Because Whipple disease is uncommon, formal prospective evaluations of therapeutic regimens have not been done. Empirical antibiotic therapy with trimethoprim-sulfamethoxazole, cephalosporins, penicillins, or tetracyclines has varied in duration and outcome. In a case series of 29 patients with Whipple disease, the duration of treatment ranged from 5 days to 4 years [2]. Alleviation of symptoms and correction of biochemical and pathologic abnormalities were not related to the duration of antimicrobial therapy. A review of antibiotic therapy in 88 patients with Whipple disease who had long-term follow-up suggested that clinical relapse occurs in as many as 35% of cases [7, 8]; the outcomes of treatment for relapse of disease at the central nervous system were particularly poor [2, 8].

No test for cure is available for Whipple disease, but the recent identification of the 16S ribosomal RNA (rRNA) of the Whipple-associated bacillus, Tropheryma whippelii, has led to the use of 16S rDNA primers for detection of this organism on polymerase chain reaction (PCR) [9, 10]. Polymerase chain reaction assays with high sensitivities have shown T. whippelii in tissues that show no evidence of disease [9].

In this study, a PCR-based test for the detection of T. whippelii was evaluated. Our results suggest that PCR may be useful for the diagnosis of Whipple disease and for monitoring patients who have been treated with antibiotics, thereby allowing the physician to better direct the duration of antibiotic therapy and, perhaps, predict clinical relapse.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Thirty patients who received a diagnosis of Whipple disease at the Mayo Clinic between 1952 and 1994 were identified through a computerized search of a diagnosis database. Medical charts of these patients were reviewed. Characteristics of 29 of the patients have been reported elsewhere [2].

Tissue Samples

Tissue samples embedded in paraffin blocks were retrieved from the archives of the Mayo Clinic tissue registry; one to seven samples were available from each patient. The anatomic origin of the samples included the small intestine, rectum, pancreas, liver, brain, spleen, and lymph nodes. Post-treatment biopsy specimens were available from 17 of the 30 patients. The medical records of these 17 patients were reviewed by a gastroenterologist who was blinded to the results of PCR and histologic review. All histologic sections were rereviewed in a blinded manner by one pathologist. In small-intestine specimens, the diagnosis of Whipple disease was confirmed on the basis of the following criteria: 1) presence of macrophages in the lamina propria that contained periodic acid-Schiff staining granules, 2) resistance of periodic acid-Schiff-positive granules to diastase, 3) distortion of villous architecture due to the expanded, infiltrated lamina propria, and 4) dilated lymphatic channels (in some instances). Tissue specimens from regions other than the small intestine were also tested for evidence of Whipple disease. Histologic criteria were the presence of foamy histiocytes with bacilli that were positive on periodic acid-Schiff staining and were resistant to diastase and, in tissues from lymph nodes, dilated sinusoids with a "Swiss cheese" appearance. A histologic section of the small bowel from 1 patient that had been read elsewhere was reexamined in the Mayo Clinic's Department of Pathology to confirm the diagnosis; a whole-blood sample (500 micro L) was obtained from this patient after a week of therapy with trimethoprim-sulfamethoxazole.

In addition to the 30 patients who had confirmed cases of Whipple disease, 8 patients who had presented with gastrointestinal symptoms and had had biopsy of the small intestine were identified by using a medical records database. Whipple disease had been considered in the differential diagnosis for all 8 patients before biopsy of the small intestine was done. Biopsy specimens were classified as suspicious for or negative for Whipple disease. Medical charts on 7 patients were available for review.

Paraffin-embedded biopsy specimens from the small intestine, colon, and brain of 42 patients who had malabsorption of undetermined cause, the irritable bowel syndrome, or viral encephalitis were used as controls. No histologic evidence of Whipple disease was found in these tissues, and all were negative for periodic acid-Schiff-positive macrophages. The control specimens were processed, interspersed with, and tested in a blinded manner in parallel with specimens from the 37 patients described above.

