Skip to main content
Device Global Header

Subscribe
or Renew

Advanced Search

Original Article

The Agent of Bacillary Angiomatosis — An Approach to the Identification of Uncultured Pathogens

List of authors.
  • David A. Relman, M.D.,
  • Jeffery S. Loutit, M.B., Ch.B.,
  • Thomas M. Schmidt, Ph.D.,
  • Stanley Falkow, Ph.D.,
  • and Lucy S. Tompkins, M.D., Ph.D.

Abstract

Background.

Bacillary angiomatosis is an infectious disease causing proliferation of small blood vessels in the skin and visceral organs of patients with human immunodeficiency virus infection and other immunocompromised hosts. The agent is often visualized in tissue sections of lesions with Warthin—Starry staining, but the bacillus has not been successfully cultured or identified. This bacillus may also cause cat scratch disease.

Methods.

In attempting to identify this organism, we used the polymerase chain reaction. We used oligonucleotide primers complementary to the 16S ribosomal RNA genes of eubacteria to amplify 16S ribosomal gene fragments directly from tissue samples of bacillary angiomatosis. The DNA sequence of these fragments was determined and analyzed for phylogenetic relatedness to other known organisms. Normal tissues were studied in parallel.

Results.

Tissue from three unrelated patients with bacillary angiomatosis yielded a unique 16S gene sequence. A sequence obtained from a fourth patient with bacillary angiomatosis differed from the sequence found in the other three patients at only 4 of 241 base positions. No related 16S gene fragment was detected in the normal tissues. These 16S sequences associated with bacillary angiomatosis belong to a previously uncharacterized microorganism, most closely related to Rochalimaea quintana.

Conclusions.

The cause of bacillary angiomatosis is a previously uncharacterized rickettsia-like organism, closely related to R. quintana. This method for the identification of an uncultured pathogen may be applicable to other infectious diseases of unknown cause. (N Engl J Med 1990; 323:1573–80.)

Introduction

BACILLARY angiomatosis (also called epithelioid angiomatosis) and cat scratch disease are important clinical disorders that are presumed to be infectious on the basis of direct visualization of microorganisms in diseased tissues. Because these organisms cannot be grown reproducibly, the identity of the presumed pathogen or pathogens remains unknown.

Bacillary angiomatosis is a distinct vascular proliferative disease initially described in the skin and lymph nodes of patients seropositive for the human immunodeficiency virus (HIV).1 , 2 Lesions from this disorder have been found to contain clusters of bacilli with positive results on Warthin—Starry staining that resemble those found in cat scratch disease,3 , 4 although the histologic appearance of the disease differs from that of classic cat scratch disease.5 Bacillary angiomatosis is known to cause disseminated visceral disease in HIV-seropositive patients and other immunocompromised hosts,6 7 8 and it has recently been found to do so in an immunocompetent host.9 Cat scratch disease is usually a benign granulomatous disease involving skin and regional lymph nodes10 , 11; however, disseminated disease has been described that involves visceral organs, particularly in immunocompromised hosts.12 13 14 The causative agent of cat scratch disease was first visualized in 1983 with the Warthin—Starry silver stain15 but has resisted identification despite reports of isolation in culture.16 Distinguishing between disseminated bacillary angiomatosis and cat scratch disease in the immunocompromised host is difficult17; the ability to identify the causative bacteria would help resolve this difficulty as well as improve diagnosis and treatment and perhaps lead to a better understanding of pathogenesis.

Two recent experimental developments — analysis of 16S ribosomal RNA (rRNA)18 , 19 and the polymerase chain reaction (PCR)20 21 22 — have allowed a novel approach to the identification of disease-associated, uncultured microorganisms. All cells contain multiple copies of a gene encoding a 16S or 16S-like small subunit of rRNA, in which hypervariable sequences are interspersed with regions of highly conserved sequence.19 The analysis of the variable portions permits the determination of phylogenetic and evolutionary relations among organisms and now forms the basis of a revised system of natural classification.19 Sequences conserved throughout each or all three kingdoms (eukaryotes, archaeobacteria, and eubacteria) can serve as targets for primer-directed DNA amplification or hybridization.23

The PCR can be used to amplify a 16S-like eukaryotic rRNA sequence from purified genomic DNA.24 It permits rapid amplification of any DNA region of interest if the sequences flanking the segment of interest are known, so that two amplification primers can be designed, and if there is an appropriate pool of DNA containing the target region.21 , 25 Therefore, one might be able to identify an unknown eubacterium with primer-directed, amplified variable regions of its 16S rRNA sequence by using primers that anneal to flanking sequences conserved within the kingdom of eubacteria.26 Although pure cultured organisms have usually been the source of the 16S ribosomal DNA (rDNA) target in PCR, infected human tissue could also be used.

A recent case of disseminated bacillary angiomatosis at our institution8 provided us with the appropriate target DNA needed for the execution of this strategy. Using broad-range eubacterial 16S rDNA primers, we amplified, cloned, and sequenced a portion of eubacterial 16S rRNA gene directly from tissue infected with the presumed etiologic agent. Subsequent analysis of this sequence and study of infected tissues from other patients suggest that bacillary angiomatosis is caused by a rickettsia-like organism that is most closely related to Rochalimaea quintana.

Methods

Patients

Figure 1. Figure 1. Splenic Nodule from Patient 1 after Warthin—Starry Staining.

In Panel A, inflammatory cells, focal areas of necrosis, and numerous epithelioid endothelial cells forming vascular channels can be seen (×670). In Panel B, at a higher magnification (×3300), pleomorphic bacillary organisms are apparent and often are in clumps (arrows). Similar organisms were seen in samples of splenic hilar lymph node and liver from the same patient.

