RA Genetics 101: Glass half-full or half-empty? Part 4

The following is part of a guest post series on rheumatoid arthritis (RA) genetics for patients as well as practicing physicians, number 4 in a series of blog posts initiated in the summer of 2011 at rawarrior.com.  The overarching goal of this blog post series is to inform readers of considerable scientific progress being made in understanding disease, with the example in this case being RA, that is not yet being incorporated into mainstream healthcare.  RA is a common, complex autoimmune disease that inflicts considerable havoc and destruction in the bodies and lives of those unfortunate enough to succumb to it.  For those who wish to know about more foundational aspects of RA, please visit Kelly Young's blog Rheumatoid Arthritis Warrior, and follow her on Facebook and Twitter.  Here, the focus  is entirely on heritable aspects of RA.  

The choice of RA as an example to achieve the overarching goal is due to my friendship with Kelly Young, and the respect I hold for her and her fellow RA warriors in combating a disease that I believe our present medical establishment could address more effectively.  As a personalized medicine evangelist, it became immediately clear that genetic and genomic information available on RA was comparatively straightforward and rather extensive when viewed in the context of many other chronic, common diseases.  This presented an opportunity to step up to the plate by informing RA patients (indeed, those with autoimmune diseases in general) of the availability of new and valuable information that might impact their healthcare.  In turn, this upholds present day trends of bypassing the paternalistic method of 20th century healthcare with a more empowered, 21st century approach that capitalizes on both internet communications and knowledge of genomic medicine, as well as open access publishing.  

In part 4 of this series "RA Genetics 101: Glass half-full or half-empty?", I take the discussion of RA genetics to a more advanced level than described in Part 1, Part 2, and Part 3.  As will quickly become clear, the take home message, at least in the case of RA, is that genetic information is already available that could improve the diagnosis (and likely, treatment) of many patients.  In addressing the present topic "Subtypes of RA", a foundation is laid for the following: 1) a more realistic perspective on how diseases generally should be diagnosed, 2) new measures available to improve diagnosis (and hence, treatment), and 3) improved attention to how patients may benefit from a personalized approach to their healthcare.  As one might imagine, to do justice to this subject matter required examination of a significant amount of data published over the past decade or more.  Hence, documentation of the working hypothesis and emerging themes presented here required significantly more than the average number of sentences and paragraphs found in a typical blog post.  For this reason, Post 4 is presented here instead of at rawarrior.com, and will be commented on there in an upcoming blog post by Kelly.  

Advanced RA Genetics- Subtypes of RA

Outline:

1.    Introduction: A Working Hypothesis and three emerging themes.

2.    Subtypes of RA: RF+/RF- & ACPA+/ACPA- serotypes.

3.    Major Histocompatibility Locus (MHC) and the Human Leukocyte Antigen (HLA).

4.    Non-HLA genes that determine RA risk.

5.    HLA-typing.

6.    Additional genetic considerations for ACPA subclasses.

7.    ACPA+ vs ACPA- pathogenesis.

8.    Subtypes of ACPA- RA (and ACPA+ RA).

9.    Prognosis and treatment of different RA subtypes.

10. Search for additional RA biomarkers reveals additional RA subtypes.

11. Genomic and Network Medicine- A step towards Personalized Medicine.  

Introduction: A Working Hypothesis and three emerging themes.

In three previous posts on this blog (see part 1,  part 2, and part 3) I described the heritable (genetic/genomic) component of RA, emphasizing the practical relevance of genetics and family history to patients (and docs) seeking support for understanding this disease.  A number of influential genes and respective genetic variants, likely influencing both age of onset and disease severity, associated with RA were previously described, as well as current methods available for identifying and tracking such genetic variants in RA patients.

Here, I develop this topic area further to include a discussion of subtypes of RA, focusing mainly on seropositive (rheumatoid factor antibody positive or RF+, and anticitrullinated protein antibody positive or ACPA+) vs. seronegative (RF-; ACPA-) forms of the disease.  This discussion, in turn, sets the foundation for a working hypothesis for autoimmune disease management from which three basic themes emerge regarding how best to manage an “RA diagnosis” and potential treatment options:

1.      “Rheumatoid Arthritis” is an artefactual entity.  By this I mean that such a disease designation is an arbitrary attempt to facilitate the diagnosis and treatment of people who undoubtedly constitute a genotypic and phenotypic continuum nevertheless manifesting joint inflammation due to an autoimmune etiology as a key symptom.  Lumping such individuals into the umbrella category “RA” in the 21^st century constitutes adherence to 19^th century reductionist medical mindsets. 

2.     Genomic approaches, including SNP analysis (part 3) and HLA-typing (described below), are available today, which if incorporated into standard rheumatological care, may facilitate the diagnosis and treatment of RA, saving needless suffering of present and future generations of patients.