Preparation and Analysis of Specimens

We extracted two sections of formalin-fixed, paraffin-embedded tissue, each 25 µm thick, that had been obtained from each patient. We used a new section of the knife for each cut to avoid cross-contamination during processing. Sections were deparaffinized with xylene and digested overnight at 55 °C with 50 µL of proteinase K (20 mg/mL) and sodium dodecyl sulfate (10%) in Tris (10 mmol/L) and EDTA (1 mmol/L). The digested tissue was boiled for 15 minutes to inactivate the proteinase K. Deoxyribonucleic acid was extracted by using a chaotropic lysis method with 200 µL of the tissue (Isoquick kit, Orca Research, Bothell, Washington).

Design of Primers

Primers W3FE and W4RB were modified to improve their specificity for T. whippelii [11]. Primers W3AF and W4AR were designed on the basis of 16S rRNA gene sequence alignments of related organisms so that the region of gene that was amplified would be unique to T. whippelii. The expected size of the W3AF and W4AR amplification product was 160 base pairs.

Evaluation of Specificity of Primers W3AF and W4AR

We determined the specificity of primers W3AF and W4AR by doing PCR on DNA that had been extracted from clinical isolates of many bacteria: Nocardia brasiliensis, N. otitidiscaviarum, N. asteroides, Mycobacterium tuberculosis, M. kansasii, M. chelonae, M. bovis, M. fortuitum, M. avium-intracellulare complex, Actinomyces bovis, A. viscosus, anaerobic Actinomyces species, Corynebacterium species, Propionibacterium species, Rhodococcus equi, unspeciated Rhodococcus, and Streptomyces griseus. Amplification products were found only in R. equi and N. otitidiscaviarum. Sequencing of amplification products of these two organisms showed four variances in nucleotide bases compared with the T. whippelii sequence: C to T at position 1092, G to A at position 1093, C to T at position 1096, and G to A at position 1098. To ensure specific detection of T. whippelii, an 18-mer oligonucleotide probe homologous to T. whippelii sequence was designed to encompass the four-nucleotide variance (see Appendix Table Table 5).

Primers W3FE and W2RB, which are also specific for T. whippelii, were used in accordance with the methods described elsewhere [9, 12].

Extraction Control

We used the ß-globin gene, which is found in all nucleated cells, to assess the efficiency of the extraction of DNA from tissue. Tissue samples that were negative for Whipple-associated bacillus 16S rDNA products were amplified for a segment of the human ß-globin gene. Samples that did not have a positive result when tested for ß-globin were considered inadequate. Primers GH20 and PCO4 were used for ß-globin amplification [13].

Polymerase Chain Reaction with Primers W3AF and W4AR

An amplification reaction mix, 45 µL, was prepared from 10 µL of 50% glycerol (10% final concentration), 5 µL of 10 x Perkin-Elmer buffer (containing 100 mmol of Tris-HCl per L [pH, 8.3] [Sigma, St. Louis, Missouri], 500 mmol of KCl per L, and 15 mmol of MgCl₂ per L) (Applied Biosystems, Perkin-Elmer, Foster City, California), 2 µL each of the four deoxyribonucleoside triphosphates, 0.25 units of AmpliTaq polymerase (Applied Biosystems, Perkin-Elmer), 5 pmol of primers W3AF and W4AR, 0.33 µL of isopsoralen (25 µg/mL), and 19.42 µL of sterile water. We added 10% or 5 µL of extracted DNA from the specimen to this master mix for a final reaction volume of 50 µL.

Thermocycling was done as follows. Initial melting was done at 94 °C for 2 minutes. This was followed by 50 cycles of denaturing at 94 °C for 30 seconds, annealing at 60 °C for 45 seconds, and extension at 72 °C for 60 seconds. Final extension was done at 72 °C for 4 minutes. After each amplification, the microtubes were exposed to ultraviolet light to promote covalent crosslinking of the isopsoralen [14]. This process prevents PCR product contamination, which can lead to a false-positive result on PCR. Amplification products were detected by performing gel electrophoresis on 20% of the reaction volume in a 3% agarose gel that contained ethidium bromide. The amplification products were transferred to a nylon membrane that was then hybridized with a digoxigenin-labeled DNA oligonucleotide probe, followed by autoradiographic detection.