Patient 1 was a 38-year-old woman in whom fever, chills, headache, myalgias, nausea, and vomiting developed five years after cardiac transplantation while she was following a maintenance regimen of immunosuppressive drugs.8 Three months earlier she had required surgical repair of a cat scratch. Abdominal CT scanning revealed hepatosplenomegaly with multiple areas of low attenuation and surrounding enhancement. Exploratory laparotomy and splenectomy were performed. Histologic examination of liver, spleen, and splenic hilar lymph node demonstrated a proliferation of histiocytoid endothelial cells forming vascular channels. Numerous bacillary organisms were seen on Warthin—Starry staining and electron microscopy (Fig. 1). Routine bacterial cultures were negative. A diagnosis of bacillary angiomatosis was made. The patient was treated with an eight-week course of erythromycin, with complete resolution of her symptoms.

Patient 2 was a 24-year-old HIV-seropositive man who had a two-month history of fever, chills, sweating, abdominal pain, anorexia, and weight loss.8 He had had several cat scratches in the previous three months. Physical examination revealed a 1.0-cm purplish nodule on his left elbow, left-axillary lymphadenopathy, and hepatosplenomegaly. A skin biopsy, bone marrow aspiration and biopsy, and core liver biopsy were performed. Pathological examination of the skin was suggestive of pyogenic granuloma; however, Warthin—Starry staining revealed bacillary organisms. Sections of bone marrow showed focal areas of epithelioid angiomatosis, but there was insufficient material for Warthin—Starry staining. Histologic findings in the liver were consistent with epithelioid angiomatosis, with Warthin—Starry staining positive for bacillary organisms. All cultures were negative. Treatment with doxycycline led to prompt clinical improvement.

Patient 3 was a 54-year-old woman infected with HIV. A biopsy of a cutaneous heel lesion was performed, and Warthin—Starry staining demonstrated clumps of bacilli. Histologic findings suggested a diagnosis of bacillary angiomatosis. (This case has been described elsewhere by LeBoit et al. [Case 8].5)

Patient 4 was a 29-year-old woman with the acquired immunodeficiency syndrome in whom diffuse lymphadenopathy developed. She had no skin lesions. Biopsy of an inguinal lymph node showed histologic features consistent with epithelioid angiomatosis. Clusters of bacilli were seen on Warthin—Starry staining.

DNA Extraction

DNA was extracted from a frozen splenic nodule and a frozen piece of splenic hilar lymph node from Patient 1, a frozen sample of bone marrow from Patient 2, paraffin-embedded, formalin-fixed tissue from a skin lesion from Patient 3, and a paraffin-embedded, formalin-fixed lymph node from Patient 4. Each of these patients had the biopsy performed at a different hospital. As a control for DNA carryover contamination, for every biopsy sample from the patients we prepared DNA extracts from at least one of the following tissues: frozen spleen from a patient with idiopathic thrombocytopenic purpura, with no clinical or cultural evidence of infection; a paraffin-embedded, formalin-fixed spleen sample obtained by splenectomy in a trauma victim; and frozen tonsil from two patients with recurrent tonsillitis.

Frozen Tissue

The methods used were modified from those described previously.27 , 28 Approximately 100 mg of each tissue was rinsed in sterile water, minced, and placed in 5 ml of digestion buffer (500 mM TRIS, pH 9; 20 mM EDTA; 10 mM sodium chloride; and 1 percent sodium dodecyl sulfate) with fresh proteinase K (Sigma, St. Louis) at a final concentration of 1 mg per milliliter. The samples were incubated at 60°C for 48 to 72 hours. Ribonuclease A (Sigma) was added to a final concentration of 4 μg per milliliter; the samples were incubated at room temperature for two hours and then extracted three times with equal volumes of TRIS-buffered phenol and chloroform—isoamyl alcohol (24:1). The DNA was precipitated with ethanol according to standard techniques29 and resuspended in 50 μl of sterile water.

Paraffin-Embedded Tissue

The method used was a modification of that published by Shibata et al.30 and described by Wright and Manos.31 Ten-micrometer sections from the various paraffin-embedded blocks of tissue were trimmed, treated with octane and ethanol, and then placed in 200 μl of digestion buffer as described.31 After digestion and heat inactivation of proteinase K, the reaction mixture was centrifuged and the supernatant was used directly in amplification reactions.

Oligonucleotide Primers

Table 1. Table 1. Oligonucleotide Primers Used for Amplification of 16S rDNA.*

The sequences of primers p11E and p13B were adapted with modifications from Chen et al.26 Primers p93E, p24E, and p12B were designed on the basis of the 16S rRNA sequence determined in this study (Table 1). Each included either an EcoRI or BamHI restriction-endonuclease site at the 5′ end to facilitate the cloning of the resultant PCR product. All primers were synthesized by Operon Technologies (Alameda, Calif.).

PCR Amplification

DNA extracts from frozen tissue were adjusted to a concentration of approximately 400 to 500 μg of total DNA per milliliter; 1 μl was used in a 100-μl reaction volume. Ten microliters (5 percent) of the paraffin-embedded DNA extract was sufficient for the amplification of human β-globin genes with primers PC04 and GH20 (Perkin–Elmer–Cetus, Norwalk, Conn.); the same amount was used for the amplification of eubacterial 16S rRNA. Each reaction contained 20 pmol of each primer and standard amounts of Gene-Amp reagents (Perkin–Elmer–Cetus), except that 2 mM magnesium chloride was used for amplification with extracts of paraffin-embedded tissue. This mixture was irradiated with ultraviolet light to reduce contamination of reagents. An overlay of sterile mineral oil was added to the tubes, followed by the DNA template. PCR cycles consisted of 1 minute of denaturation at 94°C, 1 minute of annealing at 55°C, and 1.5 minutes of extension at 72°C in a programmable thermal controller (MJ Research, Watertown, Mass.). For amplification with extracts of paraffin-embedded tissue, we used a step file in an automated DNA thermal cycler (Perkin–Elmer–Cetus). Negative controls were included during every amplification and consisted of reaction mixtures without the DNA template, as well as reaction mixtures with DNA extracts from nondiseased tissues. The product of PCR was detected by electrophoresing 10 μl of reaction solution in a 1.2 percent agarose gel containing ethidium bromide; a 1-kilobase (kb) DNA ladder (GIBCO BRL, Gaithersburg, Md.) was used as the DNA size standard on each gel.