3.     While it is becoming clear that distinct subtypes of RA are now distinguishable, such distinctions between RA and certain other autoimmune diseases may soon begin to blend, as the overlapping symptoms, physiology, and genetic parameters merge into an autoimmune disease continuum, or more likely, an “autoimmune network” (part 5- to be completed soon).  As whole genome and exome sequencing are routinely incorporated into standard healthcare practice, artificial disease boundaries will become increasingly evident, thereby requiring more personalized healthcare.

Hence, I believe the framework established in this post will begin to destroy the myth that all patients with a diagnosis of RA have the same disease, and thus are manageable with similar, if not identical treatments.  By contrast, if one particular patient’s RA is different from another patient’s RA (indeed, they cannot then both be RA), then it must be questioned as to whether or not existing treatments are sufficient to address both diseases. 

Of course, the hypothesis addressed here, that proper understanding and treatment of autoimmune diseases cannot be achieved without first dissecting the underlying genotypic portfolio, could possibly be dispatched as semantic if one considers RA as a “syndrome” (a collection of similar symptoms that are too imprecise to characterize as a disease) rather than a disease (Stanich et al., 2009;  Scott, Wolfe, Huizinga, 2010).  While syndrome may be a more accurate description of RA, it is of little help when treating patients who are refractory to therapies resulting from disease etiologies that are not “one size fits all”.  Such clinical charades as this come as no surprise to those with a provisional diagnosis of RA.  

First, obtaining a diagnosis can, in some instances, require years of visits to multiple specialists before even a tentative diagnosis can be made.  Second, among different patients with the same diagnosis, symptoms may differ markedly from one patient to another.  In all likelihood, no two patients with a diagnosis of RA have the exact same set of symptoms, with the possible exception of identical twins.  Third, even experts can’t agree on what features constitute a correct diagnosis, just as it is not clear what criteria constitute RA disease remission.  When examining external factors such as patient symptoms, it’s all guesswork.  And all too often, the doctor does not know best (said reverently!).  Clearly, more precise diagnostic tools are required to facilitate diagnosis (and prognosis) of patients with autoimmune diseases such as RA.  

While in this post I formally introduce the concept of “Network Medicine” to the RA and autoimmune communities, along with the absolute requirement to incorporate the practice of genomic approaches to medical care whenever possible, this is not the first time I’ve mentioned my chagrin with common practices of the current medical establishment.  Indeed, I have been alluding to the need for an altered perspective for some time in various social media venues, such as Twitter and in comments on various blog posts.  Below, for example, is a comment I made on Dana Symons’s blog “At the Waters Edge” regarding her post “My Take on RA Treatments & Decisions”, which nicely serves as a prologue to the present post:

Addressing the problem of RA diagnosis and treatment, by incorporating available genetic and biochemical tools, is the focus of this post, with the goal of illustrating how 20^th century “one size fits all” clinical practices leads to unnecessary suffering of 21^st century patients.  Although in this post RA is specifically used as an example, the principles outlined here should be applicable to many other chronic, complex diseases (e.g. asthma and COPD). 

Subtypes of RA: RF+/RF- & ACPA+/ACPA- serotypes.

In addition to certain apparent differences in symptoms (aka “phenotypes”), patients can also differ in molecular entities called biomarkers.  Some patients produce rheumatoid factor (RF positive), and others not (RF negative).  Some patients generate anti-citrullinated protein antibodies (ACPA+ or CCP+ [ anti-cylic citrullinated peptide), while others do not.  The latter is now considered to be the more useful biomarker to characterize patients manifesting symptoms typical of a more severe and erosive form of RA, and which distinguishes RA into two distinct classes or subtypes (see also Trouw and Mahler, 2012).  Other biomarkers, such as erythrocyte sedimentation rate (ESR or “sed rate”) and C-reactive peptide (CRP) may also be used, though most often with diminished reliability.  It is worth noting that ACPA+ is considered to be highly diagnostic for RA with respect to other autoimmune diseases, about two-thirds of RA patients are ACPA+, and most, but not all, ACPA+ patients are also RF+.

Citrullinated proteins are generated via a particular cellular enzyme, peptidylarginine deiminase encoded by the PADI4 gene (see part 3) present inside cells, and is thought to signal targeting of cellular proteins for degradation during apoptosis (cell death). (Interestingly, the PADI4 gene can be repressed by glucocorticoids (e.g. prednisone), raising the possibility that this is one mechanism by which low dose predinisone treats RA.)  The consequence of this is that citrullinated proteins would normally be unexposed to the host immune system, so they would be recognized as foreign, and thus able to invoke an immune response.   In ACPA+ RA patients, the latter is thought to trigger a cascade of events leading to autoimmunity, with the thought being that sites of inflammation will contain many dying cells, thereby amplifying the autoimmune reactivity (ibid). This may be accompanied by a “breaking of tolerance” phase, where although a certain amount of citullination of host proteins is ordinarily tolerated by the immune system, once a threshold value is exceeded, citrullination begins to evoke an immune response.  Additionally, the process of inflammation itself may give rise to increased citrullination, particularly in the immune response cells themselves. 