We proved the efficiency of isopsoralen for inactivating amplified DNA by doing the following experiment. We used two sets of 10-fold serial dilutions of amplified template ( 10^–15 dilution), each containing isopsoralen, and irradiated one of the dilutions with ultraviolet light. A PCR product was detected in the amplicon dilution set not treated with ultraviolet light even when the product amplicons were diluted by as much as 10^–12; in the sample treated with ultraviolet light, no product was found at a dilution less than 10^–1 [15].

16S Ribosomal DNA Sequence Analysis of Selected Samples

Seven specimens were analyzed by PCR and DNA sequence determination at a separate laboratory. Six specimens came from patients who had positive histologic evidence of Whipple disease; one specimen came from one of the seven patients who had clinically suspected Whipple disease. Samples were digested, subjected to DNA extraction, and amplified by using primers W2RB and W3FE. These primers amplify an intervening 230-base-pair region of the 16S rRNA gene. Negative controls were digestion buffer without tissue and paraffin-embedded tissue from patients who tested negative for Whipple disease. The latter set of controls was scraped from glass slides and run in parallel with tissues from patients with Whipple disease.

Sequencing

Products from reactions with primers W3FE and W2RB were purified in a Centricon-100 column (Amicon, Beverly, Massachusetts) and run through a DyeDeoxy (Applied Biosystems, Perkin-Elmer) terminator cycle sequencing protocol using one of five primers (W2R, W3F, W4R, 1175R, or 1175F). Both strands of each reaction product were sequenced. Sequence fragments were aligned by using AutoAssembler software (Applied Biosystems, Perkin-Elmer) to create a consensus sequence. The consensus sequence was analyzed by using the University of Wisconsin Genetics Computer Group software suite (Madison, Wisconsin) [16].

Results

Top Methods Results Discussion Author & Article Info References

Tropheryma whippelii-specific PCR was done on specimens obtained from 30 patients with histologically confirmed Whipple disease. As in other studies [17], larger amplification targets were more difficult to detect in the fixed tissues from these patients. Test results were positive in 29 of 30 specimens for which primers W3AF and W4AR (160 base pairs) were used (sensitivity, 96.6%; specificity, 100%) and in 16 of 28 specimens for which primers W3FE and W2RB (284 base pairs) were used (Table 1). We also tested specimens from 8 patients in whom the clinical differential diagnosis included Whipple disease without histologic confirmation after review of several biopsy specimens. Specimens from 7 of these patients had positive results when primers W3AF and W4AR were used; 4 samples had positive results when primers W3FE and W2RB were used. Negative control samples that were processed in the same manner as the test samples did not show any amplification with either of these primer sets (Table 1). Twelve patients from whom multiple specimens were obtained also had at least 1 specimen in which the histologic findings were typical of Whipple disease; all of these patients had a PCR positive specimen that had been obtained from a site with negative histologic findings (Table 2). Only 1 of the 12 patients had a sigmoid biopsy specimen that was negative on PCR and yielded negative histologic findings. Overall, these results seem to be consistent with the systemic nature of Whipple disease.

Antibiotic Treatment

Twenty-five of the 30 patients who had histologically confirmed Whipple disease were treated with antibiotics. Post-treatment biopsy specimens from the small bowel and, in some cases, from the rectum were obtained from 15 of 25 patients. We also did PCR on pre- and post-treatment specimens obtained from 2 additional patients who had a clinical diagnosis of Whipple disease but negative histologic findings. These patients had received antibiotics for presumptive disease despite negative test results. In this group of 17 patients, 13 were male and the median age was 54 years (range, 34 to 70 years). Patients had had symptoms for a median of 96 months before diagnosis (range, 4 to 300 months). Sixteen of the 17 patients were treated with tetracycline, alone or in combination with penicillin, streptomycin, and erythromycin; 1 patient received a 3-week course of penicillin and streptomycin (median duration of treatment, 12 months [range, 6 days to 35 months]). Patients were followed for a median of 119 months (range, 3 to 308 months) after their first Mayo Clinic visit.