Purification, Cloning, and Sequencing of the PCR Product

Half the PCR product was extracted with phenol and chloroform, precipitated with ethanol, and pelleted. The 3′ ends of DNA strands were filled in with DNA polymerase I large (Klenow) fragment.29 The product was purified by low-melting-point agarose-gel electrophoresis, cut with a combination of EcoRI and BamHI (New England Biolabs, Beverly, Mass.), and then ligated with pBluescript II KS phagemid vector (Stratagene, La Jolla, Calif.), which had been cut with EcoRI and BamHI.29 Escherichia coli DH5α (Bethesda Research Laboratories, Bethesda, Md.) was electro-transformed32 with the recombinant plasmids with a Bio-Rad gene pulser (Richmond, Calif.). After screening of the ampicillin-resistant transformants for the appropriate recombinant plasmid, both strands of the plasmid inserts were sequenced with double-stranded DNA templates (Sequenase; U.S. Biochemical, Cleveland) and the dideoxy chain-termination method.33

Data Analysis

Sequences were aligned manually on the basis of conserved regions and secondary structural elements. Regions of ambiguous alignment were omitted from the phylogenetic analysis. The evolutionary distance separating two sequences was computed from the percent similarities34; a least-squares method was used to infer the phylogenetic tree most consistent with the calculated distance.35 All reference 16S rRNA sequences were obtained from the GenBank (Los Alamos, N.M.)/European Molecular Biology Laboratory (Heidelberg, Germany) data bases.

Results

Amplification of 16S rRNA Sequence from Tissue Extract Using Broad-Range Eubacterial Primers

Frozen splenic tissue from Patient 1 was chosen as the initial target for 16S rRNA amplification because of the relatively large number of bacteria seen with Warthin—Starry staining (Fig. 1B). Eubacterial broad-range primers p22E and p13B were designed from 16S rDNA gene sequences known to be conserved in all eubacteria26 but not found in eukaryotic 16S-like genes. Previous studies have demonstrated that picogram amounts of E. coli 16S rDNA could be detected with the use of these primers as hybridization probes in the presence of a large excess of eukaryotic genomic DNA.26

Figure 2. Figure 2. Agarose-Gel Electrophoresis of PCR-Amplified 16S rDNA Fragments from Tissue DNA Extracts.

Broad-range eubacterial 16S rDNA primers — p11E and p13B — were used for the reactions in lanes 1 through 5. They prime the amplification of a 232-bp fragment. Specific primers — p24E and p12B — were used for the reactions corresponding to lanes 6 through 12; they generate a 296-bp fragment. Lanes 1 and 6 contain no added DNA extract; lanes 2 and 7, a spleen sample from a patient with idiopathic thrombocytopenic purpura; lanes 3 and 8, tonsillar tissue from a patient with recurrent tonsillitis; lanes 4 and 9, a spleen sample from Patient 1; lane 5, a sample of splenic hilar lymph node from Patient 1; lane 10, bone marrow from Patient 2; lane 11, tissue from a skin-lesion biopsy from Patient 3; and lane 12, lymph-node tissue from Patient 4. The size of some of the DNA standards is indicated in kilobases.

After 25 cycles of PCR amplification using p11E and p13B and splenic or splenic hilar lymph-node DNA from Patient 1 as the target, a product of approximately 200 base pairs (bp) was detected reproducibly with agarose-gel electrophoresis (Fig. 2, lanes 4 and 5). No DNA band could be amplified from the spleen sample from a patient with idiopathic thrombocytopenic purpura or in the absence of added tissue extract (Fig. 2, lanes 2 and 1, respectively). With DNA from "normal" human tonsillar tissue, a DNA species of similar size was sometimes barely seen although not easily reproduced (Fig. 2, lane 3). This might be expected since such tissue was obtained from a site contaminated with normal mucosal flora. After 35 or more cycles of PCR amplification with p11E and p13B, faint bands could sometimes be seen in all reaction mixtures.

Sequence Determination and Development of Specific Primers

The DNA fragment amplified from the splenic extract from Patient 1 was purified, cloned, and then sequenced. Comparison of this sequence with all known DNA sequences in the GenBank/European Molecular Biology Laboratory data libraries showed that this 232-bp fragment could readily be aligned with eubacterial 16S rRNA genes but was not identical to any known sequence. The new organism represented by this sequence was designated strain BA-TF. A new primer was designed, p93E, that corresponds to a highly conserved 16S rRNA sequence at an upstream site. With the use of this primer and p13B, a 480-bp DNA product was amplified from the splenic extract from Patient 1, but not from that from the patient with idiopathic thrombocytopenic purpura or from other nondiseased splenic DNA extracts (data not shown). On the basis of the sequence of this 480-bp fragment, two BA-TF—specific primers — p24E and p12B — were designed (Table 1). These two primer sites are separated by 241 bp and are maximally divergent from known 16S rRNA sequences of eubacteria in those regions.

The two specific primers reproducibly amplified a DNA fragment of the expected size from a spleen sample (Fig. 2, lane 9) and a sample of splenic hilar lymph node from Patient 1 after 35 cycles of PCR. Reaction mixtures without the DNA template, as well as with DNA from tonsillar tissues from two patients and splenic tissue from patients without bacillary angiomatosis, were consistently negative. The absence of detectable product with bacterially contaminated tonsillar tissue demonstrated the specificity of these primers (Fig. 2, lane 8).