By contrast with the above, less is known about how RA develops in ACPA- patients.  One thought is the latter may harbor non-citrulline antigens in dying cells, and ACPA- could itself be comprised of several subgroups (described below).  

Major Histocompatibility Locus (MHC) and the Human Leukocyte Antigen (HLA).

Most importantly, the ACPA response is highly dependent on genetic background, and no adequate description of RA susceptibility and disease treatment would be complete without first understanding genetic aspects underlying this disease.  The PRIMARY control gene, HLA-DRB1 located on chromosome 6 in the major histocompatibility complex (MHC), was briefly described in part 2 and part 3, in the context of DTC genetic testing.  This gene encodes a protein that, in most RA patients, presents an “altered self” peptide to the immune system for attack and disposal as though it were a foreign body (e.g. bacterium or virus).  A short set of 5 amino acids in the corresponding HLA-DRb1 protein, referred to as the “shared epitope” (SE; citrullinated-protein binding region on the HLA molecule for presentation by dendritic cells to T-cells), is the key portion of the protein, which is highly specific for citrullinated peptides, which turns out to be the “self”  (autoimmune) antigen.  It is much more difficult to recognize citrullinated proteins as “non-self” without inheriting an SE that is highly specific for that binding moiety. 

There are many distinct alleles of HLA-DRB1 in the human population, some of which are affiliated with ACPA+ RA, others with ACPA- RA, and still others that, surprisingly, are protective from risk of susceptibility to RA.  The primary allele affiliated with ACPA+ RA is HLA-DRB1*04, and *0401/*0401 (aka HLA-DR4) homozygotes are roughly at maximal genotypic risk for RA susceptibility and severity (see also Barton, 2011, and Fig 5 of Raychaudhuri and colleagues, 2012).  For example “The odds ration (OR) for one copy of the respective HLA-DRB1 SE allele is 4.37 (or almost 4.5-fold greater than in patients without this allele), whereas the OR for two copies is 11.79, all other factors being equal (Bax et al. 2011). 

The primary allele affiliated with ACPA- RA is HLA-DRB1*03 (aka HLA-DR3).  For both ACPA+ and ACPA- RA, there are multiple distinct alleles that provide a “spectrum of susceptibility” to RA (see, for example, Fig 5 of Raychaudhuri and colleagues, 2012), such that even within one or the other ACPA category there may be considerable variation in reactivity against self proteins.  Layered on top of this complexity is the fact that some HLA-DRB1 alleles are protective, such as HLA-DRB1*1301, although it appears that this allele provides protection only for the ACPA+ category.  Given the diploid nature of human genomes, a priori any given patient may contain a combination of HLA-DRB1 alleles that confers greater or lesser reactivity to non-self antigens.  These are most often directly inherited from the parents, though in rare instances, de novo mutations could occur that may increase or decrease reactivity to self antigens. 

The complexity of inheritance of risk of susceptibility to RA, based on the variety of specific HLA-DRB1 alleles present in the human population, is even further exacerbated by findings described in the recent studies of Raychaudhuri and colleagues, 2012.  They showed that two additional HLA genes, HLA-B and HLA-DPB1, also encoded in the MHC, have significant albeit more modest contributions than HLA-DRB1 to susceptibility to ACPA+ RA. 

To summarize, most if not all ACPA+ patients, can expect to carry at least one particularly potent HLA-DRB1 SE allele, and possibly also one or more potent HLA-B and/or HLA-DPB1 risk alleles.  These risk alleles predispose carriers to extraordinary efficiency of immune recognition (by T cells, primarily), except in this instance immunity happens to be generated against self.  Consequently, in ACPA+ patients this triggers the attack of normal human proteins, the destruction of normal human cells, and a progressive escalation of the immune response against self antigens.  Of course, greater exacerbations of RA disease activity would be experienced in the homozygous state for each of the 3 respective genes mentioned above.  The entire process is thought to be responsible for generating “non-self” antigens that invoke an autoimmune response in individuals who have the ACPA+ subtype of RA (demonstrated by a positive anti-CCP test).  For a deeper understanding of biochemical and cellular aspects of the feed-forward process resulting in joint inflammation and destruction, see Figure 3 in the review by van Venrooij and colleagues. 

Finally, it is important to not trivialize the relevance of environmental attributes when considering susceptibility to disease.  For example, there is not only correspondence between ACPA+ and HLA-DRB1 allele on RA susceptibility risk, but also correspondence between these respective factors and smoking behavior, including the extent of the behavior (pack-years).  This suggests that individuals who inherit an HLA-DRB1 allele prognostic to ACPA+ RA should most certainly avoid smoking as well as smoke-filled environments.  It is thought that smoking leads to increased citrullination of proteins in the lungs (and probably elsewhere), predisposing the immune system of an HLA-DRB1 carrier to an initial reaction which can subsequently become amplified and lead to full blown autoimmunity. 

