The dormant blood microbiome in chronic, inflammatory diseases

1‘Overall, it seems inevitable that the availability of these methods will cause the catalog of disease states recognized as having a microbial contribution to their etiology to expand enormously in the short term, particularly as improved methods for resuscitation of small cell numbers are found’ (Davey and Kell ).

2Over the years, a variety of diseases that were previously considered non-communicable have been found to have a microbial component, the role of Helicobacter pylori in ulcerogenesis (Marshall and Warren ) being a particularly well-known example. There have also been hints for a microbial component to many other non-communicable diseases, but culturing the relevant organisms has rarely been successful. However, there is increasing recognition that microbes may be present in forms that are not easily culturable, and a number of recent articles have brought these possibilities more sharply into focus. Our aim is to review these developments. The manuscript structure is shown in Fig. 1.

3In this field, much confusion has arisen historically because of a failure to recognize that most microbes reproduce by binary fission and that this reproduction must be a minimal property or hallmark of a microbial cell that possesses ‘life’ or is ‘alive’ (Proal, Albert and Marshall ). Thus, as with Schrödinger's cat (e.g. Primas; Gribbin ), we cannot say that an individual microbial cell ‘is’ alive, only (if true) that it ‘was’ alive, since it will by then have become two cells. This implies that being alive is not best treated as though it were an innate property of a cell, but the definition must be operational, and include both the cell and the ‘environment’ (experiment) used to detect the status a posteriori (Kell et al. ).

4Thus, as with Postgate (e.g. Postgate,, ), we equate viability with culturability, and stress that culturability—the ability to reproduce—is to be determined operationally. Other methods that do not determine culturability are not tests of viability per se, but merely measure what they measure (e.g. the content of a chemical such as ATP, membrane permeability to a dye, enzymatic activity, macromolecular sequences and so on). In addition, it is impossible in principle to (cor)relate macroscopic measurements of a culture with the ability of individual cells to divide (Kell et al.; Davey and Kell ). In other words, if the macroscopic ATP content of say a starving culture were to decrease by 50%, we would not know if all of the cells had lost half their ATP or half of the cells had lost all of their ATP (or anything in between). The culturability of the former would likely be 50% and of the latter 100%, despite the same macroscopic ATP content.

5A lack of culturability may mean that a cell is non-viable under the circumstances tested, but viability or non-viability are not the only two possible states here. An apparent non-culturability of a surviving cell also admits another possibility, for which the natural term is ‘dormant’ (Kaprelyants, Gottschal and Kell; Epstein ). This is that the cell is not presently culturable (viable), but it is not ‘dead’ (in the sense of an operationally irreversible loss of viability) in that it may be induced to return to a state of culturability (by a process or processes typically referred to as ‘resuscitation’). This also means that the term ‘viable-but-non-culturable’, while quite common in use, is in fact an oxymoron that is to be discouraged (Kell et al. ). The eminent microbial physiologist Howard Gest is similarly scathing about the term ‘unculturable’ (Gest ), noting that one just needs to try harder to culture organisms. Table 1 shows the three terms best suited to discuss these issues, while Fig. 2 shows a diagrammatic representation of the macroscopic physiological microbial states we mostly consider.

6The assessment of replication potential (culturability) of individual cells may be done microscopically (e.g. by microscopic counts) or macroscopically (e.g. via colony formation on an agar plate or through the ‘most probable number’ technique). The latter has the advantage of potentially assessing dormancy in the absence of any contaminating culturable cells that might proliferate during the assay (Kaprelyants, Mukamolova and Kell; Votyakova, Kaprelyants and Kell; Kell et al. ). For assessing culturability (=viability), we do not therefore include other strategies in which cells do not actually divide, such as the so-called direct viable count of Kogure, Simidu and Taga. Thus, we here highlight the point that the possibility of microbial dormancy means that a system that appears to be devoid of culturable microbes may still contain dormant cells or forms that may become culturable.

