Staphylococcus aureus colonization in hemodialysis patients: a prospective 25 months observational study

Swabs were obtained from two trained and validated investigators from the nasal mucosa and the surrounding skin at the catheter entry site by using BBL™ CultureSwabs from BD. In nasal swabs, the cotton tip was inserted about 1 cm into one nostril and rotated 4 times along the nasal septum with slight pressure against it and with slight rotational movements. The same procedure was repeated with the same swab in the other nostril. For sampling at catheter entry sites, the swab end was drizzled with a sterile physiological saline solution and rotated on the skin around the catheter entry with slight pressure. Subsequently, the swab samples were tested for S. aureus using a semiquantitative procedure. S. aureus enrichment took place by a phenol red mannitol salt broth and quantification on mannitol salt agar (BD, Heidelberg, Germany) as previously reported [3]. CFUs were determined and categorized as < 10, 10–10², 10²–10³ and > 10³. From each S. aureus-positive patient sample, a single colony was picked for preparing the glycerol stock. DNA isolation was carried out using the DNeasy® Blood & Tissue Kit (250) from QUIAGEN (Qiagen, Venlo, The Netherlands) according to the manufacturer’s instructions, but with an addition of 0.2 mg/mL lysostaphin (Sigma-Aldrich, St. Louis, Missouri, USA) to the lysis buffer.

The spa region refers to a repeat region on the protein A gene. The repeats usually have a length of 24 base pairs and vary in number from 2 to 16. Based on order and type of repeats, S. aureus strains are classified by the Ridom software and entered into the Ridom SpaServer database ([15], Ridom SpaServer (http://​www.​spaserver.​ridom.​de/​). The amplification of the spa sequence was based on the protocol of ridom BIOINFORMATICS using the corresponding primers (http://​www.​ridom.​de/​staphtype/​spa_​sequencing.​shtml). The initial temperature for denaturation of the strands was raised from 80 °C to 94 °C and the number of cycles decreased from 35 to 30. Subsequently, 10 μl of amplified spa DNA were applied to a 1.5% agarose gel for control for unspecific bands. 5 μl of amplified spa DNA per sample were incubated with 2 μl of ExoSAP-IT® from Affymetrix for 15 min at 37 °C and for further 15 min at 80 °C. The amount of purified PCR product was then determined by the DS-11+ Spectrophotometer from DeNovix and checked qualitatively. Based on the shown fragment length on the gel, the purified PCR product was diluted to 1 ng/μl (fragments to 300 bp) or 5 ng/μl (fragments from 300 bp) according to the protocol provided by eurofins. Each 15 μl of this dilution was separately combined with 2 μl (10 μM) of the spa-forward primer and the spa-reverse primer and sent to eurofins for sequencing. High-quality sequences were then integrated into the Ridom SpaServer database and assigned to a specific spa type.

The genetic composition of the obtained S. aureus isolates was determined with the S. aureus Genotyping Kit 2.0 from Alere Technologies GmbH [16]. This array contains single strand DNA probes for various virulence and resistance genes. Hybridization with biotin-labelled bacterial DNA allows the simultaneous detection of numerous staphylococcal genes. PCR for the linear amplification, biotin labeling of the nucleotide sequences and the subsequent hybridization of the PCR product on the chip surface were performed using the protocol provided. Bound bacterial DNA was visualized by precipitation staining and analyzed using the ArrayMate Reader from Alere Technologies GmbH. The result files include a qualitative statement about the gene expression (positive/negative/ambiguous) as well as signal intensities of all sequences spotted on the chip. Since each lineage is defined by its core variable genome and linked to defined mobile genetic elements, the virulence and resistance gene patterns detected by the Staph array can be used to predict the lineage type. Hence, the software automatically predicts the CC and provides a list of common multi-locus sequence typing (MLST) and spa types [16].

In total, 86 patients from an outpatient dialysis center were included in the SaDial-study. 70% were male; the median age was 65 years and 59% were older than 60 years (Table 1). 38% were dialyzed at least temporarily respectively at the time of sampling via CVC. 40% of the patients were diabetics and 62% stayed at least once on an intensive care unit (ICU). Only 5% were living at a nursing home. Almost half of the patients were dialyzed for more than 5 years. During follow-up time, 22 patients died, 7 withdrew from the study for other reasons, 4 patients were transplanted, 2 moved away and in further 3 patients, a dialysis was no longer necessary.

