically approved) biomarkers. Of course genomics is meant to be straightforward — DNA is digital, and so progress in this field probably follows something very much like Moore's Law. Protein science is by comparison very messy, technically much more difficult, and represents an altogether deeper layer of biology. Despite this excuse, it was a frustrating time. At the end of the day, however, protein biomarkers, together with some metabolites, represent the best tools we have for objectively tracking changes in wellness and the emergence, classification and control of most diseases (personalized medicine). Persistence in addressing the challenges is therefore justified. By my reckoning, something like proteomics has actually existed for most of my life. Here are a few decade-sized snapshots, including one that has not yet occurred.
103
In 1974, Pat O'Farrell and Joachim Klose were preparing to publish the initial descriptions of 2-D electrophoresis [3,4] — the revolutionary method that first sparked widespread belief in the idea that we could look at many proteins at once. I went to visit Pat in Boulder Colorado where he had just finished his PhD, and he described the difficulty of getting his “mere method” published. One look at the gel picture (of the Escherichia coli proteome) in the finally-accepted JBC paper had made me a believer. I almost gave up on the technique after the first month of completely blank gels — “luckily” it turned out that my brand new bottle of Coomassie Blue from Fisher Scientific contained a completely unrelated blue dye with no affinity for proteins (Fisher apologized and said they would put the right compound in future bottles). Re-staining with real Coomassie Blue showed lots of spots, and pretty picture addiction set in. In 1975 I started working with my father to improve 2-D technology and we built a group at Argonne National Laboratory to bring some real engineering resources to bear on the problem (resulting in the original Iso-Dalt 2-D system [5]). Our first significant paper using the method described the 2-D pattern of human blood plasma [6], and in it we identified all the major spots by immunoprecipitating each protein with a specific antibody from the collection generated by the Behring Institute in Marburg, Germany for diagnostic purposes. The plasma 2-D pattern (still the most beautiful) summarizes an enormous amount of useful information about the most useful diagnostic sample. Slowly the technology matured, the stains got more sensitive [7], and image analysis systems began to deliver meaningful measurements of protein amounts (though the tiny computing power available then would make these approaches seem ridiculous now). In parallel, a different protein survey approach, based on selecting monoclonal antibodies recognizing some expressed epitope, was beginning to emerge from Kohler and Milstein's work at the MRC lab in Cambridge [8], where I had gone to do a PhD in protein crystallography with Max Perutz. There I met Terry Pearson, a close friend and collaborator ever since, who made the first ever commercial monoclonal at about this time
D
49
large collections of molecules (and no one had a good idea how large those collections actually were). Neither the separative power available nor the analytical sensitivity was sufficient. Nevertheless the idea that we should enumerate all the building blocks of living things, as we had systematically explored all the isotopes in the periodic table of the elements (using preparative mass spectrometry at Oak Ridge), was correct and made sense as a large-scale scientific objective. Pursuit of the goal did lead to the development of a number of important pieces of technology including zonal ultracentrifuges [1], high speed parallel biochemical analyzers [2], and the first liquid chromatography systems operating at “high” pressures (5000 psi). I spent a lot of time in my father's lab at Oak Ridge (some of it trying to “borrow” components for rocket fuel), and became accustomed to the idea that developing new tools can be fun, and that these open up new vistas in biology.
E
1. Introduction
T
48 47
Looking back, of course the tools available at the time were not really up to the task: centrifugation, disc gel electrophoresis, Edman peptide sequencing (DNA sequencing was not yet possible), polyclonal antibodies and large-bore chromatography were not capable of dealing effectively with really
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
3
JOURNAL OF P ROTEOM IC S XX ( 2014) X XX–X XX
167 168 169 170 171 172 173 174 175 176
4. 1984: large scale biology
179
By the mid-1980s it seemed to my father and me that 2-D gel technology was ready for “serious” applications, including use in diagnostics and pharmaceutical R&D. We had organized two international meetings on 2-D electrophoresis (1982 at Argonne and 1984 at the Mayo Clinic) that resulted in thick, resultspacked issues of Clinical Chemistry [10,11]. With this encouragement we decided to start a company (Large Scale Biology Corp.; LSB) to build protein databases, where we were fortunately joined by my wife Constance Seniff, who had real business experience and the wisdom to temper most, but not all, of our more grandiose ideas. During this period our software group, led by ex-physicist John Taylor, came up with programs (TYCHO [12]) that allowed hundreds of gels to be compared easily, based on spot modeling as 2-D Gaussians. A number of our gel and computer analysis systems were installed in pharma companies, and in our own lab we began to study the effects (particularly toxic effects) of drugs on protein expression pattern in rat liver [13]. A favorite series of experiments from that period examined 2-D patterns of more than a hundred mouse liver samples [14] and showed that about 35% of all the liver proteins were regulated by sex difference. Clearly a lot was going on in terms of gene regulation if we could measure it.