7The circulation is a closed system and the blood in healthy organisms was first believed to be a sterile environment (Drennan; Proal, Albert and Marshall ). This definition is used in the most usual sense of an absence of culturable microbes, since blood can of course provide a suitable growth medium for microbes (as in blood culture; Wilson and Weinstein; Weinstein; Schroeter et al.; cf. Valencia-Shelton and Loeffelholz ), and any bacteraemia or sepsis, even at 1–10 cells mL−1 (Murray ), is potentially life-threatening (e.g. Vincent et al.; Eleftheriadis et al.; Havey, Fowler and Daneman; Montassier et al. ). However, the principle of the presence of truly sterile blood in healthy humans has been challenged, as operationally it does not mean that dormant or non-culturable forms of organisms are absent (Kaprelyants, Gottschal and Kell; Kell et al.; McLaughlin et al. ) (see Table 1). Nearly 50 years ago, the existence of a novel bacteriological system was noted in 71% of blood samples taken from diseased humans and from 7% of supposedly healthy humans, when RBCs were lysed (Domingue and Schlegel ). A year later, corynebacteria-like microorganisms developing in hemocultures were shown within RBCs (Tedeschi et al. ), and in 2001 it was found that even ‘healthy’ blood specimens can contain bacterial 16S ribosomal DNA (Nikkari et al. ). Domingue and Woody and Domingue summarizes much of this earlier literature. L-forms are bacterial variants that lack some or all of a cell wall. Nonetheless they can divide, especially in osmotically stabilized media, by processes that variously involve membrane blebbing, tubulation, vesiculation and fission (Allan, Hoischen and Gumpert; Errington; Mercier, Kawai and Errington ). While it remains unclear whether what was seen in these earlier studies (Domingue and Woody; Domingue ) may have been L-forms (Mattman ), that could in time revert to normal bacteria under the correct conditions (Casadesús ), L-forms are becoming a topic of considerable current research (Devine; Domínguez-Cuevas et al.; Mercier, Kawai and Errington, ).

8The presence of a blood bacterial microbiome has also been associated with a variety of infectious, as well as non-infectious disease states (Huang et al.; Thwaites and Gant; Nielsen et al.; Prajsnar et al.; Wang et al.; Kibru et al.; Sato et al. ). It is, for example, known that H. pylori can exist not only in the gastric mucosa but also in peripheral blood, where it could cause bacteremia (Huang et al. ), and could contribute to Parkinson's disease (PD) or related pathologies that precede motor symptoms (Nielsen et al. ). Helicobacter pylori was also previously implicated in the development of anemia (Wang et al.; Kibru et al. ). Staphylococcus aureus can also use neutrophils as ‘Trojan horses’ to disseminate infection (Thwaites and Gant; Prajsnar et al. ), while many other pathogens, such as Listeria monocytogenes (Xayarath and Freitag ), Salmonella typhimurium (Eisenreich et al.; Claudi et al.; Holden ) and Yersinia pestis (Isberg ), are well known to persist intracellularly; Gest gives other historical examples. The same is true for viruses, which are not discussed here.

9The presence of an aberrant blood microbiota (as assessed by sequencing) has been implicated in type II diabetes and cardiovascular disease (Amar et al.,; Sato et al. ). There is also growing evidence that periodontal disease and gingivitis are closely linked to cardiovascular disease (Yang et al.; Ramírez et al. ). Oral bacterial translocation into the blood has been implicated in the development of periodontal disease-induced endocarditis and myocardial and/or cerebral infarction, especially in patients with heart valve dysfunction (Koren et al.; Amar and Engelke; Seringec et al. ).

10We will argue in the next sections that the existence of potentially viable (but possibly non-proliferating) pleomorphic bacteria in the blood of healthy humans (McLaughlin et al. ) may therefore be of some significance in pathology. If such a microbiome can disrupt homeostasis, it can ultimately play a fundamental role in disease development and progression. It has therefore been proposed that the blood microbiota might therefore represent or contribute to the first step in the kinetics of atherosclerosis (Sato et al. ), cardiovascular disease and type II diabetes (Amar et al. ), and therefore ultimately serve as biomarkers for cardiovascular disease risk (Amar et al. ). However, in the quest to use the blood microbiota as biomarkers, the question of detectability and cultivability are key concepts.