During the study period, the colonization status of the dialysis patients was evaluated five times, every six months. The S. aureus carriage rate in the nasal mucosa varied cross-sectionally between 33 and 46%, depending on the sampling month (Fig. 1). Related to patients dialyzed via AF, the carrier rate was at 0.39, 0.46 for patients with a CVC access respectively.

In only five cases and three different patients, the bacterium could be detected at the catheter entry site (Additional file 1, ID 003, 005, 022). In February 2016, the S. aureus isolates from nose and CVC were identical in two cases. In the other three cases (two in February 2017, one in September 2017) the strains were unrelated.

Among all S. aureus-positive samples, only two MRSA were detected (ID 001, 072). The first patient died shortly before the following sampling took place due to respiratory failure with positive detection of P. aeruginosa in sputum. The second patient died three months after sampling due to a septic shock with unclear focus but possible CVC sepsis without pathogen detection.

Investigation of nasal colonization density revealed differences in the detected amount of colony forming units per milliliter (CFU/ml) of S. aureus in the nasal mucosa depending on the vascular access type (Fig. 2). While in CVC patients only in two cases a concentration of more than 1000 CFU/ml was detected, patients dialyzing via AV fistula were generally more densely populated with S. aureus (p-value = 0.045). In those five cases in which the bacterium was detected at the catheter entry site, the amount of CFUs was generally in the range of the amount of colony forming units detected in the nasal mucosa of the individual patient.

Longitudinally, 65% of hemodialysis patients showed a S. aureus colonization. A closer look at the population dynamics of the nasal colonization revealed that 14/86 patients (16%) were persistent S. aureus carriers, defined as being positive in at least 75% of samplings (Additional file 1). In 5 of these 14 patients the same strain was detected in the nasal mucosa during regular sampling. In the remaining 9 patients, the S. aureus strain changed to another unrelated strain. Another 37/86 (43%) patients were intermittent carriers with only one or two S. aureus-positive nose swabs during regular sampling. 5/86 patients (6%) could only be sampled once due to withdrawals, deaths or later study participation but showed a S. aureus colonization. The remaining 30/86 patients (35%) had no nasal S. aureus colonization and were thus classified as non-carriers.

The colonizing S. aureus population in hemodialysis patients was very diverse. The most common sequence type was the CC8 (22%) followed by CC45 (19%) and CC7 (14%) while the CC30 (11%) was only placed on position four (Fig. 3). The two positive MRSA nasal swabs were assigned to the sequence types CC5 and CC22. These, as well as the clonal complexes CC1, CC10, CC25, CC101, CC133 and CC182 could also be detected sporadically (Table 2). This distribution differs from the sex- and age-adjusted SHIP-TREND-0 general population (p = 0.001). The most common S. aureus strain here was the CC30 (21%) (twice as frequent compared to SaDial) followed by CC45 (16%), CC15 (12%), CC8 (10%) (half as frequent), CC25 (8%), CC22 (8%) and CC7 (5%) (third as frequent). All other sequence types matched in relative numbers in both studies.

A closer look at the spa types showed that the spa type t008 was highly prevalent in the dialysis cohort (19%, considering all positive samples) (Table 2). A comparison of all t008 spa types on genetic level by DNA microarray revealed that these isolates were not clonally identical. All other spa types were only occasionally detected.

We frequently detected one or more strain shifts in persistent carriers (9/14) (Additional file 1). In 24 cases, the primary and the subsequent isolate were closely related based on their spa type, suggesting the presence of a diversifying S. aureus population in the nose rather than a strain replacement. In 11 cases, the subsequent isolate belonged to a different spa type, implying a strain replacement. However, since only one colony from each patient was analyzed, we cannot exclude a nasal co-colonization with different S. aureus clones. Patient 005, for instance, was co-colonized with two different spa types (t593 (CC15) and t091 (CC7)) in February 2016 with one of the types also being detected at the CVC entry site.

Contrary to expectations, there were only six blood stream infections with S. aureus in the dialysis cohort within the 25 sampling months (Table 3). Affected patients showed the corresponding symptoms (leukocytosis, fever and systemic inflammatory response syndrome) and were tested positive for S. aureus in blood culture. Three of these infections were attributed to the CVC access. All patients but one survived the S. aureus-associated infection. There was no link between the infection and the S. aureus carrier status or the type of dialysis access. Four of these six patients were dialyzed by a CVC at the time of infection (ID 005, 018, 019, 028), while two patients were dialyzed via AV fistula continuously over all sampling months (ID 032, 034). This corresponds to a rate of 0.002 S. aureus infections per month for AF patients and 0.01 for CVC patients respectively.