Analysis of a Novel 16S rRNA Sequence Found in Unrelated Patients with Bacillary Angiomatosis

To determine whether strain BA-TF was associated with other cases of bacillary angiomatosis, we obtained diseased tissue from three other patients, each seen at other institutions — frozen bone marrow from Patient 2, paraffin-embedded, formalin-fixed skin tissue from Patient 3, and paraffin-embedded, formalin-fixed lymph-node tissue from Patient 4. All three samples yielded a PCR product of the same expected size when the specific primers were used (Fig. 2, lanes 10 through 12). Forty cycles of amplification were performed with DNA extracted from formalin-fixed, paraffin-embedded tissues. These DNA products were cloned and sequenced. To minimize any chance of carryover contamination from our laboratory, tissue from Patients 3 and 4 was studied at a site distant from the Stanford campus.

The DNA sequence determined from the frozen tissue from Patient 2 was identical to that of BA-TF throughout the 241 base positions in the amplified fragment. One of two recombinant clones from the formalin-fixed tissue from Patient 4 had the same sequence as BA-TF; the other clone had a sequence that differed from BA-TF at 2 of 241 positions. Two recombinant clones from Patient 3 were sequenced; these sequences were found to be identical and differed from the BA-TF sequence at only 4 positions. To determine whether this sequence heterogeneity might be found with frozen tissue samples as well, the product amplified from the frozen splenic hilar lymph-node tissue from Patient 1 with the specific primers was cloned. In comparison with the original splenic BA-TF sequence, there was 1 nucleotide substitution of a total of 1350 positions from eight clones. This result is consistent with previously published error frequencies for thermostable Taq polymerase during amplification of undamaged human genomic DNA.22

Figure 3. Figure 3. Phylogenetic Trees.

The tree in Panel A, based on 16S rDNA sequences, indicates evolutionary relations between the new organism, BA-TF, and representative purple eubacteria (see Methods for details).

The evolutionary distance is proportional only to the horizontal scale, expressed as the number of point mutations per sequence position. The tree was constructed with the use of sequences from Pseudomonas testosteroni, Rhodocyclus gelatinosa, and Rhodopseudomonas marina, which are not shown. Panel B shows an "unrooted" phylogenetic tree with the three kingdoms —eukaryotes, archaeobacteria, and eubacteria. In this tree the length of the line segments reflects the evolutionary distance. Crosshatching indicates the location of the purple eubacterial phylum. The scale bar represents 0.1 point mutation per sequence position. (Adapted from Turner et al.37)

Table 2. Table 2. 16S rRNA Sequence Similarities between the New Organism, BA-TF, and Selected Rickettsia and Other Eubacteria.

Optimized sequence alignments were prepared with use of the 480-bp 16S rRNA sequence of BA-TF (excluding the primer sequences) and those of other organisms. These alignments were used to construct a phylogenetic tree according to standard techniques.36 , 37 As shown in Figure 3, strain BA-TF falls within the α subdivision of purple eubacteria and specifically within the R. quintana—Agrobacterium tumefaciens complex. This tree is consistent with the evolutionary relation of the purple eubacteria and the tribe Rickettsieae, on the basis of comparisons of full-length 16S rRNA gene sequences.38 Table 2 shows the percent sequence similarities between BA-TF and other eubacteria. Even though the data in the table are based on selected 16S rRNA sequence positions, calculations of sequence similarities using the entire available BA-TF sequence gave equivalent results (data not shown). In a similar fashion we analyzed the sequence from Patient 3 and the deviant sequence from Patient 4. The results for each showed no significant difference from the analysis for BA-TF: the same relations to other organisms were confirmed. The sequence from Patient 3 was even more closely related to R. quintana; this limited portion of the 16S rRNA sequence was identical to the corresponding region of the R. quintana 16S rRNA sequence. Phylogenetic trees, constructed from each of these two deviant sequences, were nearly identical to the original (data not shown). The 480-bp 16S rRNA sequence from BA-TF has been deposited in the GenBank/European Molecular Biology Laboratory data libraries.

Using the broad-range primers p93E and p13B, no irradiation of reagents, and 40 cycles of PCR, we occasionally obtained a visible DNA product when no DNA template or extract was added to the reaction mixture. The DNA product from two such reactions was cloned and sequenced. In these cases the sequences were identical to each other and were most closely related to Pseudomonas cepacia, a β purple eubacterium commonly found in aquatic environments (data not shown). These findings strengthen the specificity of the association between the BA-TF sequence and bacillary angiomatosis.

Many of the human pathogens closely related to strain BA-TF are in the tribe Rickettsieae, although this tribe is phylogenetically dispersed throughout the purple eubacterial phylum.38 Brucella abortus is the second most closely related organism to BA-TF among those whose 16S rRNA sequences have been characterized. According to our analysis, BA-TF is a unique rickettsia-like organism that is most closely related to R. quintana, the causative agent of trench fever. A. tumefaciens and other bacteria in this complex are plant pathogens and soil organisms, suggesting that BA-TF, as well as some other rickettsia, may have arisen as a plant-associated organism.39

Discussion

Our understanding of microbial pathogens and the diseases they cause is limited by our ability to cultivate these organisms. A number of human diseases are linked to microbes that can be consistently visualized but not grown. Their identity has been a mystery. We have applied a technique that circumvents the absolute need to isolate or grow a disease-associated organism but that nonetheless may reveal its phylogenetic relation. Such relations may predict important and useful features of an organism, since evolutionarily related organisms are expected to share some characteristics.