Non-HLA genes that determine RA risk.

Further increasing the overall complexity of susceptibility to RA is the presence of additional, non-HLA, genes located elsewhere in the genome that can attenuate the response to self antigens provided mainly through HLA genes.  Eight of these, including several key ones such as PTPN22 are tested by 23andMe, and were described in part 2 and part 3 of this blog post series.  Of these eight (and others- approximately 30 total with low effect size; not discussed here), only partial overlap is observed between ACPA+ AND ACPA- patients (see Fig. 1 and Table 3 of Viatte and colleagues).  Such differences are consistent with ACPA+ and ACPA- RA representing distinct diseases with similar presenting symptoms.  While non-HLA genes contribute to the overall risk of susceptibility to RA, even their combined effect size (OR) is quite low, indicating that HLA remains the predominant susceptibility locus for RA risk.  The vast majority of the genetic contribution to ACPA+ RA risk actually can be explained by two genes, HLA-DRB1 plus PTPN22. 

HLA-typing.

From the above discussion, it would seem that testing of HLA would provide a significant indication of potential susceptibility of RA.  Although HLA typing is clinically uncommon (and likely expensive: see ProImmune HLA Tissue Typing Service), a first approximation of HLA status can be obtained rather inexpensively by DTC genetic testing via 23andMe (HLA region- rs6457617) or deCodeMe (HLA-DRB1- rs660895), as previously described in part 2.  Again, however, it is important to keep in mind that common genetic variants represent risk factors that, of themselves and often even when present in combinations, are not solely determinative of disease susceptibility.  Other risk alleles, scattered across the genome, some of which may have a protective capacity but have not yet been characterized, are possibly involved, and 40 to 50% of the total RA disease risk is impacted by environmental factors (Part 2).  

So, a priori, might it be advantageous to evaluate HLA genotype for pre-determinative diagnostic purposes, especially if one is of female gender in a family with a history of autoimmune disease?  The party line of most clinical experts is NO (see, for example, UpToDate: HLA and other susceptibility genes in rheumatoid arthritis).  However, much of the rationale behind this view may be related to current clinical inadequacy in obtaining and interpreting genetic data, and ensuing legal liabilities inherent to the medical arena.  Additionally, it’s possible that genetic testing could cause unnecessary worry on the part of patients with a family history of RA who, although testing positive for HLA and possibly other RA risk alleles, may never contract the disease for any one of a variety of reasons (non-smoking family, etc.).  Also, much of the discussion on the topic of pre-determinative genetic testing has focused on “population-based” screening, rather than screening individuals at high risk.  Such considerations prompted Karlson et al, 2011 to investigate the possibility of pre-determinative diagnosis, and while their results showed that genetic-based risk analysis is presently inadequate for population-based screening, it “significantly stratifies individuals for RA risk beyond clinical risk factors alone”. 

Personally speaking, if I myself felt at risk of due to family history of RA or other autoimmune disease, I would desire to be tested for HLA as soon as possible, with the goal of limiting potential disease damage via preemptory drug intervention.  While there’s always the issue of possible rare mutations present elsewhere in the genome that may contribute to a patient’s RA risk, within a decade this potential caveat may become less germane, as genome sequencing becomes a standard part of the patient medical record, although identifying such rare mutations and related issues of genome interpretation may require additional time to process (Kohane and Shendure, 2012; MacArthur and Lek, 2012; Kobolt, 2012). 

Since it’s likely that not all RA patients or others with a family history of autoimmune disease, will be able to obtain their respective genotypic information in the near term, disparities in medical treatment may inevitably arise.  Alternatively, some patients may wish to remain ignorant of their personal genotype, in spite of obvious advantages to knowing it.  Certainly, however, over time there will be greater advantages to testing as opposed to not testing.  And with time, it seems likely that a broad combination of genetic and non-genetic (e.g. environmental) factors together will be identified as useful for diagnosis of RA and other autoimmune diseases (Klareskog et al, 2004; Javierre et al., 2011).  Such variables must necessarily be determined empirically from one human disease to the next.  Such investigations will be facilitated by the ability of IBM’s Watson supercomputer to rapidly sort through clinical algorithms to make first-approximation, hierarchical predictions of disease susceptibility, pathological etiology, and best-practice treatment options available to a broad spectrum of ethnically diverse patients. 

Additional genetic considerations for ACPA subclasses.

Both major subclasses of RA, ACPA+ and ACPA-, are equally heritable; twin studies showed 68% heritability for ACPA+ vs. 66% heritability for ACPA-, although most of the ~30 common gene variants presently associated with RA (including HLA-DR4) are affiliated with the ACPA+ category (Note: most GWAS studies have been enriched with patients of the ACPA+ category; see also Plenge, 2009).  It is nonetheless compelling that known genetic risk factors explain more of the genetic variance of RA than observed for just about any other common, complex disease, including other autoimmune diseases (e.g. ankylosing spondylitis, psoriatic arthritis, systemic lupus erythematosis, etc.). 