During the 25-months study period 22/86 patients died (mortality: 26%). Main cause of death was a sepsis or septic shock (16/22 patients), but only in one case (ID 019) S. aureus could be confirmed as the causative pathogen (Table 4). This patient was a non-carrier during all samplings and died due to a S. aureus-induced infection of the central venous catheter with further involvement of gram-negative bacteria (Table 4). Seven septic patients had no clear evidence of S. aureus because the focus remained unclear or could only be limited to cocci in general. In another 8 septic patients, sepsis was induced by other pathogens (E. faecalis, S. epidermidis, E. coli, C. difficile, K. oxytoca, P. aeruginosa and Influenza B). Only one patient died from sepsis due to a proven CVC infection caused by Klebsiella oxytoca. Other five patients had a suspected CVC infection without pathogen detection. Only 6 patients overall did not die from sepsis but from either ischemic or embolic diseases (Table 4).

To investigate the impact of demographic data and patients´ medical histories from our dialysis cohort, we performed a multivariate analysis (Cox regression). As expected, advanced age was an independent factor positively associated with mortality (p = 0.005). Unexpectedly, S. aureus carriers had a 5-fold higher survival probability compared to non-carriers (HR, 0.2 for carriers to die [95% CI, 0.1–0.6]). Overall mortality considering the sub-cohort of S. aureus carriers was in turn only significantly influenced by patient age.

The survival probability in the hemodialysis study population over 25 months was 74% with a standard deviation of 3%. Figure 4 displays the Kaplan-Meier survival curves of all dialysis patients as well as for the sub-cohorts S. aureus carrier and non-carrier. In patients with at least one S. aureus positive nasal swab during all samplings, the overall survival probability raised to 86%. In contrast, the survival probability diminished to 55% in non-carriers.

The SaDial study revealed a cross-sectional S. aureus prevalence in hemodialysis patients between 33 and 46%, which is higher than the local general population (SHIP-TREND-0, mean 25%), but rather in the lower area compared to previous studies on hemodialysis patients (ranging from 30 to 84%) [8]. The strong variations in the carrier rate between different study cohorts are still common today and are likely influenced by the type of dialysis [17]. Among S. aureus carriers in our study, hemodialysis patients with an AV fistula had been on renal replacement therapy for longer time than those with CVC and had a higher S. aureus density at nasal swabs. Usually, patients with an AV fistula are those whose renal disease has a longer history. Thus, they had more chances to come into contact with hospital environment, medical staff and also with S. aureus. This way, both the likelihood of colonization and the colonization density increase.

The MRSA prevalence in the SaDial hemodialysis patients (2/86 patients; 2.3%) was well below the average MRSA prevalence as summarized in a meta-analysis of > 5.000 hemodialysis patients worldwide (7.2%, range 4.9–9.9%; [18]). Related to the same geographical region, the MRSA prevalence of the SaDial-study is above the normal population (0.34%) [3] and slightly above German health care facilities of northeastern Germany: 2.1% in homecare services, 1.4% in long-term care facilities, and 0.5% in rehabilitation clinics [19]. A comparative study by Neuhaus et al. in 2003 considering the MRSA prevalence in German hospitals and nursing homes showed no significant difference between both medical institutions in case of an active exchange of patients [20]. Similar to our outpatient dialysis center and the University Medicine Greifswald, this leads to an adjustment of MRSA prevalence to the latter one. However, established prevention strategies against MRSA spread, such as information exchange to the patient’s carrier status, the subsequent isolation of affected ones or eventual decolonization actions of MRSA carriers, prevent a spread of MRSA strains and show in a worldwide comparison a low MRSA prevalence in our outpatient dialysis center.