Results of PCR testing, histologic findings of all pre- and post-treatment specimens, and duration of treatment for all patients are summarized in Table 3. Before antibiotic treatment was initiated, results of PCR were positive in all patients with Whipple disease for whom primers W3AF and W4AR were used and in 9 of 17 patients for whom primers W3FE and W2RB were used. Fifteen of 17 patients (88.2%) had a good initial clinical response to antibiotics; however, 1 patient died before post-treatment biopsy could be done. Five patients (29.4%) had clinical relapse after treatment; relapse occurred a median of 28 months (range, 7 to 272 months) after diagnosis. Results of PCR were positive in post-treatment specimens from 12 of 17 patients for whom primers W3AF and W4AR were used and in specimens from 3 of 17 patients for whom primers W3FE and W2RB were used. Seven of the 12 patients who had a positive result on PCR (positive predictive value, 58% [95% CI, 28% to 85%]) had one or more clinical relapses or did not respond to antibiotic treatment. In contrast, none of the 5 patients whose post-treatment samples were negative on PCR had relapse (negative predictive value, 100% [CI, 48% to 100%]; P = 0.044, Fisher exact test.).

The correlation of histologic results to treatment outcomes was less clear. Post-treatment specimens from 6 of 17 patients had positive histologic results. Two of these 6 patients had no clinical response or had relapse, whereas 4 patients had a good clinical response. Of the 11 patients whose post-treatment samples were histologically negative, 5 patients responded well, 1 died of hepatoma, 1 did not respond, and 4 had clinical relapse. Thus, no correlation was observed between histologic findings in post-treatment specimens and clinical outcome (P > 0.2, Fisher exact test). Six of 17 patients died during follow-up; two deaths were directly related to progressive Whipple disease.

Sequence Analysis of Whipple-Associated Bacillus 16S Ribosomal DNA Products

Specimens from seven patients were analyzed for the T. whippelii 16S rDNA sequence. Six patients had clinical and histologic manifestations of Whipple disease; one patient had suspicious but nonconfirmatory histologic findings. Samples from these seven patients had positive results on PCR when primers W3FE and W4RB were used in a second laboratory; use of these primers produced a 154-base pair amplification product [11]. Five of 7 samples tested using primers W2RB and W3FE, which generated a 284-base pair product, also had positive results. Sequencing was done on the 284-base-pair products amplified from these five patients. In three patients, the sequence was completely homologous with the previously published T. whippelii 16S rDNA sequence (GenBank M87484). A single base-pair mismatch was found in one patient (A for T at position 990), and two single base-pair mismatches were found in a patient whose histologic findings were suspicious for Whipple disease (A for T at position 990; T for A at position 1147).

Discussion

Top Methods Results Discussion Author & Article Info References

In this retrospective study, we used highly sensitive PCR-based methods to detect T. whippelii in formalin-fixed, paraffin-embedded biopsy specimens taken from the small intestines and lymph nodes of 29 patients in whom Whipple disease was diagnosed clinically and histologically. In addition, we showed the presence of this pathogen in clinical specimens from patients in whom Whipple disease was considered in the differential diagnosis but could not be confirmed by histologic testing. Our study confirms that PCR is helpful in identifying Whipple-associated bacillus in tissues with no direct manifestation of disease and underscores further the occult systemic distribution of the implicated organism [11]. Post-treatment biopsies of the small bowel were done for 17 patients; biopsy results showed persistence of detectable T. whippelii DNA in 12 of these patients. Seven of the 12 patients either had clinical relapse or never responded to treatment. In contrast, the remaining 5 patients who had undetectable DNA on follow-up biopsy did not have clinical relapse (P = 0.044). We found no correlation between post-treatment histologic results and clinical outcome (P > 0.2).

Whipple disease is often considered in the differential diagnosis of patients with signs and symptoms of gastrointestinal malabsorption, with or without central nervous system involvement. However, many cases of this disease primarily involve the central nervous system or other organ systems without antecedent gastrointestinal symptoms: Thus, the disease might be mistaken for other chronic infectious or inflammatory diseases, such as rheumatoid arthritis or sarcoidosis. Indeed, most cases of Whipple disease in our series were recognized only after the infection caused clinical manifestations in the small intestine that led to biopsy and histologic diagnosis; others were diagnosed only after autopsy [2]. Thus, the early diagnosis of Whipple disease is problematic, in large part because of its nonspecific presentation.