ACPA+ vs ACPA- pathogenesis.

ACPA+ RA is more straightforward in terms of diagnosis, prognosis, and pathogenesis than ACPA- RA, or indeed most any other autoimmune disease, given the identification of citrullinated proteins as “self” antigens responsible for initiating autoimmunity.  Unfortunately, ACPA+ RA prognosis is most often worse than that of ACPA- RA.  Differences in pathogenesis and prognosis of ACPA+ RA and ACPA- RA is consistent with the fact that each is associated with different genetic risk factors as well as environmental factors. 

ACPA+ autoantibodies may be present up to a decade prior to the start of symptoms, and both increased levels and expanded specificities coincide with the appearance of clinical symptoms and a full-blown RA diagnosis (see also excellent review by Huizinga and colleagues, 2012).  Additional factors, such as the presence of two copies in homozygous carriers of HLA-DRB1*04 confer the highest odds of early mortality from extraarticular diseases, such as cardiovascular disease.  In the case of ACPA+ RA, it appears that the autoantibody is more than a mere biomarker for RA, being actually involved in the pathogenesis itself, including bone destruction.  ACPA+ autoantibodies effectively ramp up the immune system (i.e. inflammation) and inflict more damage, including greater radiological joint damage, than observed with ACPA- RA, and response to various treatments is worse with ACPA+ RA (Huizinga and colleagues, 2012).  Although ACPA- patients have more fibrosis and increased thickness of the synovial lining layer, they are also more likely than ACPA+ patients to achieve drug-free remission.  For ACPA+ RA, it is unclear why systemic loss of tolerance against citrullinated peptides most often presents specifically at the joints at comparatively early stages of the disease. 

The summation of evidence suggests that ACPA+ RA and ACPA- RA are, in fact, two different diseases, and that further studies on each disease should thus involve respectively separated populations of patients.  In the broader context of the complexity of autoimmune disease, this is consistent with the fact that ACPA+ RA shares PTPN22 function with type I diabetes but not with ACPA- RA.  This raises the possibility that ACPA+ RA may actually have more in common genetically and immunologically with type 1 diabetes than with ACPA- RA, in spite of their respective symptomatic differences.  

Subtypes of ACPA- RA (and ACPA+ RA).

Perhaps it may seem satisfying if the ACPA+ and ACPA- disease subtypes were, of themselves, discrete disease entities, but as one might anticipate that is not the case.  For example, Terao et. al., 2012 demonstrated that in a Japanese population ACPA- RA consists of at least two subtypes based on whether patients were RF+ or RF-.  The distinction between the two ACPA- subtypes correlated with their respective HLA-DRB1 genotypes, consistent with studies showing that specific HLA-DRB1 alleles are associated with ACPA- RA vs. ACPA+ RA (see also Mackie et al., 2012). 

In addition to the above example, Lundberg et. al., 2012 showed that in a Swedish population, seventeen distinct RA subsets could be identified based on their ACPA fine specificity profiles (with limited cross-reactivity), to just four different citrullinated peptides (enolase, vimentin, fibrinogen, and type II collagen).  It’s particularly interesting to note that in this study 18% of ACPA-negative patients were positive for at least one ACPA fine specificity, suggesting that even the ordinary ACPA status designation is provisional.  Moreover, 14% of ACPA+ patients (as determined by reactivity using the standard CCP2 test) were negative for all four ACPA fine specificities used here, suggesting other citrullinated targets exist (meaning additional opportunity to identify ACPA specificities in ACPA- patients).  Interestingly, this study also showed that ACPA+ RA, associated with both SE and PTPN22 as well as with smoking, corresponds essentially with only the citrullinated enolase and vimentin antigens. 

The difference in odds ratios for susceptibility risk among patients in the seventeen distinct RA subsets (with respect to: HLA-DRB1 SE, PTPN22 status, smoking, and enolase or vimentin autoantibodies) was striking, ranging from 1 (negative for each of the four variables) to 50 (positive for each variable; see Table 3 of Lundberg et al., 2012).  Unfortunately, no correlation between ACPA fine specificity and clinical characteristics was demonstrated in this study.  (Below, this point is discussed further in terms of potential prognostic value of ACPA fine specificities.) 

It is important to realize the extent of ambiguity with which studies such as these are capable of successfully categorizing patients into particular subtypes; for example, up to 1% of healthy controls, and up to 6% of non-RA disease controls, have been found to be ACPA+.  These and other findings suggest that ACPA autoantibody is neither necessary (ACPA- patients) nor sufficient (ACPA+ healthy controls) for RA disease.  This is in keeping with our hypothesis that “RA” is simply an umbrella term for a spectrum of different diseases that share overlapping phenotypes (e.g. joint pain).  HLA-DRB1 alone confers a “spectrum of risk susceptibility” to RA disease.

Prognosis and treatment of RA subtypes.