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
C
184
E
183
R
182
R
181
N C O
180
T
178 177
5. 1994: big 2-D
203
The first Siena meeting in 1994 marked a turning point, with the near-instantaneous adoption of the word “proteome” to characterize the objective (if not the result) of this field. The funding picture in proteomics improved substantially, and we finally succeeded in automating most aspects of the 2-D process, including an elegant kinetic silver stain machine developed with the help of Thierry Rabilloud, with the result that we could finally run and analyze hundreds of gels per week. Very high speed spot-cutters made it possible to ‘manufacture’ authentic, pure human low-abundance proteins for use as immunogens by pooling a specific spot from 1000 gels of the same sample. Plasma 2-D gels still suffered from overloading with the huge amount of albumin, and we decided to use brute force to
204 205 206 207 208 209 210 211 212 213 214 215 216
U
202 201
6. 2004: mass spec
254 253
In 2002 we left 2-D gels (and LSB) behind, and began to rethink our approach. To start with a clean slate, we wrote a review [18] covering the plasma proteome and this convinced me to return to the study of this particularly fascinating sample, the subject of our first 2-D efforts, and start a small organization called the Plasma Proteome Institute. The depth of the plasma proteome as explained in that review, encompassing a dynamic range of more than 10 [10] between albumin and some cytokines, is so far beyond the dynamic range of 2-D or conventional immunoassays that a different route was required. Mass spectrometry (MS) provides direct physical detection of analyte molecules, and seemed to offer a practical alternative if one could combine its specificity and precision [19] with the sensitivity of immunoassays. This line of thought led to the invention of the SISCAPA® technology [20], which uses a specific antibody to capture a target peptide from a tryptic digest of plasma, along with a stable isotope labeled internal standard version of the peptide. While it takes some time and expertise to generate anti-peptide antibodies of high enough affinity [21], the singular advantages of this approach in sensitivity, throughput and
255
F
166 Q3
O
165
217
R O
164
remove it from samples before analysis. Many years previously we had built recycling affinity chromatography systems [15] in which rabbit polyclonal antibody columns could be used to bind an antigen and then recycled by acid elution hundreds of times. This approach turned out to work well for depleting 10 or more high abundance plasma proteins [16]. The resulting depleted plasma 2-D maps were much cleaner and, combined with additional fractionation, finally allowed us to see 3700 spots comprising 325 identified proteins [17], although the effort involved limited this work to only one sample! This depletion technology was later licensed to Agilent as the basis for their widely used MARS columns. Along the way we merged LSB with a California company (Biosource Technologies) developing large scale protein production in tobacco plants, and went public in one of the most successful IPOs of 2000 (which we celebrated at the ‘extraordinary’ millennial Palio held that year in Siena before the 2-D meeting). Looking to the next generation of tools for proteomics, we initiated a partnership with Biosite Diagnostics to use their unique phage display and immunoassay technologies to make large-scale antibody arrays. The plan was to generate high-affinity human antibodies to 200 proteins per month covering all the tissue-specific proteins we observed in comparing gels of more than 100 human tissues in the prototype 2-D gel-based HPI (produced by a superb staff that grew to more than 50 people). A surprise patent lawsuit (not involving us directly) disrupted this effort before momentum was established, and we never got to the anticipated more sensitive, high-throughput proteomics array platform. I still wonder whether this might have overcome many of the barriers impeding biomarker progress, but in retrospect, knowing now how difficult it is to make multiplex immunoassays work, and the specificity issues they suffer from, we were lucky that fate provided an opportunity at this point to open our eyes to vacuum electrophoresis, otherwise known as mass spectrometry.
P
163
(sold to a British biotech company for four bottles of fetal bovine serum). Between that time and now, it must be said that monoclonal antibody technology has generated more successful protein biomarkers than our increasingly elegant proteomics methods, probably because an antibody can serve as both discovery tool and clinical assay. The absence of a technology gap between discovery and application for monoclonals was a feature we would later recognize as critical. Putting these two approaches together, on the theory that animals could be immunized with pure protein spots cut from 2-D gels, we put forward the idea of the “Human Protein Index” (HPI [9]), an attempt to catalog all the human proteins. This proposal was discussed at high levels of government as a possible large-scale scientific effort in 1980/81, but did not survive Ronald Reagan's election and the ensuing changes in personnel and direction.
D
162
E
161
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
4
289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
F
O
288
R O
287
337
P
286
The tools of proteomics are now finally capable of addressing biomarker questions very effectively. At the discovery level, many thousands of proteins can now be confidently detected in clinically relevant, and challenging, samples such as blood plasma, providing rough quantitative estimates of relative concentration. However this in itself is not enough to open up the biomarker field because of an unfortunate limitation of the technology being used for biomarker discovery — its complexity and low throughput [28]. In an earlier era, protein biomarkers were discovered through large scale screening of monoclonal antibodies, and the same antibodies could be used to implement simple, large-scale assays in clinical samples. In essence the gap between finding a candidate biomarker and testing it in clinical samples sets was small — the antibody that defined the biomarker also served as a high-throughput assay. Current mass spectrometry-based deep discovery platforms, like the 2-D gels before them, are, however, not directly capable of conducting large-scale biomarker validation studies. As a result, to date almost no proteomic biomarker “discovery” efforts have been followed up by clinical validation. This has proven to be a critical limitation. While observing 10,000 proteins in 10 samples may be considered good discovery technology, measuring 10 proteins in 10,000 samples is likely to be far more effective in finding a real biomarker (assuming of course that one chooses the 10 proteins wisely, or luckily). This is so because biology accommodates a huge amount of variation, both across populations and across time, that often seems to mask the “invariant” causes and effects underlying the mechanisms and indicators we want to find. Fortunately, MS-based directed assay technologies, such as the SISCAPA peptide capture approach, allow thousands of samples to be tested, and provide the technical solution to the key problem in biomarker translation. To me it provided a motivational breakthrough as well: I have progressed from protein crystallography, where it took roughly a year to complete a new structure; to 2-D gels, taking a few weeks to run, stain, scan and analyze a useful set of samples; and then to long sample prep protocols ending in nanoflow LC–MS (with 60 minute cycles — which are tantalizing, but in the end infinitely tedious to watch). Now, finally, the sample prep is automated at the level of hundreds of samples per day, and the LC–MS run takes 3 min for a 10-plex assay, while generating much more precise measurements. It can be satisfying to watch information emerge at 3 min/sample, but that turns out to be just the beginning: we found that the peptides enriched by SISCAPA capture are pure enough in many cases to analyze by MS without any chromatography. This makes it possible to use both MALDI-TOF [29,30] and direct injection MRM [23] (with Agilent's RapidFire®) analysis, which are capable of processing samples in 20 or 7 s, respectively. Instead of a ball and chain holding us back, the MS became a hungry mouth crying out for more samples. Such technology finally makes it possible to imagine proteomics experiments that can be done in one day, and places the intellectual emphasis where it really belongs — on the samples and the biology they embody. Good samples are
