METHODS

4.1 ETHICAL CONSIDERATIONS

The projects included in this thesis include in vitro and in vivo experiments, as well as an epidemiological study. All studies including animals or patients were approved by the regional ethics committee in Stockholm. By conducting studies that translate pre-clinical findings into an epidemiological study (from bench to “bedside”), and envisioning in the future to utilize epidemiological data to design pre-clinical studies (from bedside to bench), we believe that the results from our experiments will benefit patient care and that these benefits exceed any discomfort that our studies might cause to human or animal subjects.

Performing in vitro experiments with commercially available cell lines does not require ethical approval and provide us with the tools to study host-bacterial interactions in detail and with high control (compared to clinical research or in vivo experiments). We have developed and utilized in vitro experimental setups that mimic the environment in the renal tubule and expand beyond static cell culture setups, without having to use in vivo models.

The use of UPEC pyelonephritis strains originally isolated from patients, such as CFT073⁶⁸, increases the applicability of pre-clinical studies of host-pathogen interactions.

Despite the advances in in vitro cell culture models, these models cannot currently incorporate all interplaying factors. Therefore, we must also conduct in vivo experiments to fully understand host-pathogen interactions. Our in vivo Tissue Microbiology model of UPEC kidney infection is designed to reduce the number of animals used in one study. As we can study the progression of infection in the same animal at different time points, instead of sacrificing several animals at each selected timepoint, we use up to 70% less animals than conventional methodology. In some cases, when only local responses are studied, one animal can serve as its own control, since we can infuse UPEC into one tubule and a PBS solution into another tubule in the same kidney. Further, we always base our experimental designs on earlier results, so that no unnecessary animal experiments are carried out.

Whenever it is possible, we try to replace animal experiments with in vitro experiments.

However, when animal studies are crucial to study certain pathophysiological changes at the site of infection, our aim has always been to refine the experiments and induce as little discomfort as possible. In our microinfusion kidney infection model the animal is anesthetized throughout the whole infection experiment and sacrificed at end point while still under anesthesia. In our ascending UTI model the animals are anesthetized during the procedure of transurethral introduction of UPEC into the bladder, and during the 4 days of infection progress they are monitored daily for any signs of discomfort, in which case they are sacrificed immediately.

With regards to the epidemiological study included in this thesis, the study was completely retrospective, meaning that the researchers have not influenced the diagnostics performed or treatments used. Data collected from registers and patient medical records were anonymized

immediately and handled with coded index numbers instead of the patients’ personal identity numbers. Data collection was performed through an automated search, and only prespecified data were extracted from each medical record. Although we did not obtain formal consent from each patient, the benefits of the work outweighed any risks according to the ethical committee approval.

4.2 IN VITRO METHODS TO STUDY UPEC KIDNEY INFECTION

UTIs can be studied in in vitro cell culture models where monocultures of epithelial (bladder or renal) cells are stimulated with bacteria or bacterial compounds^54,82. Such in vitro studies have helped scientists map important details about bacterial pathogenesis and pathophysiological changes during UTIs, including bacterial attachment, internalization, and intracellular survival²⁰⁷, as well as host immune responses^82,208. In order to mimic the more complex environment during infection, trans-well models that incorporate several host cell types²⁰⁹, as well as models of ex vivo organ culture have been developed⁹¹. Experiments utilizing such models have ascribed major roles to cytokines and chemokines in immune cell recruitment⁹¹. More recently, novel in vitro biomimetic platforms, the so called organ-on-a-chip, have been developed²¹⁰. In these platforms physiological parameters, such as the shear stress of the urine flow, can be introduced in infection experiments²¹¹.

In this thesis we have utilized numerous in vitro model systems to study details of host-pathogen interactions during UPEC kidney infection. In Paper I we utilize a biomimetic flow model to mimic the flow conditions in the renal tubule. For this model we worked with the human renal carcinoma (epithelial) cell line A-498 (ATCC® HTB-44). One day prior to infection experiments, cells were seeded into the biomimetic flow chamber, and allowed to attach before the start of the experiment. On the day of experiment the flow chamber was connected to a peristaltic pump, which allowed media to flow over the attached cells.

Infection was induced by slowly injecting UPEC upstream of the cells. Other chemicals (e.g.

IFNγ) could also be added to the media at specific timepoints. The progression of the infection and the status of the host cells could then be followed in real time with bright field or fluorescence microscopy, and flow through could be collected at certain timepoints for later analyses. These analyses included hemolysis assays, enzyme-linked immunosorbent assays (ELISA), ATP determination assays and Luminex assays.