Bacillary angiomatosis is a well-defined, increasingly common infection of the skin and visceral organs, most often seen in immunocompromised hosts, especially patients with HIV infection.8 , 40 It is potentially fatal if not treated. The numerous bacilli found in the lesions of this disease have been compared to those seen in lesions from cat scratch disease, although the vascular proliferative response in bacillary angiomatosis is not a common feature of cat scratch disease. The bacilli in each disease are pleomorphic, measure 0.2 to 0.5 μm by 1 to 3 μm, grow in clumps, and have trilaminar walls.4 These bacilli have resisted identification. We have used eubacterial broad-range primers to amplify a novel 16S rDNA sequence from diseased tissue from four unrelated patients with bacillary angiomatosis. The sequence obtained from three of these patients was identical; we designated it the reference sequence, BA-TF. The sequences found in the third patient and in one clone from the fourth differed from BA-TF only minimally. None of these sequences were detected in tissues from patients without bacillary angiomatosis. On the basis of our phylogenetic analysis of these sequences, we conclude that bacillary angiomatosis is caused by a rickettsia-like organism, most closely related to R. quintana. Even though the sequence in the third patient was identical to the corresponding region of the R. quintana 16S rRNA gene, the limited amount of available sequences from this patient and the predominance of the BA-TF sequence in the other patients with bacillary angiomatosis suggest that BA-TF more accurately reflects the true 16S rRNA sequence of the pathogen. Given the paucity of a purple bacterial 16S rRNA sequences, it is not yet possible to assign BA-TF to a genus or species; however, the sequence similarity between BA-TF and R. quintana is the same as that between Rickettsia typhi and Rickettsia rickettsii (98.3 percent).

PCR is an extremely sensitive method of detecting target DNA sequences. Sequences of 16S rRNA can be amplified and compared with a reference-sequence data base in order to infer phylogenetic relations.24 , 35 The use of broad-range or "universal" 16S rRNA primers,23 however, presents special problems. Contamination of PCR reagents and target with a few bacteria, bacterial genomes, or very small amounts of eukaryotic DNA is virtually impossible to prevent and can lead to false positive results. Exquisite care must be taken to keep contamination to a minimum.41 , 42 We found that the use of no more than 25 PCR cycles with broad-range primers prevented the detection of product with uninfected tissues (unpublished data). Ultraviolet irradiation has been shown to reduce carryover contamination of PCR reagents and was useful in our 40-cycle amplifications.43 , 44 The use of specific primers reduces the need for such stringency. Our association of the BA-TF sequence with tissue from patients with bacillary angiomatosis was strengthened by the independent processing and amplification of the various samples in different laboratories. In addition, samples of internal tissue from patients without bacillary angiomatosis never produced a visible amplification product.

Despite the small number of nucleotide differences between BA-TF and the 16S rRNA sequence of Patient 3 and of a single clone from Patient 4, these sequences are all related to 16S rRNA sequences of the same organisms. On the basis of the 16S rRNA nucleotide positions included in our phylogenetic analysis, the sequences from these two patients differed by only 0.7 percent from the BA-TF sequence. However, these differences do raise some important issues. Could sequence heterogeneity indicate the presence of multiple but related strains within the diseased tissue — i.e., a mixed infection? This is unlikely for two reasons. First, at rRNA positions known to participate in the formation of base pairs in secondary structures, we have seen nucleotide substitutions without compensatory substitutions at the corresponding paired position; such changes would be disruptive to the rRNA secondary structure. Second, sequence heterogeneity was apparent only in Patients 3 and 4, for whom formalin-fixed tissue was studied. Another possibility is that the heterogeneity is due in part to the amplification of different copies of the 16S rRNA gene from the same organism. A small number of nucleotide differences are known to exist within the seven copies of the E. coli rRNA genes.45 , 46 This possibility cannot be ruled out for BA-TF; however, the explanation we most favor for the observed sequence differences is Taq polymerase error. Taq polymerase can mistakenly incorporate nucleotides, particularly during the amplification of damaged DNA.47 , 48 Formalin fixation can damage nucleic acids at particular positions and therefore contributes to higher polymerase-error rates. Under these conditions, when there are few original templates in a tissue sample, it is imperative to sequence multiple clones of the amplified product.

Complete 16S rRNA sequences are not necessary for the construction of meaningful phylogenetic trees23; however, the amount of partial sequence available and the composition of the reference 16S rRNA sequence collection are important variables. Our evolutionary-distance tree duplicated the relations among the Rickettsieae and other purple bacteria established by Weisburg et al. using full 16S rRNA sequences.38 Thus, we are confident about our phylogenetic analysis of strain BA-TF. In addition, a rickettsial origin is consistent with previous morphologic descriptions of the bacillus causing bacillary angiomatosis.

What are the implications of the relation between BA-TF, R. quintana, and other α purple bacteria? R. quintana is transmitted by the body louse and is the cause of trench fever, a systemic illness characterized by persistent or recurrent fever, bone and muscle pain, rash, and splenomegaly.49 R. quintana is sensitive in vitro to erythromycin and the tetracyclines,50 agents that have been used successfully to treat bacillary angiomatosis.40 Although trench fever and disseminated bacillary angiomatosis share some clinical features, the vascular lesions of bacillary angiomatosis distinguish it from trench fever. A separate member of the order Rickettsiales, Bartonella bacilliformis, can produce lesions (verruga peruana) that are indistinguishable from those caused by bacillary angiomatosis, and this has led to suggestions that it might be the cause of this disease.4 , 51 There are no 16S rRNA sequence data available for Bart. bacilliformis; therefore, its relation to BA-TF remains unclear. B. abortus has features in common with the agent of bacillary angiomatosis: both are transmitted to humans from domestic animals by percutaneous inoculation, frequently involve the reticuloendothelial system, and can cause chronic febrile illnesses. On the other hand, infrequent skin involvement, a granulomatous rather than a vascular proliferative host-tissue response, a gram-negative staining pattern, and 16S rRNA sequence distinguish brucellosis from bacillary angiomatosis.52

The grouping of R. quintana and BA-TF with the agrobacteria and rhizobacteria suggests a plant or soil origin.39 Humans are the only known reservoir for R. quintana, but the other characterized member of this genus, R. vinsonii, is carried by voles.53 The grouping of BA-TF with these organisms would be consistent with its transmission from the environment to humans by means of a break in the skin. What role, if any, cats have, and the exact relation of BA-TF to the bacillus of cat scratch disease must await further analysis of the latter organism. We are currently using the techniques described in this report to identify the cat-scratch-disease bacillus.

