Tissue distribution and cell tropism of Brucella canis in naturally infected canine foetuses and neonates

Brucella canis infection is an underdiagnosed zoonotic disease. Knowledge about perinatal brucellosis in dogs is extremely limited, although foetuses and neonates are under risk of infection due to vertical transmission. In this study, immunohistochemistry was used to determine tissue distribution and cell tropism of B. canis in canine foetuses and neonates. Diagnosis of B. canis in tissues of naturally infected pups was based on PCR and sequencing of amplicons, bacterial isolation, and immunohistochemistry, whose specificity was confirmed by laser capture microdissection. PCR positivity among 200 puppies was 21%, and nine isolates of B. canis were obtained. Tissues from 13 PCR-positive puppies (4 stillborn and 9 neonates) presented widespread immunolabeling. Stomach, intestines, kidney, nervous system, and umbilicus were positive in all animals tested. Other frequently infected organs included the liver (92%), lungs (85%), lymph nodes (69%), and spleen (62%). Immunolabeled coccobacilli occurred mostly in macrophages, but they were also observed in erythrocytes, epithelial cells of gastrointestinal mucosa, renal tubules, epidermis, adipocytes, choroid plexus, ependyma, neuroblasts, blood vessels endothelium, muscle cells, and in the intestinal lumen. These results largely expand our knowledge about perinatal brucellosis in the dog, clearly demonstrating a pantropic distribution of B. canis in naturally infected foetuses and neonates.

Pantropic distribution of Brucella canis in foetal and neonatal canine tissues. Immunolabeled coccobacilli were widely distributed in several organs and tissues from all 13 Brucella PCR-positive puppies ( Table 1). Organs with the highest frequency of positivity (100%) included the stomach, intestine, kidney, and umbilicus, which contained numerous intracytoplasmic immunolabeled coccobacilli located mostly within macrophages (Fig. 1). Macrophages were also positive in the hepatic parenchyma as well as in portal spaces and around central lobular veins (12/13), in lungs (11/13), and heart (11/13). Reproductive organs, including the gonads, uterus and prostate, although pre-pubertal and in a very young age, were frequently positive (8/11), with immunolabeling mostly in macrophages, in the albuginea (Fig. 1), prostate, myometrium, uterine broad ligament. Spleen and lymph nodes, positive in 62-69% of the puppies, most of the time presented low intensity (−or : Not available to be tested. + = only one positive mark in ten fields (400X) or more; ++ = at least one positive mark in every two fields; +++ = more than two positive marks in more than ten fields. +, 10/13; − or +11/13), respectively. In these tissues, positive macrophages often had perivascular localization ( Fig. 1), and infiltrated the wall of arteries (pulmonary, umbilical, coronary, and hepatic). There was also positive coccoid staining in intravascular erythrocytes (7/13) in different organs, including the lungs, genital organs, kidneys, liver, heart chambers, and cerebrum (Fig. 2). In addition to macrophages and erythrocytes, positivity was also observed in other cell types in different organs, as in superficial epithelium of gastric mucosa, in enterocytes, and in goblet cells in the small and large intestines (Fig. 3). Meconium was positive in 75% of stillborn puppies while only 33% of neonates had immunolabeled coccobacilli in the intestinal luminal content, with a lower intensity of immunolabeling (+, 3/9). The gastrointestinal mucosa was intensely positive in all puppies, although immunolabeled coccobacilli were observed mostly intracytoplasmically in macrophages diffusely distributed in the lamina propria, submucosa, muscular layer, mesentery and gastro-splenic ligament, in stillborn and neonates. Only two puppies had low intensity of immunolabeling in stomach and intestine: one premature stillborn and one four-day-old premature neonate who did not nurse in the bitch and was fed artificial milk formula by orogastric tube due to palatoschisis. In five puppies, immature adipocytes in the mesenteric fat contained immunolabeled coccobacilli in the cytoplasm (Fig. 3).
Occasionally, intracytoplasmic immunolabeled coccobacilli were present in smooth muscle cells in umbilical, pulmonary and coronary arteries, where few or no macrophages and neutrophils were present (Fig. 4). The same finding was observed in myocardium cells, with variable number of coccobacilli intracytoplasmically in the absence of inflammation or associated with rare interstitial macrophages and neutrophils (Fig. 3).