D
285
336 335
E
284
7. 2014: a biomarker era
T
283
C
282
E
281
R
280
R
279
O
278
C
277
elimination of immunoassay interferences have proved persuasive. SISCAPA technology was developed over the decade through a series of collaborations with Terry Pearson and the Proteomics Centre at the University of Victoria [22,23], as well as a consortium effort with Mandy Paulovich (at the Fred Hutch in Seattle) and Steve Carr (at the Broad Institute) funded by the National Cancer Institute's CPTAC program [24]. While attending the Siena meeting of 2002, I had the great good fortune to receive a phone call inviting me to join the board of directors of a global diagnostics company: Dade Behring, which included the Behring Institute that provided the antibodies we used for the first plasma protein identifications by 2-D more than 35 years earlier. Dade's brilliant executive team, charged with developing, marketing and supporting sophisticated hospital clinical analyzers, provided an eye-opening education about the factors constraining progress in this field and make it appear glacially slow from the outside. One oft-quoted factor is government regulation, but this is not the largest impediment (as the FDA will cheerfully confirm). The experts at Dade Behring responsible for deciding whether to adopt a new test (300–400 candidates were presented to them every year by academic or biotech ‘discoverers’) quickly explained why proteomics was having no impact in diagnostics [25]. We in proteomics simply had no idea what was involved in verifying and validating a new protein biomarker to the level that made its medical success likely, and thereby justify spending $3–5 M to bring it to market as an FDA-cleared test. Everything we were doing was extremely early stage, and we were not equipped, either mentally or technologically, to carry candidate markers to the required maturity. It may be obvious that scientists are often amateurs in business, just as business people are usually amateurs in science, but the lesson can take most of a lifetime to learn: make friends with really good business people if you want your work applied in the real world. After 5 years on the Dade board we made the decision to sell the company to Siemens, where today it forms the core of the second largest diagnostic test vendor in the world. By 2012 the SISCAPA approach had advanced from a curiosity to an enabling technology. Andy Hoofnagle (at the University of Washington) recognized that the intractable problems with one specific clinical immunoassay, for thyroglobulin (Tg), could be solved using the peptide capture approach [26]. Thyroid cancer patients typically have their thyroids removed as part of treatment, and hence if the thyroid-specific protein Tg ever reappears in their blood, it is a sign of cancer recurrence. However, thyroid cancer patients frequently develop interfering autoantibodies to Tg that prevent the immunoassay from registering this recurrence — a false negative test with serious consequences. The SISCAPA approach, in which those autoantibodies are destroyed during the initial sample trypsinization, was subsequently adopted by large clinical reference labs [27] in the US, with widespread adoption in most parts of the world on the horizon. The prospect of real clinical use of mass spec-based protein assays, and the growing demand for the technology to create new multiplex cancer tests, convinced us to set up a new company together with Terry Pearson (SISCAPA Assay Technologies, Inc.) where we could assemble a dedicated team to develop commercial SISCAPA platforms for research and clinical use.
N
276
U
275
JOUR NAL OF P ROTEOM ICS XX ( 2014) X XX–XX X
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393
5
JOURNAL OF P ROTEOM IC S XX ( 2014) X XX–X XX
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
100
0 0
C
8. 2024: personalized, patient-centric medicine
424
Prediction of events 10 years in the future is risky, if not foolish, but nevertheless worth trying: that which cannot be envisioned is rarely achieved. Ten years from now, a substantial fraction of medical effort will be redirected to disease prevention based on early detection of developing chronic conditions. This will occur through a major shift of emphasis and funding from drug development to biomarker development. The scientific challenges involved in
429 430 431
R
R
428
N C O
427
400
8.1. Technology
442
Analytical technology will finally cease to be a major limiting factor. The world of bioanalysis will resolve into two main branches: everything with a nucleic acid sequence will be detected and measured by next-gen sequencing (no more arrays); and almost everything else (proteins, metabolites, drugs, environmental toxicants) will be measured by mass spectrometry. Based on an extrapolation from past technical
443
110 90
Pneumonia
300
70
200
50
100 0
30
-100
10
-200
-10 11
21
31
41
40
432
Nose cold
1
30
developing new drugs and biomarkers are similar, but a significant change of mindset is required to recognize that they have comparable value and should be funded at similar levels (rather than the 100:1 current ratio in favor of therapeutics). The change is justified by the simple fact that prevention or early treatment is much less expensive than late-stage intervention, a difference that can make a decisive contribution to healthcare economics, as well as being better for the patient. Several major developments will enable this change in healthcare strategy (Fig. 3).
500
CRP Peak Area Ratio
426
U
425
20
Fig. 2 – Correlation between CRP and LPS-BP levels in all the samples shown in Fig. 1.
E
423 422
10
LPS-BP
T
421
200
F
401
O
400
300
51
61
71
81
91
101
LPS-Binding Protein Peak Area Ratio
399
R O
398
400
P
397
CRP
396
500
D
395
hard to find, and sometimes the best are those you can make yourself, literally. I started periodically collecting dried blood spot (DBS) samples in 2008 — it just requires pricking a finger, collecting 5 drops of blood on a Whatman 903 filter paper card, allowing 2 h to dry and then keeping the card in a dry envelope at + 4 °C or − 20 °C. Such cards were originally developed as an easy way to collect small blood samples from newborn babies in order to test them for a series of inherited metabolic diseases [31]. We recently ran a panel of SISCAPA assays on protein eluted from a set of 104 of these serial samples, and Fig. 1 shows the results obtained for C-reactive protein (CRP) and lipopolysaccharide binding protein (LPS-BP), both of which are acute phase biomarkers increased by inflammation and infection (CRP is also used as a cardiovascular risk marker). During the period covered here I had three head colds and one short bout of bacterial pneumonia (just before the last Siena meeting — the antibiotics probably cured me, but Santa Maria della Scala, which has cared for pilgrims in Siena for hundreds of years, contributed as well). During the latter episode, CRP increased to 350 standard deviations above my otherwise extremely constant baseline value. If this were typical of biomarkers in general, statistics would be almost unnecessary! CRP and LPS-BP proved, as expected, to be very highly correlated (Fig. 2), confirming the mechanistic consistency of the acute phase response. The experiment also confirms that proteins on DBS are well enough preserved over years to yield stable values after digestion and SISCAPA measurement.