Differences in prognosis are relevant when considering treatments that may work for a patient of one RA subtype versus another.  ACPA+ and ACPA- subtypes differ in response to methotrexate treatment (no effect of methotrexate on progression to RA in ACPA- individuals), and for patients with high levels of ACPA, methotrexate alone is insufficient to control the disease.  ACPA+ RA is is associated with greater radiological joint damage, increased extra-articular manifestations (e.g. ischemic heart disease), decreased likelihood of remission, as well as with different response to therapy, indicating the importance of determining ACPA status early in the course of disease.  Since seroconversion of ACPA status is uncommon, repeating ACPA measurements in daily practice is essentially unnecessary.  With ACPA status alone reflecting significant differences in pathogenesis and treatment requirements, it’s reasonable to imagine that further discrimination among RA subtypes would be advantageous to physicians and their patients.  This illustrates the need for additional biochemical methods of distinguishing various RA subclasses from one another.

Search for additional RA biomarkers reveals additional RA subtypes.

Ideally, future analyses of RA etiology might take a direction similar to that recently reported for Sjorgren’s syndrome, where multiple subtypes of the disease are distinguishable by their various serological profiles.  While it is not understood how each different autoantibody identified for Sjogren’s syndrome correlates with disease etiology, it nevertheless provides a mechanism for improving disease management based on correlating serotypes with respective clinical associations.  

Recently, such an attempt to correlate ACPA status with clinical features of ACPA+ RA patients was made by Willemze et al., 2012, who identified 64 different subgroups of 661 patients that could be distinguished based on reactivity to 9 different citrullinated antigens.  While considerable heterogeneity was observed in terms of ACPA fine specificity, unfortunately no correlation could be made between the various subgroups and the clinical characteristics chosen for analysis (e.g. morning stiffness, swollen joint count, radiographically-assessed joint destruction, ESR, CRP, RF, and DMARD-free remission).  This study would suggest that stratification of ACPA+ patients by citrullinated antigen specificity with respect to clinical manifestations provides no further insight into potential subtypes of ACPA+ RA.  Sadly, however, this study is limited by the equivocal nature of symptom-reporting by various patients, as well as the general reliability of the particular inflammatory markers used as proxies for true clinical features (Figure 4: ESR, CRP, RF).  Furthermore, as the authors themselves noted, they may have simply missed key citrullinated antigens that would reveal distinct subgroup specificities.  Additionally, an analysis of ACPA- patients by one of the same authors, using a somewhat similar approach to that mentioned above, gave the similar conclusion that subgroups of ACPA- patients could not readily be catalogued. 

Notwithstanding the above results, a study by Shi et al, 2011, involving one of the same authors (T. Huizinga), showed that a different autoantibody, one recognizing carbamylated peptides (anti-carP), is found in up to 45% of RA patients, including up to 30% of ACPA- patients.  Most importantly, in this specific instance the presence of the anti-carbamylated protein autoantibody in ACPA- patients correlated with a more severe course of the disease.  These results suggest that there are actually no fewer than two readily distinguishable subgroups of ACPA- RA, anti-carP+ and anti-carP-.  Conceivably, these two groups might be subdivided further based on RF reactivity, or ACPA fine specificity (found even in patients classified as ACPA-), as described above. 

Similarly, the results of Shi et al, 2011 revealed that there are no fewer than two subgroups of ACPA+ patients.  While presence of anti-carP in ACPA- patients predicted a more severe course of disease, as assessed particularly by radiological damage estimated using the Sharp-van der Heijde method, their presence in ACPA+ patients did not increase further the damage already resulting from ACPA+ status. 

Carbamylation is mediated by cyanate, which is increased in both smoking and inflammation.  This provides a possible rationale whereby ACPA- patients, while lacking the HLA-DRB1 allele ordinarily associated with the effect of smoking (HLA-DRB1 *04), may nevertheless be at increased risk upon exposure to this environmental agent.  ACPA autoantibodies are highly specific to RA, but presently it is unknown whether anti-carP is similarly unique to RA or present in other autoimmune disorders. 

Summarizing the above, identification of anti-carP autoantibody, and its correlation with disease severity, suggests no fewer than four distinct subsets of RA: ACPA+ anti-carP+, ACPA+ anti-carP-, ACPA- anti-carP+, and ACPA- anti-carP- (Shi et al, 2011).  Conceivably, this could be subdivided into 6 respective subsets based on RF status (i.e. ACPA- only), though to my knowledge no clear evidence supports a specific role of RF itself in impacting RA disease course.  The above findings are but a preliminary indication of the potential granularity that may distinguish a normalized population of autoimmune patients of provisional RA diagnosis.  Indeed, closer inspection has revealed that additional autoantibody specificities, in combination, may have the potential to distinguish up to 70% of ACPA- RA patients.  Whether or not these new specificities will serve as reliable as ACPA autoantibody as prognostic biomarkers remains to be determined. 

Genomic and Network Medicine- A step towards Personalized Medicine. 

