Efficacy of Rifampicin in the Treatment of Experimental Acute Canine Monocytic Ehrlichiosis

Objectives: To assess the efficacy of rifampicin in achieving clinical and haematological recovery and clearing infection in dogs with experimentally induced acute monocytic ehrlichiosis. Methods: Five Ehrlichia canis-infected Beagle dogs were treated with rifampicin (10 mg/kg/24 h orally for 3 weeks), nine E. canis-infected dogs received no treatment (infected untreated dogs) and two dogs served as uninfected controls. Clinical score, platelet counts, immunofluorescent antibody titres and PCR detection of E. canis-specific DNA in blood, bone marrow and spleen aspirates were evaluated on post-inoculation days 21 (start of rifampicin), 42 (end of rifampicin) and 98 (end of the study). Results: By day 21 post-inoculation, all infected dogs became clinically ill and thrombocytopenic, seroconverted and were PCR positive in at least one tissue. Clinical scores and antibody titres did not differ between the treated and infected untreated dogs throughout the study. The rifampicin-treated dogs experienced an earlier resolution of their thrombocytopenia (Kaplan–Meier survival plot, P ¼0.048), and the median platelet counts were significantly higher in the treated compared with the infected untreated dogs on post-inoculation days 42 (P ¼0.0233) and 98 (P ¼ 0.0195). At the end of the study, three treated and six untreated infected dogs remained PCR positive in one tissue each. Conclusions: The rifampicin treatment regimen applied in this study hastened haematological recovery, but was inconsistent in eliminating the acute E. canis infection.


Introduction
Tetracyclines are considered to be first-line drugs in the treatment of canine monocytic ehrlichiosis (CME) caused by Ehrlichia canis. 1,2However, only doxycycline has been critically evaluated in experimental or natural disease, and there are conflicting reports on the clearance of the infection following different doxycycline treatment protocols.In experimental acute or subclinical CME, it was very effective in eliminating the infection. 3 -6In contrast, in other experimental or clinical studies, treatment with various doxycycline regimens failed to eradicate the infection in 25% -100% of the acutely, subclinically or chronically infected dogs, despite their clinical and haematological recovery. 7 -11Collectively, current evidence implies that doxycycline effectively ameliorates the E. canis-induced clinical and haematological abnormalities, but is not consistently effective in clearing the infection.In addition, some dogs may be intolerant to doxycycline, 12 thus justifying the evaluation of medications that could be used as alternatives to doxycycline.
Recent data in a limited number of dogs have indicated that rifampicin, an inhibitor of the B subunit of DNA-dependent RNA polymerase, may be partially effective in eliminating ehrlichiaemia in dogs experiencing the subclinical or chronic phases of the disease. 9,13Of comparative interest, in human ehrlichiosis patients with a specific doxycycline contraindication, rifampicin is used as an alternative chemotherapeutic agent, affording rapid clinical recovery, although its efficacy in clearing the infection has not been evaluated. 14,15There are currently no data on the potential effectiveness of rifampicin in acute CME or on the elimination of E. canis from tissues other than blood.
This study was undertaken to evaluate the efficacy of rifampicin in clearing E. canis from blood, bone marrow (BM) and spleen aspirates, and in achieving clinical and haematological recovery in dogs with experimental acute CME.

Experimental dogs
Sixteen Beagle dogs, seven males and nine females, with an age ranging from 5 to 70 months (median 11 months), purchased from the School of Veterinary Medicine (University of Thessaly, Greece) experimental dog colony were used in the study.All dogs were fully vaccinated against canine parvovirus-2, canine adenovirus-2, canine distemper virus, Leptospira icterohaemorrhagiae/Leptospira canicola and rabies, and were subject to strict ectoparasite and endoparasite control measures.Before the experimental infection, they were healthy and had no haematological or biochemical abnormalities in serial evaluations over a 30 day acclimatization period.The dogs were seronegative to E. canis, Babesia canis (Fluo BABESIA canis; Agrolabo, Scarmagno, Italy; positive titre: ≥1/100) and Leishmania infantum antibodies (Fluoleish BVT, Virbac, Carros, France; positive titre: ≥1/100) by indirect fluorescence antibody (IFA) testing, as well as to Dirofilaria immitis antigens (Snap 3Dx; IDEXX, Westbrook, ME, USA).All dogs were PCR negative for E. canis DNA in blood, spleen and BM aspirates before the inoculation with E. canis.The duration of the study after the experimental inoculation was 98 days (14 weeks).The study was approved by the Research and Ethics Committee, Faculty of Veterinary Medicine, Aristotle University of Thessaloniki (458/ 23-6-2009) and the Directorate of Veterinary Services, Thessaloniki (13/3152, 16-03-2010), Greece.In every aspect, handling of the animals strictly adhered to the rules and regulations governing animal experimentation.