E
394
Serial Dried Blood Spot Number (4.5 yrspan) Fig. 1 – CRP and LPS-BP levels (MRM peak area ratios vs stable isotope labeled internal standards at constant levels) measured by a SISCAPA assay panel in 104 serial dried blood spots collected by one individual at varying intervals over 4.5 years. CRP (blue diamonds, scale to the left) and LPS-BP (orange dots, scale to the right). Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
433 434 435 436 437 438 439 440 441 Q4
444 445 446 447 448 449
6
JOUR NAL OF P ROTEOM ICS XX ( 2014) X XX–XX X
1.E+02
Historical
1.E+01
Projection
1.E+00 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Year Fig. 3 – Sensitivity of state-of-the-art mass spectrometers as a function of time. Historical estimates obtained from two MS vendors yield an approximate log-linear curve from which performance expected in 2024 can be estimated.
450
458
8.2. Menu
459
473
The menu of useful protein biomarkers will have changed substantially compared to 2014. Approximately half the current clinical menu of 204 proteins [32] will be obsolete, while several hundred previously unmeasured proteins will have shown clinically validated importance. Most importantly, a majority of clinical tests will consist of multiple analytes considered as a multiplex test delivering a single answer. Among other advantages, mass spectrometry makes it easy to measure multiplex protein panels — something that is not routinely practical with the previous generation (immunoassay) technology. Given this ability, it makes intuitive sense to pair a marker that rises in a disease with a second that decreases — the ratio of these two is likely to be more robust than either alone. Larger multiplex panels can provide even more statistical power, and cover a variety of potential indicators of the same disease.
474
8.3. Dried blood spots
475
Access to diagnostic testing will expand dramatically through use of dried blood spots: samples that can be collected by an individual wherever they are, including at home. Selfcollection of dried blood spots will transform the collection of diagnostic samples for people who have difficulty traveling to a doctor's office or who need regular repeat testing. The long-term stability of proteins in dried blood spots stored at
462 463 464 465 466 467 468 469 470 471 472
476 477 478 479 480 481
C
461
E
460
R
456
R
455
O
454
C
453
N
452
U
451
1
490
The availability of a longitudinal series of samples for each patient will radically increase the diagnostic value of existing protein tests, as well as new ones. It will become standard practice to interpret each test value against the individual's personalized baseline, instead of against a population distribution, as is currently the case. The striking improvement in power for cancer tests like CA-125 applied in a longitudinal manner [33] will be generalized and applied for most indicators of chronic disease, revealing quantitative details of a person's trajectory from health to illness (and back). Access to stored dried blood spot samples will also make it possible to establish personal pre-disease baselines for newly-discovered tests.
491
8.5. Patient-managed testing and the “Quantified Self”
Five microliters of plasma contains ~1 amol (600,000 molecules) of a 50 kD protein present at 10 pg/mL. If these could be detected with near-perfect efficiency, counting statistics suggests a measurement CV of
Available online at www.sciencedirect.com
ScienceDirect www.elsevier.com/locate/jprot
Review
2
Six decades searching for meaning in the proteome☆
3Q1
Leigh Anderson⁎
4
SISCAPA Assay Technologies, Inc., P.O. Box 53309, Washington, DC 20009, USA
R O
O
F
1
5 87
AR TIC LE I NFO
ABSTR ACT
91Q6 16 0
Keywords:
This review describes one thread in a fabric of developments leading to the present state
11 17
Proteomics
of proteomics, stretching over 60 years and ending with a prediction for 2024. While
12 18
HPI
composed largely of personal reminiscences, the story offers some instructive success and
13 19
SISCAPA
failures, and appears to be nearing the long-sought goal of deep insights into real biology.
14 20
MRM
This article is part of a Special Issue entitled: 20 years of Proteomics.
15 21
MALDI
22
Biomarker
23
Plasma
35 36 37 38 39 40 41 42 43 44 45
P
D
E T E
R
34
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1964: molecular anatomy . . . . . . . . . . . . . . . . . . 1974: 2-D gels . . . . . . . . . . . . . . . . . . . . . . . . 1984: large scale biology . . . . . . . . . . . . . . . . . . 1994: big 2-D . . . . . . . . . . . . . . . . . . . . . . . . . 2004: mass spec . . . . . . . . . . . . . . . . . . . . . . . 2014: a biomarker era . . . . . . . . . . . . . . . . . . . . 2024: personalized, patient-centric medicine . . . . . . . 8.1. Technology . . . . . . . . . . . . . . . . . . . . . . 8.2. Menu . . . . . . . . . . . . . . . . . . . . . . . . . 8.3. Dried blood spots . . . . . . . . . . . . . . . . . . 8.4. Personal baselines . . . . . . . . . . . . . . . . . . 8.5. Patient-managed testing and the “Quantified Self” 9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
R
33
1. 2. 3. 4. 5. 6. 7. 8.
N C O
32
Contents
U
31
© 2014 Published by Elsevier B.V.
