Bacterial DNAemia is associated with serum zonulin levels in older subjects

The increased presence of bacteria in blood is a plausible contributing factor in the development and progression of aging-associated diseases. In this context, we performed the quantification and the taxonomic profiling of the bacterial DNA in blood samples collected from forty-three older subjects enrolled in a nursing home. Quantitative PCR targeting the 16S rRNA gene revealed that all samples contained detectable amounts of bacterial DNA with a concentration that varied considerably between subjects. Correlation analyses revealed that the bacterial DNAemia (expressed as concentration of 16S rRNA gene copies in blood) significantly associated with the serum levels of zonulin, a marker of intestinal permeability. This result was confirmed by the analysis of a second set of blood samples collected from the same subjects. 16S rRNA gene profiling revealed that most of the bacterial DNA detected in blood was ascribable to the phylum Proteobacteria with a predominance of the genus Pseudomonas. Several control samples were also analyzed to assess the influence of contaminant bacterial DNA potentially originating from reagents and materials. The data reported here suggest that para-cellular permeability of epithelial (and, potentially, endothelial) cell layers may play an important role in bacterial migration into the bloodstream. Bacterial DNAemia is likely to impact on several aspects of host physiology and could underpin the development and prognosis of various diseases in older subjects.


P s e u d o m o n a s A r t h r o b a c t e r A c i n e t o b a c t e r E s c h e r i c h i a -S h i g e l l a P h y l l o b a c t e r i u m P a r a c o c c u s T e p i d i
Escherichia-Shigella Cluster_10 Arthrobacter Solirubrobacterales 67-14 genus Caulobacteraceae genus Cluster_364 Burkholderia Clostridium  Fig. S5. Abundance of 16S rRNA gene copies of taxonomic units detected in second set of blood samples (n=42) that significantly correlated with the serum levels of zonulin. ρ, Spearman's rank correlation coefficient; P, P value of the Kendall's rank correlation. Taxa that resulted significantly correlated with zonulin also from the analysis of the first set of blood samples are indicated in bold and red color.

Fig. S6. Correlations of the taxonomic units detected in blood (expressed as relative abundances) toward age, BMI, and metabolic and functional markers
determined in blood of the older subjects under study (n=43). This figure only includes taxa whose abundance significantly correlated with at least one parameter.

Technical issues concerning zonulin quantification
In this study, zonulin quantification in serum samples was carried out by means of the most commonly used commercial ELISA kit. Recently, the specificity of this and others ELISA assays has been questioned 2 and, consequently, it was suggested to interpret with caution data collected as direct assessment of intestinal permeability 3 . In this context, it is noteworthy that already Scheffer and colleagues 4 , previously identified through the use of the same kit a variety of proteins structurally related to zonulin (in particular properdin).
Consequently, the authors suggested that although the assay was not specific for pre-haptoglobin2 quantification, other members of permeability-regulating proteins belonging to the mannose-associated serine protease family could be determined 4 .

Technical issues concerning the detection and taxonomic profiling of bacteria DNA in blood
In a recent publication, circulating cell-free DNA isolated from human blood plasma was subjected to massive shotgun sequencing 5 ; more than half of the identified contigs had little or no homology with sequences in available databases and, interestingly, were assigned to hundreds of entirely novel microbial taxa. In our study, we did not find such a large presence of unknown microorganisms. Nonetheless, two main aspects distinguish the research by Kowarsky et al. from ours: (i) we performed 16S rRNA gene profiling and not shotgun metagenomic sequencing and (ii) we analyzed DNA isolated from whole blood and not plasma. This second aspect is particularly important considering the presence of bacterial DNA in blood cells such as erythrocytes and antigen-presenting cells 6,7 .
In this study, the bacterial DNA isolated from blood was taxonomically profiled through MiSeq sequencing of 16S rRNA gene amplicons. We presented above numerous similarities, both quantitatively (i.e., abundance of 16S rRNA gene copies) and qualitatively (i.e., detected taxa), between the results of our study ad what reported in several other studies available in literature. However, none of the papers we referenced above focused specifically on the evaluation of potential contaminant DNA, originating from any possible experimental step. The use of 16S rRNA gene profiling for the bacterial taxonomic characterization of low microbial biomass samples, such as blood, has been criticized as being at high risk of microbial contamination that may occur at any step of the protocol, from sample collection until sequencing 8,9 . In our study, we analyzed several control samples to assess the potential presence of contaminants in labware (e.g. vacutainer and EDTA tubes) and reagents (e.g. solutions used during extraction, library preparation, sequencing, and qPCR). According to qPCR experiments, we always detected in control samples a quantity of bacterial DNA much lower than that quantified in blood samples, suggesting the potential contaminants should not have significantly affected the taxonomic profiling of blood samples. However, the confirmation of a significant correlation between zonulin and 16S rRNA gene copies in blood (total and ascribed to Pseudomonas) also in the second set of blood samples investigated supports the conclusion that the bacterial DNA detected in blood largely do not derive from contamination. Nonetheless, it is also important to mention that most of the bacterial