The availability of the DNA sequence of the 16S rRNA gene from T. whippelii has facilitated the development of a relatively specific and sensitive diagnostic test for Whipple disease [9, 11, 18, 19]. Although the 42 blindly tested negative control samples were indeed negative in our study (which would argue against substantial cross-contamination of PCR products), nucleic acids specific to T. whippelii were found in 7 samples from patients for whom a diagnosis of Whipple disease was strongly considered despite the absence of convincing histologic evidence. In 2 of these patients, empirical treatment with doxycycline for possible Whipple disease was successful. Table 4 briefly describes the clinical presentations of six of these patients.

We used the primers W3FE and W2RB (which are specific to Whipple-associated bacillus) to detect Whipple-associated bacillus in paraffin-embedded specimens and compared the sensitivity of these primers with that of primers W3AF and W4AR [9]. When compared with histologic testing, the sensitivity for detecting the bacillus was 59% with primers W3FE and W2RB and 96.6% with primers W3AF and W4AR. This is in agreement with a recent study in which the primers W3FE and W2RB did not detect T. whippelii in a sample of infected vitreous [11]. A new, more sensitive antisense primer for detection of a 154-base-pair product was then designed; primers W3AF and W4AR were designed in the present study to increase the sensitivity of detection of T. whippelii, especially in fixed, paraffin-embedded material.

A negative result on PCR in one patient whose condition was histologically confirmed might be explained by the presence of a different organism with a histologic appearance similar to that of T. whippelii, degradation of target DNA, or excessive fixation of the tissue, all of which may lead to lower PCR sensitivity. Testing fresh biopsy material or material fixed in ethanol may help to avoid some of these problems.

In addition to its role in the diagnosis of Whipple disease, PCR may be useful for monitoring the course of patients with this infection. Long-term treatment with such antibiotics as penicillin and trimethoprim-sulfamethoxazole fails to prevent recurrence of disease in many patients; some relapses have been fatal [21]. Histologic findings seemed to improve overall after long-term treatment was initiated, but this did not predict complete recovery [22, 23]. However, the use of PCR as a test for cure has some potential limitations. At present, we cannot determine whether a positive result on a PCR test done after treatment is caused by persistence of infection or by the persistence of DNA from dead bacilli in the specimen. Cultivation of Whipple bacillus on artificial media would be the best test for proving the persistence of this pathogen; however, because an in vitro or in vivo cultivation system does not yet exist, methods based on nucleic acid amplification are the only tools available for direct detection. Polymerase chain reaction may be useful for follow-up testing with the understanding that this test has a low positive predictive value. Nevertheless, loss of detectable DNA was observed in 5 of 17 patients from whom biopsy specimens were available after treatment. The first course of antibiotic therapy in these patients lasted 6 days to 35 months. During a median follow-up period of 156 months (range, 82 to 208 months), none of the 5 PCR-negative patients had relapse. Thus, the negative predictive value of PCR seemed high, although this observation was based on a small sample. A delay in the clearance of organism-specific DNA after clinically efficacious antimicrobial therapy is also a feature of mycobacterial infection [24] and, perhaps, Lyme disease [25].

We conclude that PCR may be valuable for the diagnosis of Whipple disease, especially when the diagnosis cannot be confirmed by histologic testing. Judicious use of a PCR-based test may help to identify infections with T. whippelii more definitively, thus leading to better-informed management decisions. It may also be possible to detect infection before histologic findings become abnormal and to differentiate this infection from others that have a similar histologic picture, such as infections with M. avium complex [20]. Polymerase chain reaction may also have a role, albeit a more limited one, in monitoring the eradication or persistence of infection and thereby providing a more rational approach to such treatment decisions as antibiotic selection and duration of therapy.

From the Mayo Clinic and Mayo Graduate School of Medicine, Rochester, Minnesota; State University of New York at Syracuse, Syracuse, New York; Stanford University Medical Center, Stanford, California; and the Veterans Affairs Palo Alto Health Care System, Palo Alto, California.

Dr. Rooney: Department of Anatomic Pathology, State University of New York at Syracuse, Health Science Center, 750 East Adams Street, Syracuse, NY 13210.

Drs. Relman and Fredricks: Department of Veterans Affairs Medical Center, 3801 Miranda Avenue, Palo Alto, CA 94304.