C
24 27 26 25 28 30 29
Q2
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
46
☆ This article is part of a Special Issue entitled: 20 years of Proteomics. ⁎ Tel.: + 1 301 728 1451. E-mail address: [email protected].
http://dx.doi.org/10.1016/j.jprot.2014.03.005 1874-3919/© 2014 Published by Elsevier B.V.
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2
JOUR NAL OF P ROTEOM ICS XX ( 2014) X XX–XX X
58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83
2. 1964: molecular anatomy
86
In 1964, when I was 15 years old, my father (Norman G. Anderson, who has attended several of the Siena meetings and, at 95, is still contributing important ideas to science) was formulating what he called the Molecular Anatomy (MAN) Program at Oak Ridge National Lab. In summarizing the idea, he wrote in Science [1]:
90 91 92 93 94 95 96
N
89
U
88
C
85 84
87
Molecular anatomy is concerned with the description, at the molecular level, of the structure and organization of cells and tissues. It is the logical extension of microscopic anatomy, and it will ultimately be the basis of the molecular pathology of human cells.
97 98 99 100 101 102
3. 1974: 2-D gels
121 120
F
57
O
56
R O
55
C
54
E
53
R
52
R
51
O
50
P
September 2014 is an interesting point from which to look back at the history of what has been called proteomics, and I am grateful to the organizers of the 10th Siena meeting for the suggestion to do so — especially given that it will be 20 years since the first of these seminal conferences. I have been lucky enough to be present for all of them and in the process fallen in love with the city and its many treasures, along with the amazing, unique and wonderful circle of enthusiasts who organize and attend every other year. In the following I will look at the last 20 years of proteomics embedded in a personal reminiscence covering a larger swath of time — the five decades from 1964 to 2014, plus a peak 10 years into the future. Today we are in a period of real optimism regarding the part of proteomics that interests me most — proteins as diagnostic biomarkers. This is a huge improvement on the recent history of the biomarker field, in which enormous effort was expended with disappointing results. While genomics made major strides and produced significant clinical results, particularly in cancer (a genetic disease), proteomics did not produce any bona fide (i.e., clinically approved) biomarkers. Of course genomics is meant to be straightforward — DNA is digital, and so progress in this field probably follows something very much like Moore's Law. Protein science is by comparison very messy, technically much more difficult, and represents an altogether deeper layer of biology. Despite this excuse, it was a frustrating time. At the end of the day, however, protein biomarkers, together with some metabolites, represent the best tools we have for objectively tracking changes in wellness and the emergence, classification and control of most diseases (personalized medicine). Persistence in addressing the challenges is therefore justified. By my reckoning, something like proteomics has actually existed for most of my life. Here are a few decade-sized snapshots, including one that has not yet occurred.
103
In 1974, Pat O'Farrell and Joachim Klose were preparing to publish the initial descriptions of 2-D electrophoresis [3,4] — the revolutionary method that first sparked widespread belief in the idea that we could look at many proteins at once. I went to visit Pat in Boulder Colorado where he had just finished his PhD, and he described the difficulty of getting his “mere method” published. One look at the gel picture (of the Escherichia coli proteome) in the finally-accepted JBC paper had made me a believer. I almost gave up on the technique after the first month of completely blank gels — “luckily” it turned out that my brand new bottle of Coomassie Blue from Fisher Scientific contained a completely unrelated blue dye with no affinity for proteins (Fisher apologized and said they would put the right compound in future bottles). Re-staining with real Coomassie Blue showed lots of spots, and pretty picture addiction set in. In 1975 I started working with my father to improve 2-D technology and we built a group at Argonne National Laboratory to bring some real engineering resources to bear on the problem (resulting in the original Iso-Dalt 2-D system [5]). Our first significant paper using the method described the 2-D pattern of human blood plasma [6], and in it we identified all the major spots by immunoprecipitating each protein with a specific antibody from the collection generated by the Behring Institute in Marburg, Germany for diagnostic purposes. The plasma 2-D pattern (still the most beautiful) summarizes an enormous amount of useful information about the most useful diagnostic sample. Slowly the technology matured, the stains got more sensitive [7], and image analysis systems began to deliver meaningful measurements of protein amounts (though the tiny computing power available then would make these approaches seem ridiculous now). In parallel, a different protein survey approach, based on selecting monoclonal antibodies recognizing some expressed epitope, was beginning to emerge from Kohler and Milstein's work at the MRC lab in Cambridge [8], where I had gone to do a PhD in protein crystallography with Max Perutz. There I met Terry Pearson, a close friend and collaborator ever since, who made the first ever commercial monoclonal at about this time
D
49
large collections of molecules (and no one had a good idea how large those collections actually were). Neither the separative power available nor the analytical sensitivity was sufficient. Nevertheless the idea that we should enumerate all the building blocks of living things, as we had systematically explored all the isotopes in the periodic table of the elements (using preparative mass spectrometry at Oak Ridge), was correct and made sense as a large-scale scientific objective. Pursuit of the goal did lead to the development of a number of important pieces of technology including zonal ultracentrifuges [1], high speed parallel biochemical analyzers [2], and the first liquid chromatography systems operating at “high” pressures (5000 psi). I spent a lot of time in my father's lab at Oak Ridge (some of it trying to “borrow” components for rocket fuel), and became accustomed to the idea that developing new tools can be fun, and that these open up new vistas in biology.