R. quintana is the only pathogenic rickettsia to grow on cell-free medium.54 The relation of BA-TF to this organism may help us develop appropriate culture conditions for the growth of this new pathogen. Growth of BA-TF would facilitate the investigation of the angioproliferative response it elicits in the host.40 Our results also have implications for the diagnosis of bacillary angiomatosis: divergent portions of the BA-TF 16S rRNA gene or adjacent regions of the genome may serve as probes for in situ hybridization. If such probes were labeled with fluorescent dyes, hybridization to the agent of bacillary angiomatosis could be visualized with epifluorescence.55 Alternatively, an in situ PCR-based technique might be feasible.

The discovery of this novel rickettsia-like organism in patients with bacillary angiomatosis is one example of the application of a potentially powerful technique. We expect that this approach will be applicable to other infectious diseases with unclear causes and will expand our understanding of interactions between humans and microbes.

Funding and Disclosures

Supported by Public Health Service grants (AI 26195 [Dr. Falkow] and AI 23796 [Dr. Tompkins]) from the National Institutes of Health, a contract (N14–87-K-0813 [Dr. Schmidt]) with the Office of Naval Research, and an unrestricted gift from Praxis Biologics, Rochester, N. Y. Dr. Relman is a recipient of a Lucille P. Markey Scholar Award in Biomedical Science.

We are indebted to Drs. Carol Kemper and Stanley Deresinski and our other clinical colleagues for their support and enthusiasm; to Dr. Norman Pace, Indiana University, for advice, constructive suggestions, and access to data bases; to Dr. Donald Regula, Department of Pathology, Stanford University, for providing invaluable assistance with tissue samples from Patient 1 and photographs; to Dr. Kemper, Santa Clara Valley Medical Center, San Jose, for providing tissue from Patient 2; to Dr. Philip LeBoit, Department of Pathology, University of California, San Francisco, for providing tissue from Patient 3; to Dr. Ronald F. Dorfman, Department of Pathology, Stanford University, and Dr. Klaus J. Lewin, Department of Pathology, University of California, Los Angeles, for providing tissue from Patient 4; and to Dr. Michele Manos and Ms. Catherine Greer, Cetus Corporation, Emeryville, California, for giving us valuable advice about and assistance with the extraction of DNA from formalin-fixed, paraffin-embedded tissue.

Author Affiliations

From the Department of Microbiology and Immunology (D.A.R., S.F., L.S.T.) and the Division of Infectious Diseases, Department of Medicine (D.A.R., J.S.L., S.F., L.S.T.), Stanford University, Stanford, Calif., and the Department of Biology, Indiana University, Bloomington (T.M.S.). Address reprint requests to Dr. Relman at the Department of Microbiology and Immunology, Fairchild Building D309, Stanford University, Stanford, CA 94305–5402.

References (57)

  1. 1. Stoler MH, Bonfiglio TA, Steigbigel RT, Pereira M. . An atypical subcutaneous infection associated with acquired immune deficiency syndrome . Am J Clin Pathol 1983; 80:714–8.

  2. 2. Cockerell CJ, Whitlow MA, Webster GF, Friedman-Kien AE. . Epithelioid angiomatosis: a distinct vascular disorder in patients with the acquired immunodeficiency syndrome or AIDS-related complex . Lancet 1987; 2:654–6.

  3. 3. Angritt P, Tuur SM, Macher AM, et al. . Epithelioid angiomatosis in HIV infection: neoplasm or cat-scratch disease? Lancet 1988; 1:996.

  4. 4. LeBoit PE, Berger TG, Egbert BM, et al. . Epithelioid haemangioma-like vascular proliferation in AIDS: manifestation of cat scratch disease bacillus infection? Lancet 1988; 1:960–3.

  5. 5. LeBoit PE, Berger TG, Egbert BM, Beckstead JH, Yen TS, Stoler MH. . Bacillary angiomatosis: the histopathology and differential diagnosis of a pseudoneoplastic infection in patients with human immunodeficiency virus disease . Am J Surg Pathol 1989; 13:909–20.

  6. 6. Koehler JE, LeBoit PE, Egbert BM, Berger TG. . Cutaneous vascular lesions and disseminated cat-scratch disease in patients with the acquired immunodeficiency syndrome (AIDS) and AIDS-related complex . Ann Intern Med 1988; 109:449–55.

  7. 7. Milam MW, Balerdi MJ, Toney JF, Foulis PR, Milam CP, Behnke RH. . Epithelioid angiomatosis secondary to disseminated cat scratch disease involving the bone marrow and skin in a patient with acquired immune deficiency syndrome: a case report . Am J Med 1990; 88:180–3.

  8. 8. Kemper CA, Lombard CM, Deresinski SC, Tompkins LS. . Visceral bacillary epithelioid angiomatosis: possible manifestations of disseminated cat scratch disease in the immunocompromised host: a report of two cases . Am J Med 1990; 89:216–22.

  9. 9. Cockerell CJ, Bergstresser PR, Myrie-Williams C, Tierno PM. . Bacillary epithelioid angiomatosis occurring in an immunocompetent individual . Arch Dermatol 1990; 126:787–90.

  10. 10. Carithers HA, Carithers CM, Edwards RO Jr. . Cat-scratch disease: its natural history . JAMA 1969; 207:312–6.

  11. 11. Carithers HA. . Cat-scratch disease: an overview based on a study of 1,200 patients . Am J Dis Child 1985; 139:1124–33.