Group allocation and treatment protocol
The dogs were randomized into three groups by utilizing a table of random numbers.Treated dogs (n ¼5) included E. canis-infected dogs that were treated with rifampicin (Rifadin) at 10 mg/kg/24 h, orally, for 3 weeks.All treated dogs were also given dexamethasone (0.5 mg/kg dexamethasone sodium, intravenously) once, 15 days after the completion of rifampicin, to precipitate any latent E. canis infection.Infected untreated dogs (n¼9) included E. canis-infected dogs that received no treatment.Two Beagles served as uninfected controls.Administration of rifampicin started on post-inoculation day (PID) 21, when all treated dogs matched the following criteria: development of thrombocytopenia, seroconversion to E. canis and positive PCR in at least one of the tissues tested (blood, BM or spleen), with or without clinical manifestations.
The rifampicin dosage applied in the present study was based on previous suggestions 16 and a preliminary pharmacokinetic evaluation conducted 2 months prior to this study on eight healthy Beagles, in which a high serum concentration of rifampicin (mean C max 0.01554+ 0.00183 mg/L and mean concentration of 0.00431+0.00056mg/L after 24 h) was achieved following a single oral dose of 10 mg/kg of the product used in the present study (G.Batzias, K. Theodorou and M. E. Mylonakis, unpublished results).

Experimental infection
Thirteen dogs were inoculated intravenously with 5 mL of heparinized E. canis-infected blood that contained 12.5×10 4 16S rDNA copies, quantified by an E. canis-specific real-time PCR. 17The blood was drawn from an acutely ill reservoir Beagle, artificially infected with the Israeli E. canis strain 611 (propagated in DH82 cell culture), 18,19 donated by the Koret School of Veterinary Medicine, The Hebrew University of Jerusalem, Israel, as soon as the dog became thrombocytopenic (platelet count 44×10 9 /L) and seroconverted to E. canis (IFA titre 1/1600) and E. canis DNA could be detected in blood and spleen samples on PID 19.The two uninfected dogs were inoculated with 5 mL of heparinized blood from a healthy uninfected Beagle.

Clinical and haematological examination
Each dog had serial clinical examinations performed on days 30, 14, 8, 6, 4 and 2 before and on the day of inoculation (day 0); thereafter, this was carried out every 2 days until PID 42 and weekly until PID 98. Physical examinations were performed by two assessors (M.E. M. and C. K. K.) who had no knowledge of the group allocation.A clinical evaluation sheet was filled out qualitatively (i.e.0 ¼'absence' and 1 ¼'presence' of the clinical finding).For each dog, the cumulative clinical score per examination was calculated.Clinical manifestations systematically evaluated included depression/lethargy, fever (.39.58C), lymphadenomegaly, bleeding tendency, mucosal pallor and palpable splenomegaly.Anorexia was assessed by the primary caregiver of the dogs (K.T.) in an open-label manner and was also included in the cumulative clinical score.
Complete blood counts were performed on an Advia Hematology Analyzer (Advia 120 Haematology System; Bayer, Tarrytown, NY, USA).All dogs were tested 2 weeks before and on inoculation day.Thereafter, treated and infected untreated dogs were tested on PIDs 7, 14, 21, 28, 35, 42, 70, 84 and 98, while uninfected dogs were tested on PIDs 7, 14, 35, 70 and 98. Giemsa-stained blood smears were reviewed to confirm the presence of thrombocytopenia throughout the study.Aliquots of EDTA-treated blood samples were stored at 2208C until molecularly analysed.