Supplementary Material
13 genera detected in blood in our study have been reported as contaminants occurring during microbiome research in other studies (reviewed in 8 ).
Considering the relative abundance of bacterial taxa detected in blood and control samples, we hypothesize that the most probable contaminants belong to the families Enterobacteriaceae, Micrococcaceae and Moraxellaceae (the second, third and fifth most abundant families detected in blood, respectively), whereas at least most part of the DNA ascribed to Pseudomonadaceae (the most abundant family detected in blood) is less likely to derive from contaminants. Lists of bacterial taxa that were identified in negative controls during different independent studies have been proposed 8,10 , cataloging up to 70 different genera to be considered as potential contaminants 8 . These lists contain numerous Proteobacteria including Pseudomonas, which was found to be the most prevalent and abundant bacterial genus in the blood samples investigated in our study. Pseudomonas is a ubiquitous bacterium, which colonizes numerous environments, such as soil, water and various plant and animal organisms, due to minimal survival requirements and remarkable adaptation ability 11 . Notably, Pseudomonas is also one of the microorganisms most frequently isolated from patients with bacteremia, particularly the species P. aeruginosa 12 . In this report, the partial sequence of the 16S rRNA gene belonging to the most prevalent and abundant OTUs found in the analyzed blood samples (Cluster 1 and Cluster 3, Fig. 5) shared 100% similarity with P. fluorescens and other species of the same phylogenetic lineage. Although far less pathogenic than P. aeruginosa, P. fluorescens has been often reported as the aetiologic agent of opportunistic infections in lungs, mouth, stomach, urinary tract, skin, and, most commonly, blood 13,14 . Notably, P. fluorescens is recognized as the most important cause of iatrogenic sepsis, attributed to contaminated blood transfusion or contaminated equipment used in intravenous infusions [15][16][17] .
Although the literature evidence discussed above suggests that P. fluorescens and related species can be contaminants (see Supplementary Discussion in Additional file 1), on the other hand, these bacteria were also reported to possess numerous functional properties that support their survival and growth in mammalian hosts 14 . Furthermore, an interesting association was found between the presence of serum antibodies against the I2 peptide encoded by P. fluorescens and Crohn's disease 18 , celiac disease 19 , ankylosing spondylitis 20 , and chronic granulomatous disease 21 . In addition, P. fluorescens was reported to be regularly cultured from clinical samples even in the absence of acute infection 14 . Finally, P. fluorescens was demonstrated to induce zonulin expression and decreased intestinal permeability in a time dependent manner in an in vitro model of intestinal epithelium 22 . In the same study, the authors found increased zonulin levels and higher abundance of Pseudomonas 16S rRNA gene copies (as determined through qPCR with genus-specific primers) in coronary artery disease (CAD) patients compared to non-CAD subjects 22 . Altogether, these reports support the hypothesis that human-adapted P. fluorescens strains constitute low-abundance indigenous members of the microbial ecosystem of various body sites, such as the lungs, mouth, and stomach 14,[23][24][25] . Contextually, we can speculate that certain P. fluorescens-related strains are highly adaptable and poorly pathogenic members of the microbiota in several body sites that may frequently translocate into the bloodstream, providing a dominant contribution to bacterial DNAemia. However, we are conscious that our results do not conclusively demonstrate the actual presence of Pseudomonas (cells or free DNA) in blood. We believe that DNA-