E
1. Introduction
T
48 47
Looking back, of course the tools available at the time were not really up to the task: centrifugation, disc gel electrophoresis, Edman peptide sequencing (DNA sequencing was not yet possible), polyclonal antibodies and large-bore chromatography were not capable of dealing effectively with really
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
3
JOURNAL OF P ROTEOM IC S XX ( 2014) X XX–X XX
167 168 169 170 171 172 173 174 175 176
4. 1984: large scale biology
179
By the mid-1980s it seemed to my father and me that 2-D gel technology was ready for “serious” applications, including use in diagnostics and pharmaceutical R&D. We had organized two international meetings on 2-D electrophoresis (1982 at Argonne and 1984 at the Mayo Clinic) that resulted in thick, resultspacked issues of Clinical Chemistry [10,11]. With this encouragement we decided to start a company (Large Scale Biology Corp.; LSB) to build protein databases, where we were fortunately joined by my wife Constance Seniff, who had real business experience and the wisdom to temper most, but not all, of our more grandiose ideas. During this period our software group, led by ex-physicist John Taylor, came up with programs (TYCHO [12]) that allowed hundreds of gels to be compared easily, based on spot modeling as 2-D Gaussians. A number of our gel and computer analysis systems were installed in pharma companies, and in our own lab we began to study the effects (particularly toxic effects) of drugs on protein expression pattern in rat liver [13]. A favorite series of experiments from that period examined 2-D patterns of more than a hundred mouse liver samples [14] and showed that about 35% of all the liver proteins were regulated by sex difference. Clearly a lot was going on in terms of gene regulation if we could measure it.
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
C
184
E
183
R
182
R
181
N C O
180
T
178 177
5. 1994: big 2-D
203
The first Siena meeting in 1994 marked a turning point, with the near-instantaneous adoption of the word “proteome” to characterize the objective (if not the result) of this field. The funding picture in proteomics improved substantially, and we finally succeeded in automating most aspects of the 2-D process, including an elegant kinetic silver stain machine developed with the help of Thierry Rabilloud, with the result that we could finally run and analyze hundreds of gels per week. Very high speed spot-cutters made it possible to ‘manufacture’ authentic, pure human low-abundance proteins for use as immunogens by pooling a specific spot from 1000 gels of the same sample. Plasma 2-D gels still suffered from overloading with the huge amount of albumin, and we decided to use brute force to
204 205 206 207 208 209 210 211 212 213 214 215 216
U
202 201
6. 2004: mass spec
254 253
In 2002 we left 2-D gels (and LSB) behind, and began to rethink our approach. To start with a clean slate, we wrote a review [18] covering the plasma proteome and this convinced me to return to the study of this particularly fascinating sample, the subject of our first 2-D efforts, and start a small organization called the Plasma Proteome Institute. The depth of the plasma proteome as explained in that review, encompassing a dynamic range of more than 10 [10] between albumin and some cytokines, is so far beyond the dynamic range of 2-D or conventional immunoassays that a different route was required. Mass spectrometry (MS) provides direct physical detection of analyte molecules, and seemed to offer a practical alternative if one could combine its specificity and precision [19] with the sensitivity of immunoassays. This line of thought led to the invention of the SISCAPA® technology [20], which uses a specific antibody to capture a target peptide from a tryptic digest of plasma, along with a stable isotope labeled internal standard version of the peptide. While it takes some time and expertise to generate anti-peptide antibodies of high enough affinity [21], the singular advantages of this approach in sensitivity, throughput and
255
F
166 Q3
O
165
217
R O
164
remove it from samples before analysis. Many years previously we had built recycling affinity chromatography systems [15] in which rabbit polyclonal antibody columns could be used to bind an antigen and then recycled by acid elution hundreds of times. This approach turned out to work well for depleting 10 or more high abundance plasma proteins [16]. The resulting depleted plasma 2-D maps were much cleaner and, combined with additional fractionation, finally allowed us to see 3700 spots comprising 325 identified proteins [17], although the effort involved limited this work to only one sample! This depletion technology was later licensed to Agilent as the basis for their widely used MARS columns. Along the way we merged LSB with a California company (Biosource Technologies) developing large scale protein production in tobacco plants, and went public in one of the most successful IPOs of 2000 (which we celebrated at the ‘extraordinary’ millennial Palio held that year in Siena before the 2-D meeting). Looking to the next generation of tools for proteomics, we initiated a partnership with Biosite Diagnostics to use their unique phage display and immunoassay technologies to make large-scale antibody arrays. The plan was to generate high-affinity human antibodies to 200 proteins per month covering all the tissue-specific proteins we observed in comparing gels of more than 100 human tissues in the prototype 2-D gel-based HPI (produced by a superb staff that grew to more than 50 people). A surprise patent lawsuit (not involving us directly) disrupted this effort before momentum was established, and we never got to the anticipated more sensitive, high-throughput proteomics array platform. I still wonder whether this might have overcome many of the barriers impeding biomarker progress, but in retrospect, knowing now how difficult it is to make multiplex immunoassays work, and the specificity issues they suffer from, we were lucky that fate provided an opportunity at this point to open our eyes to vacuum electrophoresis, otherwise known as mass spectrometry.
P
163
(sold to a British biotech company for four bottles of fetal bovine serum). Between that time and now, it must be said that monoclonal antibody technology has generated more successful protein biomarkers than our increasingly elegant proteomics methods, probably because an antibody can serve as both discovery tool and clinical assay. The absence of a technology gap between discovery and application for monoclonals was a feature we would later recognize as critical. Putting these two approaches together, on the theory that animals could be immunized with pure protein spots cut from 2-D gels, we put forward the idea of the “Human Protein Index” (HPI [9]), an attempt to catalog all the human proteins. This proposal was discussed at high levels of government as a possible large-scale scientific effort in 1980/81, but did not survive Ronald Reagan's election and the ensuing changes in personnel and direction.