11In particular, the existence of a blood microbiome is only really meaningful and of scientific interest if it represents an undisturbed state, and is not, for instance, an artefact caused by the external introduction of microbes through human intervention, reagent contamination (Schroeter et al.; Salter et al. ) and so forth. We therefore rehearse the evidence that while such artefacts are certainly possible, and must be excluded rigorously, the phenomenon of a human blood microbiome cannot be dismissed as such an artefact in toto.

12While contamination from reagents (e.g. Schroeter et al.; Salter et al. ), or simply poor sterile technique with needles and so on, can lead to an artefactual appearance of a blood microbiome, we consider that the following arguments, taken together, exclude the thought that the entire (and considerable) literature on a blood microbiome can be explained via contamination. The first argument is that there are significant differences between the blood microbiomes of individuals harboring disease states and nominally healthy controls, despite the fact that samples are treated identically (see later). Some similar arguments apply to the assessment of drug transporters under different conditions (Kell and Oliver ).A second argument is that the morphological type of organism (e.g. coccus versus bacillus) seems to be characteristic of particular diseases.A third argument is that in many cases (see below) relevant organisms lurk intracellularly, which is hard to explain by contamination.A fourth argument is that there are just too many diseases where bacteria have been found to play a role in the pathogenesis, that all of them may be caused by contamination.Finally, the actual numbers of cells involved seem far too great to be explicable by contamination; given that blood contains more than 109 erythrocytes mL−1, if there was just one bacterial cell per 100 000 erythrocytes (see below and Amar et al. ), this will equate to 104 bacteria mL−1. These are not small numbers.

13The first argument is that there are significant differences between the blood microbiomes of individuals harboring disease states and nominally healthy controls, despite the fact that samples are treated identically (see later). Some similar arguments apply to the assessment of drug transporters under different conditions (Kell and Oliver ).

14A second argument is that the morphological type of organism (e.g. coccus versus bacillus) seems to be characteristic of particular diseases.

15A third argument is that in many cases (see below) relevant organisms lurk intracellularly, which is hard to explain by contamination.

16A fourth argument is that there are just too many diseases where bacteria have been found to play a role in the pathogenesis, that all of them may be caused by contamination.

17Finally, the actual numbers of cells involved seem far too great to be explicable by contamination; given that blood contains more than 109 erythrocytes mL−1, if there was just one bacterial cell per 100 000 erythrocytes (see below and Amar et al. ), this will equate to 104 bacteria mL−1. These are not small numbers.

18It is important to point out that molecular methods have been used frequently to detect active sepsis. These selfsame methods are also used in environmental biology (as we pointed out in this review), without undue concern about the potential for contamination. Contamination will always be a concern, of course, as noted by Nikkari et al., but many papers since 2001 have documented strategies for detecting prokaryotic DNA in blood and serum using appropriate and careful controls (Anthony et al.; Mylotte and Tayara; Jiang et al.; Varani et al.; Mancini et al.; Chang et al.; Grif et al.; Fernández-Cruz et al.; Gaibani et al. ). Also, detecting bacteria in blood cultures during sepsis is considered the standard diagnostic tool for blood stream infections (Muñoz et al.; Varani et al. ), and some laboratories consider that e.g. PCR testing should always be a complement for the traditional blood culture test (Grif et al. ).

19Dormancy in microbiology is of course well known, even for non-sporulating bacteria, and has been defined as a stable but reversible nonreplicating state (Mariotti et al.; see also Table 1 and Kaprelyants, Gottschal and Kell; Kell et al., ). The importance of dormant or non-cultured (as opposed to ‘non-culturable’) organisms has long been recognized in environmental microbiology (e.g. Mason, Hamer and Bryers; Amann, Ludwig and Schleifer; Eilers et al.; Hugenholtz; Keller and Zengler; Pham and Kim; Epstein ), because of the 100-fold or greater difference between microscopically observable cells and those capable of forming a colony on an agar plate (‘the great plate count anomaly’, see below).