BM and spleen sampling
BM and spleen aspirates (0.5-1 mL) were obtained as previously described 20,21 under short-term sedation induced by acetylpromazine (0.05 mg/kg acepromazine, intravenously) and thiopental sodium (Pentothal; 5 mg/kg, intravenously).The aspirates were collected in EDTA-treated tubes and stored at 2208C until molecularly analysed.Giemsa-stained cytology smears were reviewed after each aspiration to confirm the adequacy of the samples.Aspiration of BM and spleen was performed for all dogs 2 days prior to the experimental inoculation; thereafter, treated and infected untreated dogs were sampled on PIDs 7, 21, 42, 70 and 98, and uninfected dogs on PIDs 42 and 98.

Serological testing for E. canis
An IFA test was used to detect anti-E.canis IgG antibodies on 15-well slides using E. canis acetone-fixed antigen suspension (Fluo EHRLICHIA canis; Agrolabo).Serial 2-fold dilutions of serum samples from the dogs were reacted with fluorescein-conjugated rabbit anti-canine immunoglobulin G (anti-dog IgG; Sigma-Aldrich, St Louis, MO, USA) with an end-titre of 1/1600.A titre equal to, or higher, than 1/100 was considered positive.Paired serum samples (14 days prior to inoculation and on the day of inoculation) were initially tested for all dogs; thereafter, Theodorou et al.

PCR amplification and analysis of E. canis DNA
Two PCR assays were carried out in this study in two independent laboratories for confirmation of ehrlichial infection and response to treatment.They were performed with whole-genome DNA extracted by use of a commercial kit according to the manufacturer's protocol (NucleoSpin Tissue kit; Macherey-Nagel, Dueren, Germany), from 200 mL of blood samples and 50 mL of BM and spleen aspirates, and stored at 2208C until analysed.
An E. canis-specific nested PCR assay detecting the E. canis 16S rRNA gene was performed at the Laboratory of Microbiology, Faculty of Veterinary Medicine, Aristotle University of Thessaloniki, Greece, using the primers ECC 5 ′ -AGAACGAACGCTGGCGGCAAGCC-3 ′ and ECB 5 ′ -CGTATT ACCGCGGCTGCTGGCA-3 ′ for the primary amplification, and 'canis' 5 ′ -CAAT TATTTATAGCCTCTGGCTATAGGA-3 ′ and HE3 5 ′ -TATAGGTACCGTCATTATCTT CCCTAT-3 ′ for the secondary amplification, as previously described, 11,22 except that Platinum Taq DNA Polymerase High Fidelity (Invitrogen, Carlsbad, CA, USA) was used for optimum amplification.In each set of amplifications, two positive controls (blood and BM DNA from a molecularly confirmed E. canis-infected dog) and a negative control (without DNA template) were included.The identity of 'canis'/HE3 PCR amplicons (389 bp) as E. canis was established by subsequent sequencing of both strands of the purified amplicons performed by VBC-BIOTECH Service GmbH (Vienna, Austria) using the representative PCR primers.Amplicon sequences were compared with the reference '611' (accession number U26740) by multiple sequence alignment by use of the ClustalW program (Conway Institute, Dublin, Ireland). 23olecular evidence of infection was also assessed by the Koret School of Veterinary Medicine, which performed an E. canis-specific real-time PCR assay targetting a 123 bp fragment from a single copy of E. canis 16S rRNA gene, as previously described. 17Primers used were Ec-16S-F 5 ′ -TCGCTATTAGATGAGCCTACGT and Ec-16S-R 5 ′ -GAGTCTGGACCGTATC TCAG.The reactions were carried out using Thermo Start Master Mix (Thermo Scientific, Germany) and SYTO 9 fluorescent dye (Thermo Scientific) in a Corbett Research Rotor-Gene 6000 (Corbett Research Pty Ltd, Sydney, Australia).The protocol used was the same as that described in Peleg et al., 17 with a minor change to the start-up temperature: 15 min at 958C instead of 7 min, owing to the use of a different enzyme.Sequence analysis and comparisons were conducted using the forward primer Ec-16S-F and performed at the Hebrew University Sequencing Core.Sequences were evaluated with ChromasPro software version 1.33 (Technelysium Pty Ltd, South Brisbane, Australia) and compared with sequence data available from GenBank using the BLAST 2.2.9 program (http://www.ncbi.nlm.nih.gov/BLAST/).
All 16 dogs were tested by nested PCR assay of blood, BM and spleen prior to E. canis inoculation.Thereafter, treated and infected untreated dogs were tested by the nested PCR assay on PID 7, and by both nested and real-time PCR on PIDs 21, 42, 70 and 98, while uninfected dogs were tested on PIDs 42 and 98.With the exception of the E. canis-infected blood of the reservoir dog, which was quantified by the real-time PCR, both nested and real-time PCR results were recorded as either negative or positive.