D
162
E
161
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
4
289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
F
O
288
R O
287
337
P
286
The tools of proteomics are now finally capable of addressing biomarker questions very effectively. At the discovery level, many thousands of proteins can now be confidently detected in clinically relevant, and challenging, samples such as blood plasma, providing rough quantitative estimates of relative concentration. However this in itself is not enough to open up the biomarker field because of an unfortunate limitation of the technology being used for biomarker discovery — its complexity and low throughput [28]. In an earlier era, protein biomarkers were discovered through large scale screening of monoclonal antibodies, and the same antibodies could be used to implement simple, large-scale assays in clinical samples. In essence the gap between finding a candidate biomarker and testing it in clinical samples sets was small — the antibody that defined the biomarker also served as a high-throughput assay. Current mass spectrometry-based deep discovery platforms, like the 2-D gels before them, are, however, not directly capable of conducting large-scale biomarker validation studies. As a result, to date almost no proteomic biomarker “discovery” efforts have been followed up by clinical validation. This has proven to be a critical limitation. While observing 10,000 proteins in 10 samples may be considered good discovery technology, measuring 10 proteins in 10,000 samples is likely to be far more effective in finding a real biomarker (assuming of course that one chooses the 10 proteins wisely, or luckily). This is so because biology accommodates a huge amount of variation, both across populations and across time, that often seems to mask the “invariant” causes and effects underlying the mechanisms and indicators we want to find. Fortunately, MS-based directed assay technologies, such as the SISCAPA peptide capture approach, allow thousands of samples to be tested, and provide the technical solution to the key problem in biomarker translation. To me it provided a motivational breakthrough as well: I have progressed from protein crystallography, where it took roughly a year to complete a new structure; to 2-D gels, taking a few weeks to run, stain, scan and analyze a useful set of samples; and then to long sample prep protocols ending in nanoflow LC–MS (with 60 minute cycles — which are tantalizing, but in the end infinitely tedious to watch). Now, finally, the sample prep is automated at the level of hundreds of samples per day, and the LC–MS run takes 3 min for a 10-plex assay, while generating much more precise measurements. It can be satisfying to watch information emerge at 3 min/sample, but that turns out to be just the beginning: we found that the peptides enriched by SISCAPA capture are pure enough in many cases to analyze by MS without any chromatography. This makes it possible to use both MALDI-TOF [29,30] and direct injection MRM [23] (with Agilent's RapidFire®) analysis, which are capable of processing samples in 20 or 7 s, respectively. Instead of a ball and chain holding us back, the MS became a hungry mouth crying out for more samples. Such technology finally makes it possible to imagine proteomics experiments that can be done in one day, and places the intellectual emphasis where it really belongs — on the samples and the biology they embody. Good samples are
D
285
336 335
E
284
7. 2014: a biomarker era
T
283
C
282
E
281
R
280
R
279
O
278
C
277
elimination of immunoassay interferences have proved persuasive. SISCAPA technology was developed over the decade through a series of collaborations with Terry Pearson and the Proteomics Centre at the University of Victoria [22,23], as well as a consortium effort with Mandy Paulovich (at the Fred Hutch in Seattle) and Steve Carr (at the Broad Institute) funded by the National Cancer Institute's CPTAC program [24]. While attending the Siena meeting of 2002, I had the great good fortune to receive a phone call inviting me to join the board of directors of a global diagnostics company: Dade Behring, which included the Behring Institute that provided the antibodies we used for the first plasma protein identifications by 2-D more than 35 years earlier. Dade's brilliant executive team, charged with developing, marketing and supporting sophisticated hospital clinical analyzers, provided an eye-opening education about the factors constraining progress in this field and make it appear glacially slow from the outside. One oft-quoted factor is government regulation, but this is not the largest impediment (as the FDA will cheerfully confirm). The experts at Dade Behring responsible for deciding whether to adopt a new test (300–400 candidates were presented to them every year by academic or biotech ‘discoverers’) quickly explained why proteomics was having no impact in diagnostics [25]. We in proteomics simply had no idea what was involved in verifying and validating a new protein biomarker to the level that made its medical success likely, and thereby justify spending $3–5 M to bring it to market as an FDA-cleared test. Everything we were doing was extremely early stage, and we were not equipped, either mentally or technologically, to carry candidate markers to the required maturity. It may be obvious that scientists are often amateurs in business, just as business people are usually amateurs in science, but the lesson can take most of a lifetime to learn: make friends with really good business people if you want your work applied in the real world. After 5 years on the Dade board we made the decision to sell the company to Siemens, where today it forms the core of the second largest diagnostic test vendor in the world. By 2012 the SISCAPA approach had advanced from a curiosity to an enabling technology. Andy Hoofnagle (at the University of Washington) recognized that the intractable problems with one specific clinical immunoassay, for thyroglobulin (Tg), could be solved using the peptide capture approach [26]. Thyroid cancer patients typically have their thyroids removed as part of treatment, and hence if the thyroid-specific protein Tg ever reappears in their blood, it is a sign of cancer recurrence. However, thyroid cancer patients frequently develop interfering autoantibodies to Tg that prevent the immunoassay from registering this recurrence — a false negative test with serious consequences. The SISCAPA approach, in which those autoantibodies are destroyed during the initial sample trypsinization, was subsequently adopted by large clinical reference labs [27] in the US, with widespread adoption in most parts of the world on the horizon. The prospect of real clinical use of mass spec-based protein assays, and the growing demand for the technology to create new multiplex cancer tests, convinced us to set up a new company together with Terry Pearson (SISCAPA Assay Technologies, Inc.) where we could assemble a dedicated team to develop commercial SISCAPA platforms for research and clinical use.