In the umbilical cord, immunostained coccobacilli occurred on the epidermis in the stratum corneum ( Fig. 4) in stillborn and neonates (+ to +++), and also intracytoplasmically in the apocrine gland epithelial cells (++) in only one sample, and in adipocytes in subcutaneous fat (+, 3/10), besides the positivity in macrophages and in arterial smooth muscle cells already mentioned above.
There were only two placentas available for this study, and they were both positive by immunohistochemistry, with intracellularly immunolabeled B. canis predominantly in macrophages located in the foetal mesenchyme, and occasionally within trophoblasts (Fig. 5). Although, immunolabeling was observed in macrophages, there was no evidence of inflammatory reaction in haematoxylin and eosin-stained sections (Fig. 5), and those cells were interpreted as resident macrophages of the placenta.
Considering the originality of these findings, immunolabeled epithelial cells of renal tubules were microdissected by using a laser capture microdissection apparatus. DNA extracted from these immunolabeled microdissected immunolabeled renal tubular epithelia was PCR-positive for Brucella sp., further confirming the specificity of immunolabeling.

Discussion
This study largely expanded our knowledge regarding the tissue and cell tropism of B. canis in canine foetuses and neonates, clearly demonstrating a previously unknown pantropic distribution of this pathogen in several organs and tissues. This is the first thorough description of B. canis tissue distribution and cell tropism in its preferential host, and therefore this study contributed to a better understanding of canine brucellosis.
Abundant immunohistochemistry positivity in the gastrointestinal mucosa and in intestinal contents suggested that the oral route of infection may play an important role in canine perinatal brucellosis, possibly through ingestion of bacteria in suspension in the amniotic fluid. Carmichael and Kenney 26 hypothesized that foetal infection could occur through the oral route, based on their findings of high bacterial loads in the amniotic fluid and mild inflammation observed in the digestive system of canine foetuses. Furthermore, high positivity in the gastrointestinal mucosa was also observed in this study in one neonate at 12 days after birth, which may be due to transmammary transmission. Importantly, B. abortus endocytosis by M cells in the Peyer's patches had been described after oral experimental infection of calves 27 and in mice 28 , but the pathogen has not been observed within enterocytes and goblet cells. While the gastrointestinal mucosa and mesentery were highly positive in all puppies in this study, mesenteric lymph nodes were negative in most of the puppies, which is possibly due to the early stages of infection in these cases.
The ability of B. canis to infect renal tubular epithelial cells strongly suggested a previously unsuspected route of excretion of the organism. Transmission of B. canis has been attributed to prolonged contact with urine of an infected dog, but it was assumed that the organism originated from prostatic secretions of adult males 29 . Although the prostate gland in puppies was also positive, immunolabeled coccobacilli were located in interstitial macrophages, while in renal tubules they were in epithelial cells and in the lumen, indicating a source for urine contamination in the kidney itself. Zoonotic transmission to humans in close contact with young puppies has been documented 5,8 , and based on our findings, it may have occurred due to shedding of the organism in the urine, since in these cases abortion or vaginal secretions that are considered the conventional means of zoonotic transmission were absent 23 .
This study has demonstrated for first time that B. canis is associated with erythrocytes in naturally infected canine foetuses and neonates. Canine brucellosis is recognized as a disease with intermittent cell-associated bacteremia, presumably intracytoplasmically in leucocytes 2,23,26 . B. melitensis is capable of infecting erythrocytes of experimentally infected mice, in which bacterial cells were demonstrated either epi-or intracellularly in erythrocytes 30 . However, association of Brucella spp. with erythrocytes has not been previously demonstrated in naturally infected animals. Distribution in tissues and frequency of positivity in erythrocytes was variable, a finding that is consistent with the intermittent detection of B. canis in the blood of infected dogs 21 .