Statistical analysis
The Wilcoxon-Mann-Whitney test was used to compare the median values of the clinical score during the pre-treatment (PIDs 1 -21), treatment (PIDs 22 -42) and post-rifampicin treatment (PIDs 43-98) periods, and the median values of clinical score, platelet counts and E. canis IFA titres between treated and infected untreated dogs at rifampicin initiation (PID 21), rifampicin completion (PID 42) and the end of the study (PID 98).A Kaplan-Meier survival plot was developed to compare the cumulative probability of resolution of thrombocytopenia with time during rifampicin treatment between treated and infected untreated dogs.The survival curves were evaluated with a log-rank test.All tests were done using Stata Statistical Software Release 9.2 (College Station, TX, USA: StataCorp LP, 2006) and evaluated for significance at the 5% level.
The median values of haematocrit, leucocytes and platelets after E. canis inoculation are depicted in Figure 1.By PID 14, 12/14 (86%) of the infected dogs (four treated and eight untreated dogs) developed mild-to-moderate anaemia; on PID 21, only five untreated dogs remained anaemic, and by PID 42, none of the infected dogs was still anaemic.By PID 14, 13/14 (93%) of the infected dogs (five treated and eight untreated dogs) became mildly leucopenic; on PID 21, only two treated and two untreated dogs remained leucopenic, and by PID 42, none of the dogs was still leucopenic.By PID 21, all infected dogs developed thrombocytopenia (median 72.5×10 9 /L, range 44 -101×10 9 /L, reference intervals 200-500×10 9 /L).On PID 42, one treated dog and six untreated dogs were still thrombocytopenic, while on PID 98, all treated dogs had normal platelet counts and four untreated dogs were still thrombocytopenic.On PID 21, median platelet counts did not differ between treated and untreated dogs; however, treated dogs had significantly higher platelet counts compared with the untreated infected dogs on PID 42 (251×10 9 /L versus 168×10 9 /L, P ¼ 0.0233) and 98 (254×10 9 /L versus 205×10 9 /L, P ¼ 0.0195).The Kaplan -Meier survival curves showing the probability of resolution of thrombocytopenia with time (PIDs 21 -42) differed (P ¼ 0.048) between treated and untreated dogs (Figure 2).Among the treated dogs, the median time for normalization of platelets was PID 28, whereas among the untreated ones, resolution of thrombocytopenia was not observed during the period analysed.

E. canis serology
By PID 21, all infected dogs had seroconverted to E. canis antigens (median IgG titres 1/200, range 1/200 -1/1600) and remained seropositive throughout the study period.The two uninfected dogs remained seronegative during the same period.No difference was found in the median anti-E.canis titres between treated and infected untreated dogs on PIDs 21, 42 and 98.
Rifampicin treatment in canine monocytic ehrlichiosis 1621 JAC

PCR-based evaluation of E. canis infection and response to treatment
The PCR (nested and real-time assays) results before and after (7, 21, 42, 70 and 98 days post-inoculation) inoculation with E. canis, with regard to timepoint and tissue tested are presented in Table 2. On PID 21, all treated (median positive tissues 2, range 1 -2) and untreated (median positive tissues 2, range 1 -3) dogs became PCR positive.Upon termination of the study (PID 98), three treated and six untreated dogs were still PCR positive in one tissue each.The three PCR-positive treated dogs were clinically healthy and had normal haematological values; one of them, however, had been persistently thrombocytopenic from PID 42 to 84.Of the six PCR-positive untreated dogs, two were still thrombocytopenic (one with splenomegaly) and one had splenomegaly, but no thrombocytopenia.Two more untreated dogs were thrombocytopenic on PID 98 (one with splenomegaly), but PCR of the tissues was negative.Uninfected dogs remained PCR negative throughout the study.Overall, among the tissues tested post-inoculation until the completion of the study for the 14 infected dogs (a total of 208 PCR-examined samples, irrespective of the PCR assay applied), a positive PCR outcome was documented in 14/70 (20%), 30/70 (42.9%) and 17/68 (25%) blood, BM and spleen aspirates, respectively.The 16S rDNA sequences of the PCR amplicons derived from both PCR assays presented 100% homology with E. canis strain 611.