In Paper II we utilized a static UPEC kidney infection model. We chose to continue with A-498 cells to enable comparison with our results in Paper I, but now designed the experiments so that the cells were infected under static conditions instead of infection under flow. The static condition was chosen as we in this paper aimed to compare the immune responses of renal epithelial cells to those of primary mouse dorsal root ganglia (DRG) cells that are cultured and infected under static conditions. At end point, supernatants were collected, and responses were analyzed through ELISA.

In paper III we used a controlled 2-step in vitro cell culture model, where we first infected human renal epithelial cells, and then used the filtered supernatant to stimulate human

endothelial HMEC-1 (ATCC® CRL-3243) cells to investigate signaling mechanisms that might induce infection-mediated coagulation. In this study we aimed to come closer to the kidney microenvironment we study in the in vivo model and used RPTEC/TERT1 cells (ATCC® CRL-4031), a human renal epithelial cell line that has been shown to resemble primary renal proximal tubule cells^212,213. At end point of single experiments, supernatants were collected, and the epithelial and endothelial responses were analyzed through ELISA, Cytometric Bead Array or proteome array. To measure tissue factor procoagulant activity, stimulated human endothelial cells were lysed, whereafter tissue factor was extracted and finally used in chromogenic activity assays.

Primary mouse DRG cells were used in Papers I and II to study how sensory nerves might sense and respond to bacterial infection. DRG cells were extracted from BALB/c mice (8-12 weeks of age) by our collaborators as previously described²¹⁴. These cells were cultured and infected/stimulated under static conditions. Supernatants were collected at end point to analyze immune (IL-6) or neural (CGRP) responses with ELISA. For experiments investigating DRG CGRP release, capsaicin was used for positive controls. We chose to measure CGRP release as an indicator of sensory nerve activation as cultured rodent DRG cells have been found to release CGRP secondary to increases in intracellular Ca²⁺ levels, an indication that the sensory nerve cell is activated^215,216.

4.3 IN VIVO METHODS TO STUDY UPEC KIDNEY INFECTION

It is impossible to mimic the full complexity of an organ in cell culture, since infection in the mammal involves multiple systems, including the vascular system, the nervous system, the lymphatic system and the immune system. Further, a coherent host response to UPEC kidney infection also involves the integrated action of several organs^92,96. Thus, in vivo experimentation has become an important tool in Infection Biology. In the studies included in this thesis we have utilized two models of kidney infection: an ascending model of UTI and a Tissue Microbiology model of kidney infection (Figure 7).

In the ascending infection model of UTI, the bladders of female Sprague-Dawley rats were carefully catheterized and slowly instilled with 10⁸ CFU (colony forming units) of UPEC strains in 100 μl PBS, or PBS alone (uninfected control) while the animals were anesthetized.

The infection was allowed to progress for 4 days, after which the rats were sacrificed, and tissues collected for tissue analyses. While this model closely mimics the natural route of UTI infection²¹⁷, this model does not allow us to know exactly if and when the bacteria reach the kidney. Further, we do not know exactly where the infection is localized in the kidney.

Thus, it is hard to investigate early (first hours of infection) host-pathogen interactions with this model.

To overcome the shortcomings of the ascending UTI model, we employed our Tissue Microbiology kidney infection model that enables spatio-temporal control⁶. Male Sprague-Dawley rats were anaesthetized, and surgically prepared prior to infection. For all animals this preparation included a tracheotomy, cannulation of the femoral vein for infusion of fluids

or removal of blood samples, and cannulation of the left ureter for urine sampling and to prevent bacteria from reaching the bladder. In certain cases, the animals were also subjected to either splenectomy or pharmacological blocking (with bupivacaine) of the splenic nerve prior to infection. In all animals, the left kidney was exposed via a subcostal flank incision, freed from surrounding fat, and supported in a kidney cup. Induction of infection was performed by microinfusion of approximately 4 x 10⁴ CFU UPEC mixed in a PBS solution containing Fast Green dye (in order to visualize the infused tubule under a stereoscopic microscope) and a fluorophore-conjugated dextran designed to be taken up by proximal tubule cells (in order to visualize the infused tubule during multiphoton microscopy and in ex vivo immunofluorescent analyses) into single proximal tubules. The microinfusion was performed under stereoscopic microscope observation (100x), and the infection could thereafter be visualized through intravital multiphoton microscopy to follow the infection progress in real time. Blood flow were in certain cases visualized by intravenous injection of a fluorophore-conjugated dextran designed to stay in the blood stream and not be filtrated through the glomeruli. At end point, animals were sacrificed, and tissues collected for tissue analyses.

Tissue analyses based on tissues collected from the in vivo experiments included CFU counts of plated blood, urine and homogenized organs, as well as immunofluorescence analysis of fixed renal or splenic tissue, quantitative real-time PCR of splenic tissue, and ELISA analysis of homogenized tissue or plasma samples.