Returning to the overarching hypothesis and themes presented initially, it seems the time is ripe to once and for all destroy the prevailing myth that RA is actually a finite disease entity.  While it may be an improvement to characterize it as a “syndrome”, this is an even more nebulous term for lumping together patients with different disease etiologies.  Every “RA patient” is unique, and thus, doing justice to each really requires a much more personalized approach to their care.  It is wholly unfortunate that the present day medical establishment is incapable of providing such care in the 21^st century (for similar views on this point, see The Creative Destruction of Medicine, by Eric Topol). 

At a very minimum, proper medical care in this century will require health care practitioners to utilize clinically actionable genomic information to facilitate disease diagnosis and prognosis.  Ideally, actionable biomarkers identified by proteomic methods would likewise be routinely available for sub-typing purposes.  For patients suspected of having an RA diagnosis, or else having a family member with RA (or perhaps even another autoimmune disease), HLA variants alone confer a major influence on susceptibility risk, such that one might argue it is irresponsible of the medical establishment to be treating such patients without the a priori availability of respective genotypic data.  That’s not to say that genomic information from all RA or other autoimmune patients will be readily interpretable in every case; indeed, there will be many instances where considerable ambiguity exists in spite of such information.  But, for the sake of patients where a straightforward relationship between genetics (e.g. female gender and/or HLA-DRB1*04 genotype) and environment (e.g. smoking history) is evident, such individuals will undoubtedly be well served by having such information as part of their electronic health record, well in advance of onset of symptoms. 

But what about individuals for whom genomic information is not as straightforward?  What might be the advantage to them and their health care practitioners by having access to corresponding genomic information?  And, even in instances where the interpretation of genomic information is relatively straightforward, as in the instance mentioned immediately above, can management of the respective patients’ illness actually be improved? 

Plenge, 2009 noted that genetic analysis of autoimmune diseases provides at least three distinct advantages to improving patient care: 1) insight into disease pathogenesis, 2) identification of clinically relevant subsets of disease, and 3) clinical prediction.  Of these, perhaps the first, insight into disease pathogenesis, is of greatest significance.  Identification of specific genes involved in RA and other autoimmune diseases is a prerequisite to determining the biological functions of such genes, their respective proteins, and their roles in disease pathogenesis.  Once characterized, such information allows better clinical understanding and management of the disease, while simultaneously identifying potential targets for drug intervention.  As described by Zhernakova et al., 2009 (see their respective Figure 1 and Table 3), such genetic studies have already pinpointed three common pathways involved in the pathogenesis of diverse autoimmune diseases: 1) T-cell differentiation, 2) immune-cell signaling, and 3) innate immunity and TNF signaling.  Progress such as this in understanding the pathogenesis of autoimmune disease is now reflected by the availability of several new classes of biologic agents used to control RA and other autoimmune diseases. 

Consistent with a rather limited number of biological pathways being responsible for disparate autoimmune diseases, it is not surprising that risk variants which affect one particular autoimmune disease may contribute to susceptibility of another autoimmune disease.  Indeed, the results of Zhernakova et al., 2009 that defined 3 key pathways involved in autoimmunity, likewise showed the presence of considerable overlap among different autoimmune diseases of specific gene variants (see their respective Table 3 and Supplementary Information).  Hence, gene variants known to contribute specifically to risk of susceptibility of RA (see Part 3- tables) may also contribute to the risk of one or more other autoimmune diseases (e.g. ankylosing spondylitis, psoriatic arthritis, systemic lupus erythematosis, Crohn’s disease, type 1 diabetes, celiac disease, etc.).  Indeed, roughly 45% of identified immune-mediated disease risk variants are associated with multiple (but not all) common autoimmune diseases. (Daly and colleagues, 2011).

The significance of the above findings is at least two-fold.  First, based on inheritance of genetic variants common to more than one autoimmune disease, it means autoimmune diseases generally tend to cluster in families.  Thus, a parent having RA is at increased risk of having a child with, if not RA itself, another autoimmune disease such as AS or T1D (Hemminki et al., 2009a; Hemminki et al., 2009b).  This “shared familial aggregation of susceptibility to autoimmune diseases” contributes significantly to the overall maintenance of autoimmune diseases in the human population (Hemminki et al., 2009a).  Second, depending on which particular alleles are inherited, a patient may have a disease that is either similar to or significantly different from a relative having the same or a similar diagnosis, with respect to age of onset, disease severity, phenotypic properties (symptoms), and refractoriness to treatments.  Of course, as mentioned in Post 2 and Post 3, environmental variables are also part of the overall disease susceptibility equation, and can also have a major impact on clinical outcome often regardless of genotypic background. 

Direct to consumer genetic testing company 23andMe (see Post 2 and Post 3, with respect to common genes involved in RA risk) in their blog “The Spittoon” presented a nice post related to the present theme, entitiled SNPwatch: Researchers Investigate Shared Genetic Factors for Autoimmune Diseases.  Their piece not only discusses specific genetic variants shared among different autoimmune diseases, but also provides links to respective 23andMe pages of RA risk alleles that they test customers for (see respective 23andMe Table at above link).  This is especially handy for e-Patients seeking to establish a comprehensive health record. 