Discussion
Previous reports indicated that rifampicin may be an alternative chemotherapeutic agent to doxycycline for the treatment of CME. 13,24In a study on antibiotic susceptibilities, rifampicin demonstrated an anti-E.canis activity comparable to that of doxycycline, with a very low MIC (0.00003 -0.00006 mg/L). 24hen rifampicin was given (15 mg/kg/12 h orally for 7 days) to two subclinically infected, moderately pancytopenic dogs, pancytopenia was resolved and E. canis DNA was cleared from the blood, as documented by PCR. 13The same rifampicin regimen, given to two asymptomatic E. canis-infected dogs 700 days post-inoculation, after an ineffective course with doxycycline, appeared to clear the infection in one of the two dogs, based on xenodiagnosis with ticks. 9To the best of our knowledge, this is the first study evaluating the efficacy of rifampicin in acute CME in a controlled clinical trial, expanding the range of PCR-tested tissues to include BM and spleen, in addition to blood.As administered in the current study, rifampicin objectively hastened resolution of thrombocytopenia, but was inconsistent in eliminating the acute E. canis infection.
The rifampicin regimen applied in the current study differed substantially from that previously reported in CME 9,13 in terms of dosing and duration.The high bioavailability of rifampicin in the dog when given orally, once daily, at 10 mg/kg, 16 coupled with our preliminary pharmacokinetic study, the very low MIC for E. canis 24 and the fact that higher doses are more likely to be associated with side effects, 25 heavily affected the dosing scheme in this study; on the other hand, the failure of a dog to be cleared of the E. canis infection when treated with rifampicin for 1 week dictated towards a longer treatment duration. 9owever, even with the present regimen, clearance of the infection was only inconsistently achieved (2/5 dogs).
Similar to other studies, 6,9,13 all dogs experienced typical CME-associated clinical manifestations post-inoculation that had abated irrespective of treatment status prior to the institution of rifampicin (PID 21).It appeared, therefore, particularly challenging to substantiate a rifampicin-mediated clinical improvement in the acute phase of CME, which frequently demonstrates a rapid spontaneous clinical recovery. 1Similarly, although anaemia and leucopenia occurred in several of the 14 infected dogs by PID 14 (12 and 13 dogs, respectively), only a few dogs remained anaemic and leucopenic 1 week later, which also hindered their evaluation as treatment response indicators.In contrast, all the infected dogs remained persistently thrombocytopenic until initiation of rifampicin.The time for normalization of platelet counts was significantly shorter in the rifampicintreated compared with the untreated dogs, and the median platelet counts of the former dogs were significantly higher on PID 42 (the end of rifampicin treatment) and PID 98 (the end of the study), implying that rifampicin affords haematological recovery in acute CME, similar to what was previously documented in two subclinically infected dogs. 13lthough the results of several studies have suggested that the sustained post-treatment resolution of thrombocytopenia is correlated reasonably well with the elimination of E. canis infection 3 -6,26 this was not supported by the results of this study.At the end of the study, the three treated PCR-positive dogs had normal platelet counts, an experience shared by other authors, 7,9 implying that normal platelet counts, even long after treatment, do not invariably indicate clearance of E. canis infection.The high percentage of infected untreated dogs (4/6) that experienced normal platelet counts at the end of the study despite being carriers of E. canis further underlines the suboptimal performance of platelet counts as an infection clearance indicator.All infected dogs remained seropositive during the whole study period, and no difference was found in the median anti-E.canis titres between the treated and untreated dogs on PIDs 21, 42 and 98, suggesting that, similar to platelet counts, serology may be of limited value as a post-treatment monitoring tool.Theodorou et al.
In the present study, three untreated dogs experienced spontaneous clearance of infection, as evidenced from the negative PCR assays in all three tissues tested, similar to what has been documented previously in 25% -50% of subclinically or chronically infected dogs; this supports the possibility of self-elimination of E. canis infection in nature irrespective of the clinical phase of Rifampicin treatment in canine monocytic ehrlichiosis 1623 JAC the disease. 3,4,27Alternatively, the negative PCR status of the treated and untreated dogs in the present study may be due to the suppression of ehrlichiaemia and tissue bacterial loads to a level not detectable by the applied assays.This possibility was recently clearly demonstrated in E. canis-infected dogs treated with doxycycline, where xenodiagnosis was superior to PCR in demonstrating ehrlichiaemia of low level. 9The two PCR-negative untreated dogs that remained thrombocytopenic on PID 98 (one with splenomegaly) may support the possibility of persistent infection.