B. canis infection in myocardial cells and arterial smooth muscle cells had not been documented before. It may explain the frequent finding of myocarditis and arteritis associated with this disease 26 . In humans, arterial aneurysm has been diagnosed in a teenager infected with B. canis 4 , what may have occurred due to the direct infection of the smooth muscle cells. This study also demonstrated that B. canis are often associated with cells of the choroid plexus. Importantly, neurobrucellosis is a highly relevant clinical presentation in different host species 31 . B. abortus has been described infecting neuroblasts and glial cells 32 , but, in dogs, encephalic neurological signs are usually not attributed to B. canis infection. Based on our findings, research on neurobrucellosis in dogs should be encouraged, and perhaps this agent should be included among the differentials in canine neurological patients.
Interestingly, immunostaining was observed in genital organs of canine foetuses and neonates. B. canis tropism to the genital tract in males has been well documented in adult dogs 33 , but it has not been documented in prepubertal animals. The early infection in epididymis and testicles may compromise irreversibly the reproductive potential of the dogs. Interestingly, in the uterus, positivity was observed in myometrium, and seldom in endometrium. Brucella spp. are known to have tropism for the placenta 34 , although there are clearly differences in terms of clinical and pathological manifestations according to combination of Brucella and host species 35 . Interestingly, the two placentas available for this study had B. canis infection as demonstrated by bacterial isolation and immunohistochemistry, but no inflammatory reaction was observed histologically. This contrasts with the severe necrotizing and neutrophilic placentitis observed in other species, particularly in B. abortus-infected pregnant cows 34 . However, these observations require additional studies since only two samples are not sufficient for a definitive conclusion.
Ocular lesions have been reported in adult dogs infected with B. canis 17 . Therefore, immunolabeling in the cornea and retina in neonates is not unexpected, but it indicated that infection and ocular disease may occur in perinatal brucellosis, even before the opening of the eyelid.
B. canis infection of adipocytes has not been previously described, although adipocytes are permissive cells to bacterial infection. Mycobacterium tuberculosis in human adipocytes has been demonstrated by IHQ and in situ PCR, especially in adipose tissue adjacent to kidneys, lymph nodes, stomach, heart and skin 36 , the same distribution observed for B. canis in this study. Adipose tissue may represent a site for persistence of Brucella infection in dogs, a localization that is suitable to evade treatment and immune response, as is the case for M. tuberculosis in humans 37 . In adipocytes, the endoplasmic reticulum is localized adjacent to the lipid vacuole 38 , the same location where the immunohistochemistry positive coccobacilli were visualized in this study. Furthermore, adipocytes in canine foetuses and neonates are immature and undergoing differentiation. In a previous in vitro study, pre-adipocytes expressed unfolded proteins during the differentiation process 39 a response also observed during Brucella intracellular replication 40 , which suggests that differentiating adipocytes may represent a viable niche for Brucella replication. The presence of B. canis in mature adipocytes remains to be investigated.
Apart from the infected macrophages, immunolabeling was not associated with other inflammatory cells or significant inflammation in tissues, what seems to be a feature of B. canis, since naturally infected dogs do not develop signs of sepsis 26 , and experimentally infected animals develop only mild inflammation 14 . The low inflammatory tissue reaction, observed in this study, may contribute to the prolonged unapparent infection and difficulty to eradicate the organism, which is typical of canine brucellosis 2,18,20 .
The data generated in this study provided a vast body of novel information that expands our comprehension of canine brucellosis, and provided the basis for new hypothesis regarding transmission of infection to humans or other dogs, through urine, faeces and skin of canine neonates. These hypotheses should be further investigated since it may be extremely valuable to establish new preventive measures when handling neonatal dogs.
In conclusion, B. canis is widespread in macrophages in foetal and neonatal tissues, and cell tropism is diverse, including enterocytes, renal tubular epithelia, adipocytes, myocardium, and choroid plexus cells. Gastrointestinal infection occurs in congenital brucellosis in dogs. Bacteria may be excreted in the faeces and urine of young puppies.

Methods
Experimental design. Canine foetuses and neonates of 19 breeds that died spontaneously in 32 breeding kennels were submitted for necropsy by the owners. Kennels were located in the State of Espirito Santo, Brazil. Puppies were collected from November 2014 until June 2016. Necropsy, sample collection and analysis were performed according to a protocol approved by the Ethics Committee of the Federal University of Minas Gerais (CEUA/UFMG), under protocol number 197/2014. All methods were performed in accordance with the relevant guidelines and regulations.