Overlap among disparate autoimmune diseases of genetic variants implicated in disease pathogenesis, is often illustrated by various types of Venn diagrams.  An example involving RA, systemic lupus erythematosis (SLE), and systemic sclerosis (SC), is shown in a figure from Delgado-Vega et al., 2010, reproduced below.  Genes sequestered into portions of the overlapping circles include those shared by all three diseases, only two of the three diseases, or else are unique to the individual disease.  In the context of this particular guest post series, the specific genes involved are not as important as the concept that a rather limited collection of genes, some shared and others unique, dictate susceptibility to autoimmune diseases broadly, and to RA or other autoimmune diseases more specifically. 

Figure 1 of Delgado-Vega et al., 2010.  “Unique and shared genes among SLE, RA and SSC.”

Note that several of the shared genes, including HLA, PTPN22, STAT4, IRF5, TNFAIP3, TRAF1-C5, IL21, were described in Post 3.  Importantly, disease risk here is most generally and significantly affected by HLA.
Original reference: "Recent findings on genetics of systemic autoimmune diseases" in: Current Opinion in Immunology Vol. 22: 698-705. (Elsevier press) 

A priori, it is especially worth mentioning that there is nothing to rule out the possibility of a given patient having a “hybrid autoimmune disease” (e.g. “rhupus”, a hybrid version of RA and Lupus), with genetic and clinical characteristics common to more than one autoimmune disease.  Alternatively, it is likewise possible that a given patient may present a full set of symptoms typical of two or more distinct autoimmune diseases, suggesting he/she has both diseases simultaneously (i.e. comorbidity).  In Table 2 of Zhernakova et al., 2009 are shown comorbidities of greatest likelihood for each of eleven common inflammatory and autoimmune diseases.  For RA, top comorbidities reported in this study include asthma, type 1 diabetes, and autoimmune thyroid disease (e.g. Hashimoto’s thyroiditis).  Unsurprisingly, not all autoimmune diseases are equally comorbid among themselves.  Butte and colleagues, 2009 showed, for example, that allelic variants that increase risk susceptibility for one particular autoimmune disease may be protective of susceptibility to another autoimmune disease.  Thus, for example, RA is far more likely to be comorbid with ankylosing spondylitis than multiple sclerosis.   Daly and colleagues, 2011 came to similar conclusions with different data sets. 

How are the various possibilities involving comorbidity or hybrid disease characteristics to be reconciled at the level of an individual patient presenting also with RA-like autoimmune symptoms?  The most straightforward mechanism, in keeping with the present theme, is utilizing genetic/genomic methods to facilitate a diagnosis.  While the genetic architecture of a particular patient’s autoimmune disease(s) may not hold all of the cards for a proper diagnosis (minimally due to environmental variables), they at least provide a foundation for making first approximation predictions of a patient’s likely autoimmune profile.  In many instances, it’s easier to imagine that such information will improve, perhaps significantly, the management of the patient’s disease. 

Based on the foregoing discussion, there is no reason to think, a priori, that individuals diagnosed with RA would comprise a homogeneous disease cluster.  Rather, it seems more reasonable to imagine that RA patients comprise a continuum of phenotypes having corresponding genotypes that contribute to considerable heterogeneity in disease manifestations.  Thus, as mentioned earlier, it does little good to classify RA as a “syndrome” rather than a disease, since this just stifles attempts to dissect the complexity of the real situation.  That said, it should not be surprising that diagnosis and management of patients with RA or other autoimmune diseases, is so difficult for doctors, including rheumatologists, to administer.  Consequently, this accounts for patients contacting social web sites like rawarrior.com, when they become disenchanted with their health care practitioners and seem to have no place else to turn to find truth and meaning in their disease.  

In summary, proper treatment of patients with RA and/or other autoimmune illnesses requires a personalized approach, best managed using genomic and other presently-available technologies.  As the complexity of human biology continues to become unraveled, further gains in understanding of the etiology, pathogenesis, and treatment options of individual patients will be realized.  Additional complexities, and future directions of disease understanding and management, will be discussed in Part 5 of this blog post series.  Meanwhile, it behooves every physician and each and every patient to lobby for full implementation of the technology currently available to manage human illness in ways presently achievable. 

Postscript: it is important to realize that the above discourse is a working hypothesis that will experience refinement with time.  That said, this author, as a non-clinician, is tempted to recommend that current RA patients (or those at risk of RA or another autoimmune disease) discuss this information with their doctor.  Unfortunately, many if not most doctors in practice today may not understand or appreciate the significance of such information (for more on this, see The Creative Destruction of Medicine, by Eric Topol), perhaps bringing patients full circle back to relying on information in social media for personal sustenance.