Figure 7. In vivo rodent kidney infection models. (A) Ascending model of kidney infection, where bacteria are instilled transurethrally into the bladder of anesthetized rats. At end point (4 days) some bacteria may have ascended to one or both kidneys. The exact timepoint of bacterial entry or location in the kidneys is not known.

(B) Tissue Microbiology model of kidney infection, where bacteria are (i) infused into superficial proximal tubules of a kidney. This enables spatio-temporal control of the kidney infection. Intratubular infusion is confirmed under a stereoscopic microscope by co-infusing Fast Green dye. (ii) Infection progress is followed via intravital multiphoton microscopy. Top: PBS infused tubule. Bottom: UPEC Gfp+ (green) infused tubule.

Proximal renal epithelial cells are distinguished by co-infusion of fluorophore-conjugated dextran (red). Scale bars = 100 μm. Created with BioRender.com and adapted from Steiner et. al., 2021.

4.4 BACTERIAL STRAINS

All bacterial strains used in this thesis originate from the clinical pyelonephritis isolate CFT073 (Table 1). Isogenic mutant strains have been constructed to decipher the role of different UPEC virulence factors. To enable intravital imaging, mutant strains of CFT073 expressing GFP+ have been used.

Table 1. Bacterial strains used in this thesis Strain Genotypes and

serotypes

GFP+ HlyA+ LPS Reference

CFT073 O6:K2:H1 - + O6 antigen,

hexa-acylated lipid A

Mobley et al. (1990)⁶⁸; Welch et al. (2002)²¹⁸

LT002 CFT073 hlyA::km^R - - O6 antigen,

hexa-acylated lipid A

Månsson et al. (2007)⁶ LT004 CFT073 cobS::Φ(PLtetO-1

-gfp⁺), cm^R

+ + O6 antigen,

hexa-acylated lipid A

Månsson et al. (2007)⁶ LT005 LT002 cobS::Φ(PLtetO-1

-gfp⁺), cm^R, km^R

+ - O6 antigen,

hexa-acylated lipid A

Månsson et al. (2007)⁶ LT004

ΔmsbB

LT004 msbB::km^R + + O6 antigen,

penta-acylated lipid A

Schulz et al. (2018)²¹⁹

4.5 STUDYING ACUTE PYELONEPHRITIS IN HUMANS

While a lot of information can be gathered from in vitro and in vivo studies, the ultimate application is to prevent and treat kidney infections in humans. Epidemiological studies can provide insight into relationships between exposures and outcomes during UTIs. To study the association between antithrombotic therapy and risk of bacteremia or acute kidney injury in humans, we therefore conducted a retrospective cohort study collecting data from the laboratory information management system kept by Karolinska University Laboratory as well as the electronic medical record system at Karolinska University Hospital in Stockholm, Sweden, through automated searches. The study was based on all inpatients with acute pyelonephritis (defined as growth of bacteria in urine combined with body temperature

≥38^oC, or a diagnosis of pyelonephritis), and where at least one blood culture had been drawn within one day of the urine culture. Data collected included microbiological analyses, data on demographics (age and sex), antithrombotic use, vital signs (heart rate, body temperature, blood pressures, peripheral oxygen saturation, respiratory rate, and Glasgow Coma Scale), objective physical findings (body mass index [BMI]), laboratory findings, information regarding the hospital visit (ward, duration of hospital stay, reason for hospitalization), and comorbidities.

The exposure of interest was antithrombotic treatment, defined as continuous use of drugs with Anatomic Therapeutic Chemical (ATC) classification codes starting with “B01” prior to the acute pyelonephritis episode and at least until the day of the positive urine culture. The main outcome was bacteremia, defined as at least one positive blood culture drawn within one day of the urine culture, with growth of the same bacterial strain in blood and urine. In secondary analyses, we used acute kidney injury as outcome. This was defined as either a discharge diagnosis of acute kidney injury, or as an increase of serum creatinine by >26 µmol/L, or a relative increase of serum creatinine of ≥50% within 48 h from baseline at the

time of the urine culture, according to Acute Kidney Injury Network classification stage 1 or higher²²⁰.

To investigate the association between antithrombotic treatment and both bacteremia and acute kidney injury, multiple logistic regression was used to calculate odds ratios (OR) and confidence intervals (CIs). In addition to terms for antithrombotic treatment, the logistic regression models included terms for age, sex, BMI, and the following comorbidities:

malignancy, diabetes, hypertension, chronic renal failure, chronic heart failure, chronic liver disease, cardiovascular disease, thromboembolism, and coagulopathy. Variables were chosen a priori based on clinical judgment, and not statistical significance, since statistical association between a covariate and the outcome depends on the size of the effect and the sample size for the study. Thus, relevant covariates may not be statistically significant if the sample size is not large enough to provide adequate power to detect the association.