N
276
U
275
JOUR NAL OF P ROTEOM ICS XX ( 2014) X XX–XX X
Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393
5
JOURNAL OF P ROTEOM IC S XX ( 2014) X XX–X XX
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
100
0 0
C
8. 2024: personalized, patient-centric medicine
424
Prediction of events 10 years in the future is risky, if not foolish, but nevertheless worth trying: that which cannot be envisioned is rarely achieved. Ten years from now, a substantial fraction of medical effort will be redirected to disease prevention based on early detection of developing chronic conditions. This will occur through a major shift of emphasis and funding from drug development to biomarker development. The scientific challenges involved in
429 430 431
R
R
428
N C O
427
400
8.1. Technology
442
Analytical technology will finally cease to be a major limiting factor. The world of bioanalysis will resolve into two main branches: everything with a nucleic acid sequence will be detected and measured by next-gen sequencing (no more arrays); and almost everything else (proteins, metabolites, drugs, environmental toxicants) will be measured by mass spectrometry. Based on an extrapolation from past technical
443
110 90
Pneumonia
300
70
200
50
100 0
30
-100
10
-200
-10 11
21
31
41
40
432
Nose cold
1
30
developing new drugs and biomarkers are similar, but a significant change of mindset is required to recognize that they have comparable value and should be funded at similar levels (rather than the 100:1 current ratio in favor of therapeutics). The change is justified by the simple fact that prevention or early treatment is much less expensive than late-stage intervention, a difference that can make a decisive contribution to healthcare economics, as well as being better for the patient. Several major developments will enable this change in healthcare strategy (Fig. 3).
500
CRP Peak Area Ratio
426
U
425
20
Fig. 2 – Correlation between CRP and LPS-BP levels in all the samples shown in Fig. 1.
E
423 422
10
LPS-BP
T
421
200
F
401
O
400
300
51
61
71
81
91
101
LPS-Binding Protein Peak Area Ratio
399
R O
398
400
P
397
CRP
396
500
D
395
hard to find, and sometimes the best are those you can make yourself, literally. I started periodically collecting dried blood spot (DBS) samples in 2008 — it just requires pricking a finger, collecting 5 drops of blood on a Whatman 903 filter paper card, allowing 2 h to dry and then keeping the card in a dry envelope at + 4 °C or − 20 °C. Such cards were originally developed as an easy way to collect small blood samples from newborn babies in order to test them for a series of inherited metabolic diseases [31]. We recently ran a panel of SISCAPA assays on protein eluted from a set of 104 of these serial samples, and Fig. 1 shows the results obtained for C-reactive protein (CRP) and lipopolysaccharide binding protein (LPS-BP), both of which are acute phase biomarkers increased by inflammation and infection (CRP is also used as a cardiovascular risk marker). During the period covered here I had three head colds and one short bout of bacterial pneumonia (just before the last Siena meeting — the antibiotics probably cured me, but Santa Maria della Scala, which has cared for pilgrims in Siena for hundreds of years, contributed as well). During the latter episode, CRP increased to 350 standard deviations above my otherwise extremely constant baseline value. If this were typical of biomarkers in general, statistics would be almost unnecessary! CRP and LPS-BP proved, as expected, to be very highly correlated (Fig. 2), confirming the mechanistic consistency of the acute phase response. The experiment also confirms that proteins on DBS are well enough preserved over years to yield stable values after digestion and SISCAPA measurement.
E
394
Serial Dried Blood Spot Number (4.5 yrspan) Fig. 1 – CRP and LPS-BP levels (MRM peak area ratios vs stable isotope labeled internal standards at constant levels) measured by a SISCAPA assay panel in 104 serial dried blood spots collected by one individual at varying intervals over 4.5 years. CRP (blue diamonds, scale to the left) and LPS-BP (orange dots, scale to the right). Please cite this article as: Anderson L, Six decades searching for meaning in the proteome, J Prot (2014), http://dx.doi.org/10.1016/ j.jprot.2014.03.005
433 434 435 436 437 438 439 440 441 Q4
444 445 446 447 448 449
6
JOUR NAL OF P ROTEOM ICS XX ( 2014) X XX–XX X
1.E+02
Historical
1.E+01
Projection
1.E+00 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Year Fig. 3 – Sensitivity of state-of-the-art mass spectrometers as a function of time. Historical estimates obtained from two MS vendors yield an approximate log-linear curve from which performance expected in 2024 can be estimated.
450
458
8.2. Menu
459
473
The menu of useful protein biomarkers will have changed substantially compared to 2014. Approximately half the current clinical menu of 204 proteins [32] will be obsolete, while several hundred previously unmeasured proteins will have shown clinically validated importance. Most importantly, a majority of clinical tests will consist of multiple analytes considered as a multiplex test delivering a single answer. Among other advantages, mass spectrometry makes it easy to measure multiplex protein panels — something that is not routinely practical with the previous generation (immunoassay) technology. Given this ability, it makes intuitive sense to pair a marker that rises in a disease with a second that decreases — the ratio of these two is likely to be more robust than either alone. Larger multiplex panels can provide even more statistical power, and cover a variety of potential indicators of the same disease.
474
8.3. Dried blood spots
475
Access to diagnostic testing will expand dramatically through use of dried blood spots: samples that can be collected by an individual wherever they are, including at home. Selfcollection of dried blood spots will transform the collection of diagnostic samples for people who have difficulty traveling to a doctor's office or who need regular repeat testing. The long-term stability of proteins in dried blood spots stored at
462 463 464 465 466 467 468 469 470 471 472
476 477 478 479 480 481
C
461
E
460
R
456
R
455
O
454
C
453
N
452
U
451
1
490
The availability of a longitudinal series of samples for each patient will radically increase the diagnostic value of existing protein tests, as well as new ones. It will become standard practice to interpret each test value against the individual's personalized baseline, instead of against a population distribution, as is currently the case. The striking improvement in power for cancer tests like CA-125 applied in a longitudinal manner [33] will be generalized and applied for most indicators of chronic disease, revealing quantitative details of a person's trajectory from health to illness (and back). Access to stored dried blood spot samples will also make it possible to establish personal pre-disease baselines for newly-discovered tests.
491
8.5. Patient-managed testing and the “Quantified Self”
Five microliters of plasma contains ~1 amol (600,000 molecules) of a 50 kD protein present at 10 pg/mL. If these could be detected with near-perfect efficiency, counting statistics suggests a measurement CV of