Immunohistochemistry. Paraffin embedded tissues from Brucella-positive puppies were selected for immunohistochemistry based on the absence of autolysis. The 13 puppies were included in this analysis: four stillborn and nine neonates, from eight different breeds, and nine breeding kennels.
Brucella sp. immunolabeling was performed in the following organs: lung, heart, liver, spleen, stomach, intestines, kidney, bladder, prostate, uterus, gonads, umbilical cord, encephalon, and eye. Paraffin embedded tissues were sectioned (4 μm thick) and processed for immunohistochemistry as described by Xavier et al. 34 , with modifications. Tissue sections were dewaxed, hydrated, and incubated three times in hydrogen peroxide 10% for 20 minutes, after each bath tissue sections were washed three times in PBS. To block unspecific reactions, tissues sections were incubated in 2.5% powdered milk suspension for one hour. The primary antibody was applied over the slides at the dilution of 1:1000, incubated in humidified chamber overnight and washed in PBS. Secondary biotinylated universal antibody with streptavidin-biotin complex (LSAB2 System-HRP, DAKO, USA) was added and slides were incubated for 1 hour and 40 minutes at room temperature. Reaction was revealed by using the chromogen AEC (DAKO, USA). Meyer's haematoxylin was used for counterstaining.
The primary polyclonal antibody employed has been previously described 44 . Reactions included positive and negative controls. Positive controls consisted of Brucella infected mouse tissues. Negative controls consisted of infected mouse and puppy tissues incubated overnight with PBS with suppression of the primary antibody in the reaction.
Tissue sections were analysed under light microscopy and the frequency of the immunostaining was estimated under high magnification field (400x) as follows: + = one positive mark in ten fields or more; ++ = one positive mark in every two fields; +++ = more than two positive marks in more than ten fields.
Laser capture microdissection. In order to confirm that immunohistochemically Brucella sp.-positive cells indeed contained Brucella sp., we performed laser capture microdissection using the MMI CellCut ® System.
A sample of renal tissue of one stillborn, positive by immunohistochemistry, was used. Paraffin-embedded tissue was mounted on the MMI slide membrane, processed for immunohistochemistry as described, and immunostained tubular epithelial cells were microdissected and subsequently the DNA was extracted using the Dneasy blood & tissue kit (Qiagen, Germany) according to the manufacturer's instructions. To investigate the integrity of the extracted DNA, polymerase chain reaction (PCR) was performed for the β-actin gene 45  Isolation and identification of Brucella canis. Frozen samples of heart, kidney, spleen and lung, from ten PCR and immunohistochemistry Brucella-positive puppies, from eight different kennels, were submitted in duplicates to bacterial isolation 46 . Samples of placenta were available from two puppies, and were included in this study. Tissue samples stored at −80 °C were thawed and homogenized with a scalpel blade in 500 µL of PBS. Then, 100 µL of the homogenate were inoculated in selective Tryptose agar (BD Difco, USA) with and without antibiotics (2,500 UI of polymyxin B; 12,500 UI bacitracin; 50,000 UI of nystatin; 50 mg of cycloheximide; 2.5 mg of nalidixic acid; and 10 mg of vancomycin -Sigma Aldrich, USA) and tryptose broth (BD Difco, USA). Plates were incubated at 37 °C and 5% CO 2 , checked every 48 hours for colony growth. After seven days of incubation, broth was plated on tryptose soy agar without antibiotics and these plates incubated in the same conditions for a maximum of 21 days, when, if no growth was detected, they were considered negative and discarded.
B. canis isolates were identified based on colony morphology, acriflavine test 47 , PCR for the bcsp31 gene 42 , and B. ovis-specific PCR 34 . For molecular analyses, bacterial colonies were collected from agar plates into 200 µL PBS and heat-killed (100 °C/1 hour) and DNA was extracted as described by Pitcher et al. 41 .
Isolates with positive amplification of the bcsp31 Brucella gene, negative in PCR for B. ovis, and agglutination in the acriflavine test were identified as B. canis. Data availability. The datasets generated during and/or analysed during the current study are available from the corresponding author on request to anyone who wishes to repeat our analyses or collaborate with us.