  12. 12. Margileth AM, Wear DJ, English CK. . Systemic cat scratch disease: report of 23 patients with prolonged or recurrent severe bacterial infection . J Infect Dis 1987; 155:390–402.

  13. 13. Lenoir AA, Storch GA, DeSchryver-Kecskemeti K, et al. . Granulomatous hepatitis associated with cat scratch disease . Lancet 1988; 1:1132–6.

  14. 14. Delahoussaye PM, Osborne BM. . Cat-scratch disease presenting as abdominal visceral granulomas . J Infect Dis 1990; 161:71–8.

  15. 15. Wear DJ, Margileth AM, Hadfield TL, Fischer GW, Schlagel CJ, King FM. . Cat scratch disease: a bacterial infection . Science 1983; 221:1403–5.

  16. 16. English CK, Wear DJ, Margileth AM, et al. . Cat-scratch disease: isolation and culture of the bacterial agent . JAMA 1988; 259:1347–52.

  17. 17. Schlossberg D, Morad Y, Krause TB, Wear DJ, English CK. . Culture-proved disseminated cat-scratch disease in acquired immunodeficiency syndrome . Arch Intern Med 1989; 149:1437–9.

  18. 18. Fox GE, Stackebrandt E, Hespell RB, et al. . The phylogeny of prokaryotes . Science 1980; 209:457–63.

  19. 19. Woese CR. . Bacterial evolution . Microbiol Rev 1987; 51:221–71.

  20. 20. Saiki RK, Scharf S, Faloona F, et al. . Enzymatic amplification of β-globin genomic sequences and restriction site analysis for diagnosis of sickle cell anemia . Science 1985; 230:1350–4.

  21. 21. Mullis KB, Faloona FA. Specific synthesis of DNA in vitro via a polymerase-catalyzed chain reaction. In: Methods in enzymology. Vol. 155. Wu R, ed., Recombinant DNA, Part F. San Diego: Academic Press, 1987:335–50.

  22. 22. Saiki RK, Gelfand DH, Stoffel S, et al. . Primer-directed enzymatic amplification of DNA with a thermostable DNA polymerase . Science 1988; 239:487–91.

  23. 23. Lane DJ, Pace B, Olsen GJ, Stahl DA, Sogin ML, Pace NR. . Rapid determination of 16S ribosomal RNA sequences for phylogenetic analyses . Proc Natl Acad Sci U S A 1985; 82:6955–9.

  24. 24. Medlin L, Elwood HJ, Stickel S, Sogin ML. . The characterization of enzymatically amplified eukaryotic 16S-like rRNA-coding regions . Gene 1988; 71:491–9.

  25. 25. Eisenstein BI. . The polymerase chain reaction: a new method of using molecular genetics for medical diagnosis . N Engl J Med 1990; 322:178–83.

  26. 26. Chen K, Neimark H, Rumore P, Steinman CR. . Broad range DNA probes for detecting and amplifying eubacterial nucleic acids . FEMS Microbiol Lett 1989; 48:19–24.

  27. 27. Goelz SE, Hamilton SR, Vogelstein B. . Purification of DNA from formaldehyde fixed and paraffin embedded human tissue . Biochem Biophys Res Commun 1985; 130:118–26.

  28. 28. Dubeau L, Chandler LA, Gralow JR, Nichols PW, Jones PA. . Southern blot analysis of DNA extracted from formalin-fixed pathology specimens . Cancer Res 1986; 46:2964–9.

  29. 29. Sambrook J, Fritsch EF, Maniatis T. Molecular cloning: a laboratory manual. 2nd ed. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory, 1989.

  30. 30. Shibata DK, Arnheim N, Martin WJ. . Detection of human papilloma virus in paraffin-embedded tissue using the polymerase chain reaction . J Exp Med 1988; 167:225–30.

  31. 31. Wright DK, Manos MM. Sample preparation from paraffin-embedded tissues. In: Innis MA, Gelfand DH, Sninsky JJ, White TJ, eds. PCR protocols: a guide to methods and applications. San Diego, Calif.: Academic Press, 1990:153–8.

  32. 32. Dower WJ, Miller JF, Ragsdale CW. . High efficiency transformation of E. coli by high voltage electroporation . Nucleic Acids Res 1988; 16:6127–45.

  33. 33. Sanger F, Nicklen S, Coulson AR. . DNA sequencing with chain-terminating inhibitors . Proc Natl Acad Sci U S A 1977; 74:5463–7.

  34. 34. Jukes TH, Cantor CR. Evolution of protein molecules. In: Munro HN, ed. Mammalian protein metabolism. Vol. 3. New York: Academic Press, 1969:21–132.

  35. 35. Olsen GJ. . Earliest phylogenetic branchings: comparing rRNA-based evolutionary trees inferred with various techniques . Cold Spring Harb Symp Quant Biol 1987; 52:825–37.

  36. 36. Fitch WM, Smith TF. . Optimal sequence alignments . Proc Natl Acad Sci USA 1983; 80:1382–6.

  37. 37. Turner S, Burger-Wiersma T, Giovannoni SJ, Mur LR, Pace NR. . The relationship of a prochlorophyte Prochlorothrix hollandica to green chloroplasts . Nature 1989; 337:380–2.

  38. 38. Weisburg WG, Dobson ME, Samuel JE, et al. . Phylogenetic diversity of the Rickettsiae . J Bacteriol 1989; 171:4202–6.

  39. 39. Weisburg WG, Woese CR, Dobson ME, Weiss E. . A common origin of rickettsiae and certain plant pathogens . Science 1985; 230:556–8.

  40. 40. Cockerell CJ, LeBoit PE. . Bacillary angiomatosis: a newly characterized, pseudoneoplastic, infectious, cutaneous vascular disorder . J Am Acad Dermatol 1990; 22:501–12.

  41. 41. Kwok S, Higuchi R. . Avoiding false positives with PCR . Nature 1989; 339:237–8.