The results of the present study emphasize the importance of assessing the response to treatment in CME by applying molecular testing in a range of tissues rather than solely the blood, and for a sufficient time after cessation of treatment.Upon completion of rifampicin, two treated dogs were still PCR positive, but this number increased to four and three dogs, 4 and 8 weeks post-treatment, respectively, presumably due to a dexamethasone-induced recrudescence of latent E. canis infection.Of note, one treated dog that was found to be positive at 4 weeks post-treatment became negative at 8 weeks post-treatment, possibly indicating that its final clearance may have not been mediated by rifampicin.Surprisingly, 7/9 untreated dogs were PCR negative on PID 42, but this number diminished to 3/9 by the end of the study.These data suggest that the molecular assessment of infection clearance in CME should be extended for at least 2 months post-treatment.
In the current study, BM aspirates had the highest sensitivity for determining the clearance status of the infection by PCR at the end of the study.Surprisingly, all treated and untreated dogs were blood PCR negative at the end of the study, suggesting that blood should not be the sole tissue PCR-tested posttreatment, as it may lead to an overestimation of the clearance rate.Previously, spleen aspirates had been shown to be of higher sensitivity than blood and BM in assessing clearance of infection, 6,27,28 while in another study, blood, BM and spleen were of equal value in post-treatment PCR testing. 4Collectively, these data suggest that blood, BM and spleen may be a reasonable combination of materials for the PCR-based assessment of E. canis infection clearance in the clinical setting.Theodorou et al.
In the present study, we cannot exclude the possibility that failure of rifampicin to eradicate E. canis infection was associated with an insufficient duration of treatment or, most importantly, with the emergence of resistance to rifampicin.In a recent study on Staphylococcus pseudintermedius isolates from dogs, rifampicin treatment resulted rapidly in rpoB gene mutations that conferred resistance to rifampicin. 29Future studies may therefore be warranted to investigate whether the efficacy of rifampicin could be improved when given for a longer period of time, or whether rpoB gene-mediated resistance emerges in rifampicintreated E. canis infection.In the latter case, the combination of rifampicin with another drug with known in vitro efficacy against E. canis might be able to diminish this resistance potential.In this respect, doxycycline could be paired with rifampicin, similar to what has been suggested for feline and canine bartonellosis and Brucella canis-induced endophthalmitis in the dog. 30 -33Only a few other drugs have been critically evaluated for the treatment of CME.Imidocarb dipropionate and fluoroquinolones, initially thought to be efficacious in achieving clinical remission, 34 -36 were subsequently found to be ineffective in eliminating acute E. canis infection. 26,37n conclusion, rifampicin was partially effective in eliminating acute E. canis infection, but it significantly hastened the resolution of thrombocytopenia.It may therefore be considered as an alternative antibacterial treatment in CME, especially in dogs intolerant to doxycycline.The importance of assessing the response to treatment in CME by performing molecular testing in a range of tissues including blood, BM and spleen, at least 2 months after the end of treatment, is also emphasized.Platelet counts and anti-E.canis IgG antibody titres were of limited value in the evaluation of the post-treatment clearance of E. canis infection.

Figure 1 .
Figure 1.Median haematocrit, leucocyte and platelet values prior to and post-inoculation (PI) with E. canis in 14 dogs that were treated or not treated with rifampicin; uninfected dogs served as controls.I, inoculation day; Tx start, start of rifampicin treatment; Tx end, end of rifampicin treatment.*Significantly higher median platelet counts.

Figure 2 .
Figure 2. Kaplan-Meier survival curves showing the cumulative probability of resolution of thrombocytopenia with time (PIDs 21-42) in five rifampicin-treated and nine untreated E. canis-infected dogs. 1

Table 1 .
Clinical score in 14 dogs experimentally infected with E. canis treated or not treated with rifampicin

Table 2 .
Nested and real-time PCR results for blood, BM and spleen before and after inoculation of 14 dogs with E. canis; treated dogs were given rifampicin, untreated dogs received no treatment and uninfected dogs served as controls DPI, days post-inoculation; ND, not done.