We compared the S. aureus population from our study cohort with data from a population-based study of the same geographical region of Western Pomerania (Study of Health in Pomerania, SHIP-TREND-0). The study investigated 3891 subjects aged between 29 and 79 and tested for nasal colonization with S. aureus. We adjusted the data according to sex and age. Clonal complexes occurring in patients requiring hemodialysis differ from those found in the general population with CC7 and CC8 being more prevalent in the patient cohort and CC30 being underrepresented (p = 0.001). In contrast, a study of dialysis patients from England in 2002 revealed a distribution of CC types corresponding rather to the general population from the SHIP-TREND-0 study [3]. The CC30 was the most prevalent one, followed by CC25, CC22 and CC8. Only the MLST type CC45 was underrepresented [21]. The different MLST distributions of SaDial and SHIP-TREND-0 are in agreement with the general observation that S. aureus colonization characteristics differ from the general population if patients are in frequent contact with hospitals, dialysis centers or homes for the elderly.

Each S. aureus lineage is equipped with a specific set of surface-attached adhesins, as well as toxins and immune evasion molecules; we tested whether CC7 and CC8 share some properties that might facilitate transmission and colonization in a hemodialysis setting, that are lacking in CC30 (Fig. 5). Interestingly, both CC7 and CC8 harbor the fibrinogen binding protein (efb/fib), which contributes to the initiation of foreign body infections [22]. In contrast, the collagen binding adhesion (can) mediating bacterial adherence to collagen substrates and collagenous tissues [23] is not expressed in CC8 and CC7 but in CC30. These genetic differences illustrate the strong influence of the medical environment on the S. aureus population and in consequence on the distribution of lineage-associated virulence factors.

After 24 months, the mortality rate in the hemodialysis study population was 26% (Fig. 3), with the overall mortality being lower for S. aureus carriers than non-carriers (HR 0.19, p = 0.005). The overall mortality is comparable to data from literature [24, 25]. However, the fact that S. aureus carriers seem to be protected needs to be studied more closely.

Protection from invasive S. aureus infections could be explained by the stronger pre-existing immune memory in carriers. In 2005, Wertheim et al. already showed that S. aureus carriers may have a decreased risk of death compared to non-carriers in case of S. aureus infection, hypothesizing that anti-staphylococcal antibody levels are increased in carriers, induced by prior S. aureus exposure and playing a role in protection [4]. By now, it is well known that anti- S. aureus antibody titers in carriers are higher than in non-carriers [26]. In 2015, van den Berg could strengthen this hypothesis in a mouse model by improving the course of subsequent endogenous S. aureus bacteremia by mild S. aureus skin infections. S. aureus skin infected mice showed IgG levels against various S. aureus antigens and had a higher survival probability [27]. Moreover, immune memory of S. aureus was associated with protection from serious complications of bacterial invasion [28]. The mortality rate of S. aureus carriers in our study population was 32% lower compared to non-carriers after 25 months. In conclusion, S. aureus colonization along with subclinical infections may prime the adaptive immune system and lead to a more robust immune memory protecting from fatal S. aureus sepsis.

Nevertheless, only one death in the SaDial study could clearly be attributed to a S. aureus sepsis. Indeed, the main causes of death in our hemodialysis study population were bloodstream infections by pathogens other than S. aureus (73%), followed by cardiovascular events (14%), and various causes (13%). Previous studies report that the main causes of death are comorbidities that occur alongside with or result from renal insufficiency [24]: cardiovascular events (40%), bloodstream infections (30%) and sudden deaths of patients at home with unknown cause (25%) [29, 30], the latter often again being attributed to cardiovascular events [28]. How S. aureus carriage could influence the overall mortality due to non-S. aureus infections and other, non-infectious diseases remains to be clarified.

Contrary to our assumption that S. aureus was the main cause of blood stream infections, mainly other pathogens such as E. faecalis, S. epidermidis, E. coli, C. difficile K. oxytoca and Influenza B were isolated from blood culture. According to a nationwide study in the USA by Wisplinghoff et al. in 2004 considering 24,000 infections from 10,000 hospitals, 27% were caused by coagulase negative staphylococci, 24% by S. aureus, 9% by Enterococcus species, 8% by Candida species, 8% by E. coli, 6% by Klebsiella species and others [31]. Mortality rate varied according to the respective bacteria between 13 and 29% (Candida species 29%, Enterococcus species 24%, S. aureus 19%, E.coli 17%) [31]. Thus, pathogens detected in our patients are among the common causes of blood stream infections. Future studies will use serodiagnostics, i.e. the induction of a species-specific antibody response, to clarify which pathogens were recognized and combated by the adaptive immune system. Hopefully, this will shed some light on the physiologically relevant invasive pathogens that might have been missed by standard microbiological diagnosis.