© Bill McAllen Photography

For me, the dream began in the early 1960s. As a medical student, I was helping with the care of a 16-year-old girl who had classical Hashimoto thyroiditis, a disease that more commonly occurs among middle-aged women. I guessed that this girl's early onset signified a particularly strong genetic predisposition. Indeed, further investigation identified numerous cases of thyroid disease or other endocrine-system problems in family members. It occurred to me that genetic information might predict future development of autoimmune disease and, thus, open avenues to intervene before the pathologic process reached an irreversible level.

The opportunity to pursue this idea arose 15 years later at Wayne State University School of Medicine in Detroit. There, I worked with William H. Hoffman, a pediatric endocrinologist who had an active clinic of children and adolescents with autoimmune thyroid disease, and Lynne Burek, a clinical laboratory immunologist. Later, Avi Gimelfarb, a population geneticist, joined the group.

We studied children who had either chronic lymphocytic thyroiditis or Graves thyrotoxicosis, and we studied their siblings and parents as well. If both parents had thyroid autoantibodies, our primary measurement of thyroid autoimmunity, then a child had a 71% chance of also having thyroid autoantibodies. That percentage dropped to 54 for a child from a family in which only one parent had thyroid autoimmunity. If neither parent showed evidence of thyroid autoantibodies, the child had a 29% chance of developing thyroid autoantibodies. We assumed that environmental factors operated equally among these three groups, so we attributed these differences to genetics.

At that time, we had predicted that human leukocyte antigen (HLA) might indicate a person's susceptibility to autoimmune thyroid disease. We confirmed that and more.

Our work revealed that neither a single HLA haplotype nor linked groups of gene variations predominate in people who are susceptible to thyroid autoimmunity. Still, we found genetic connections. In comparison with a child being treated for thyroid autoimmunity, 89% of the siblings with the same two HLA haplotypes also showed evidence of thyroid autoimmunity. Interestingly, 32% of those siblings had biochemical or physical evidence of thyroid dysfunction. Of the siblings who shared one haplotype with the child being treated, 69% had thyroid autoantibodies, and only one had subclinical thyroid disease. If siblings shared no HLA haplotype with the patient, only 56% had thyroid autoantibodies, and none had any evidence of thyroid dysfunction.

From that work, we drew three general conclusions. First, HLA is an important determinant in thyroiditis. Second, the particular HLA haplotype conferring susceptibility differs from family to family. And last, our findings suggested that different individuals recognize different antigenic determinants on the thyroglobulin molecule.

The risk, though, depended on more than HLA haplotypes. Age and sex were also contributing risk factors. Before puberty in children with thyroid autoimmunity, the ratio of males to females was nearly equal. After puberty, two-thirds of the patients with thyroiditis were female.

In the end, thyroid autoantibodies themselves emerged as the most important risk factors. All the patients with thyroiditis had thyroid autoantibodies, and so did 50% of their siblings. By contrast, only 10% of the control children showed thyroid autoantibodies. In almost all cases of thyroiditis (75%) or Graves disease (89%), we found autoantibodies to thyroglobulin and thyroperoxidase, but a few individuals had antibodies to one or the other. Among the siblings, finding both autoantibodies signified a greater risk of subclinical disease than either antibody alone. The presence of both thyroid-specific antibodies may be taken as an early warning sign of impending disease.

Those studies showed us that a number of genetic traits might predict later thyroid disease. Each gene contributes incrementally; the aggregation of more genes leads to proportionately greater risk.

The endeavor to predict autoimmune disease before its clinical appearance should be linked to a program to prevent or abort the disease before irreversible destruction of the target organ takes place. In humans, we have a few instances of interventions that successfully prevent progression of disease, based on identifying the requisite environmental trigger. For example, we can prevent rheumatic heart disease by perennial administration of antibiotics to prevent repeated streptococcal infections in children who have had rheumatic fever. As another example, avoiding the glycoprotein gliadin (through a gluten-free diet) prevents the clinical manifestations of celiac disease. Unfortunately, we do not have good examples of interventions that terminate on-going autoimmune responses in humans. For example, efforts to treat diseases such as rheumatoid arthritis, multiple sclerosis, and diabetes by oral administration of the putative antigen have not been effective.

Still, providing the proper antigen before an autoimmune response starts could be therapeutic. For example, using Buffalo (BUF) strain rats, which spontaneously develop a form of autoimmune thyroiditis, we showed that injecting young animals with thyroid antigen completely prevents the disease. If we injected the thyroid antigen after the onset of thyroiditis, however, the disease was unchanged or even intensified.

It is clear that we are only at the beginning of efforts to develop methods for preventing autoimmune disease, especially in highly susceptible individuals. Strategies may involve isolating environmental triggers or identifying the appropriate instigating antigen. Preventing rather than treating autoimmune disease is certainly a worthy goal.

We know so much more about the basis of autoimmune diseases than I did when I started this journey four decades ago. In many areas, predictive medicine is already complementing curative medicine. Using biomarkers based on the genome and proteome, investigators can identify unusually susceptible individuals or populations, facilitate prognosis, forecast the outcome of therapy in clinical trials, and aid in developing improved treatment or preventive measures. Applying this approach to treatment of autoimmune disease requires two steps, both within our grasp. First we must predict with reasonable assurance those who will later develop overt disease; second, we need effective interventions sufficiently benign to ethically apply to apparently healthy individuals.

Noel R. Rose, MD, PhD, is professor of pathology and of molecular microbiology and immunology at Johns Hopkins University. He is also director of the Johns Hopkins Center for Autoimmune Disease Research.