  42. Erratum , Nature 1989; 339:490.

  43. 42. Kitchin PA, Szotyori Z, Fromholc C, Almond N. . Avoidance of false positives . Nature 1990; 344:201.

  44. 43. Sarkar G, Sommer SS. . Shedding light on PCR contamination . Nature 1990; 343:27.

  45. 44. Cimino GD, Metchette K, Isaacs ST, Zhu YS. . More false-positive problems . Nature 1990; 345:773–4.

  46. 45. Brownlee GG, Sanger F, Barrell BG. . The sequence of 5s ribosomal ribonucleic acid . J Mol Biol 1968; 34:379–412.

  47. 46. Jinks-Robertson S, Nomura M. Ribosomes and tRNA. In: Neidhardt FC, ed. Escherichia coli and Salmonella typhimurium: cellular and molecular biology. 2. Washington, D.C.: American Society for Microbiology, 1987:1358–85.

  48. 47. Pääbo S, Irwin DM, Wilson achéal. . DNA damage promotes jumping between templates during enzymatic amplification . J Biol Chem 1990; 265:4718–21.

  49. 48. Ennis PD, Zemmour J, Salter RD, Parham P. . Rapid cloning of HLA-A,B cDNA by using the polymerase chain reaction: frequency and nature of errors produced in amplification . Proc Natl Acad Sci U S A 1990; 87:2833–7.

  50. 49. Vinson JW, Varela G, Molina-Pasquel C. . Trench fever. 3. Induction of clinical disease in volunteers inoculated with Rickettsia quintana propagated on blood agar . Am J Trop Med Hyg 1969; 18:713–22.

  51. 50. Myers WF, Grossman DM, Wisseman CL Jr. . Antibiotic susceptibility patterns in Rochalimaea quintana, the agent of trench fever . Antimicrob Agents Chemother 1984; 25:690–3.

  52. 51. Kreier JP, Ristic M. . The biology of hemotrophic bacteria . Annu Rev Microbiol 1981; 35:325–38.

  53. 52. Spink WW. The nature of brucellosis. Minneapolis: University of Minnesota Press, 1956.

  54. 53. Weiss E, Dasch GA. . Differential characteristics of strains of Rochalimaea: Rochalimaea vinsonii sp. nov., the Canadian vole agent . Int J Syst Bacteriol 1982; 32:305–14.

  55. 54. Varela G, Vinson JW, Molina-Pasquel C . Trench fever. II. Propagation of Rickettsia quintana on cell-free medium from the blood of two patients . Am J Trop Med Hyg 1969; 18:708–12.

  56. 55. DeLong EF, Wickham GS, Pace NR. . Phylogenetic stains: ribosomal RNA-based probes for the identification of single cells . Science 1989; 243:1360—3.

  57. Erratum , Science 1989; 245:1312.

Citing Articles (650)

    Letters

    Figures/Media

    1. Figure 1. Splenic Nodule from Patient 1 after Warthin—Starry Staining.
      Figure 1. Splenic Nodule from Patient 1 after Warthin—Starry Staining.

      In Panel A, inflammatory cells, focal areas of necrosis, and numerous epithelioid endothelial cells forming vascular channels can be seen (×670). In Panel B, at a higher magnification (×3300), pleomorphic bacillary organisms are apparent and often are in clumps (arrows). Similar organisms were seen in samples of splenic hilar lymph node and liver from the same patient.

    2. Table 1. Oligonucleotide Primers Used for Amplification of 16S rDNA.*
      Table 1. Oligonucleotide Primers Used for Amplification of 16S rDNA.*
    3. Figure 2. Agarose-Gel Electrophoresis of PCR-Amplified 16S rDNA Fragments from Tissue DNA Extracts.
      Figure 2. Agarose-Gel Electrophoresis of PCR-Amplified 16S rDNA Fragments from Tissue DNA Extracts.

      Broad-range eubacterial 16S rDNA primers — p11E and p13B — were used for the reactions in lanes 1 through 5. They prime the amplification of a 232-bp fragment. Specific primers — p24E and p12B — were used for the reactions corresponding to lanes 6 through 12; they generate a 296-bp fragment. Lanes 1 and 6 contain no added DNA extract; lanes 2 and 7, a spleen sample from a patient with idiopathic thrombocytopenic purpura; lanes 3 and 8, tonsillar tissue from a patient with recurrent tonsillitis; lanes 4 and 9, a spleen sample from Patient 1; lane 5, a sample of splenic hilar lymph node from Patient 1; lane 10, bone marrow from Patient 2; lane 11, tissue from a skin-lesion biopsy from Patient 3; and lane 12, lymph-node tissue from Patient 4. The size of some of the DNA standards is indicated in kilobases.

    4. Figure 3. Phylogenetic Trees.
      Figure 3. Phylogenetic Trees.

      The tree in Panel A, based on 16S rDNA sequences, indicates evolutionary relations between the new organism, BA-TF, and representative purple eubacteria (see Methods for details).

      The evolutionary distance is proportional only to the horizontal scale, expressed as the number of point mutations per sequence position. The tree was constructed with the use of sequences from Pseudomonas testosteroni, Rhodocyclus gelatinosa, and Rhodopseudomonas marina, which are not shown. Panel B shows an "unrooted" phylogenetic tree with the three kingdoms —eukaryotes, archaeobacteria, and eubacteria. In this tree the length of the line segments reflects the evolutionary distance. Crosshatching indicates the location of the purple eubacterial phylum. The scale bar represents 0.1 point mutation per sequence position. (Adapted from Turner et al.37)

    5. Table 2. 16S rRNA Sequence Similarities between the New Organism, BA-TF, and Selected Rickettsia and Other Eubacteria.
      Table 2. 16S rRNA Sequence Similarities between the New Organism, BA-TF, and Selected Rickettsia and Other Eubacteria.

      More Research