20Of the four main possibilities, what we do not know in general is whether the ‘missing’ cells are incapable of growth on the enrichment/isolation media,are killed by the enrichment/isolation media (e.g. Tanaka et al. ),have lost viability irreversibly (i.e. are operationally dead) orare in a dormant or not-immediately-culturable state from which we might resuscitate them (to effect culturability) if only we knew how.

21are incapable of growth on the enrichment/isolation media,

22are killed by the enrichment/isolation media (e.g. Tanaka et al. ),

23have lost viability irreversibly (i.e. are operationally dead) or

24are in a dormant or not-immediately-culturable state from which we might resuscitate them (to effect culturability) if only we knew how.

25The fact that typical isolation media and incubation conditions do not admit the measurable growth of all strains is certainly well known (indeed it is the basis for selective isolation media!), and it took a good while to learn how to culture pathogens such as H. pylori (Marshall and Warren; Marshall ), Legionella pneumophila (Feeley et al.; Saito et al.; Meyer ), Tropheryma whipplei (Maiwald and Relman; Maiwald et al.; Renesto et al. ) and so on (Singh et al. ). The majority of bacteria that persist in a ‘non-culturable’ form in wounds (e.g. Dowd et al.; Percival et al. ), or in diseases such as cystic fibrosis (Lewis ) or tuberculosis (Young, Stark and Kirschner; Zhang, Yew and Barer ), and even simply in conventional cultures of Escherichia coli (e.g. Koch; Balaban et al.; Keren et al.,; Gerdes and Maisonneuve; Amato, Orman and Brynildsen; Germain et al.; Maisonneuve, Castro-Camargo and Gerdes; Maisonneuve and Gerdes; Holden ), where phenotypic culture differentiation is well established (Koch ), are also ‘normally culturable’ by established means. Thus, the existence of operationally ‘non-culturable’ forms of only moderately fastidious bacteria is very well established, and more and more bacteria previously thought ‘unculturable’ are being brought into culture (e.g. Zengler et al.; Keller and Zengler; Stevenson et al.; Gich et al.; Kamagata and Tamaki; D'Onofrio et al.; Nichols et al.; Vartoukian, Palmer and Wade; Dedysh; Pham and Kim; Puspita et al.,; Stewart; Allen-Vercoe; Narihiro and Kamagata; Singh et al.; Walker et al.; Lagier et al.,; Ling et al. ).

26In environmental microbiology, some bacteria pass through the usual 0.2 μm filters, and have been referred to as ‘ultramicrobacteria’ (Macdonell and Hood; Morita ). It was proposed (Kaprelyants, Gottschal and Kell ) that rather than being small (starved) forms of normal bacteria they were more likely to be normal forms of small bacteria, and this seems to have been accepted (Lysak et al.; Sahin et al.; Duda et al.; Soina et al. ).

27The ability to culture certain kinds of soil bacteria by preincubation in weak broth is also well established (e.g. Bakken and Olsen; Kaprelyants, Gottschal and Kell ), and our own experiments showed very high levels of resuscitability of dormant cells of Micrococcus luteus (e.g. Kaprelyants and Kell; Kaprelyants, Mukamolova and Kell; Kaprelyants et al.,; Kell et al.,; Mukamolova et al.,,,,). In a similar way, substrate-accelerated death of non- or slowly growing microorganisms has been known for decades (Postgate; Calcott and Postgate; Calcott and Calvert ).

28Thus, any of several well-established mechanisms may contribute to the (often) large differences observable between microscopic counts and the number of operationally culturable microbes, with the greatest likelihood being that we simply have to develop more and better methods to bring these strains back into culture, i.e. to resuscitate them. In particular, however, this ‘great plate count anomaly’ has, of course, been brought into much sharper focus because of the advent of culture-independent, sequence-based means for detecting and (to a certain extent) enumerating microbes (though not, of course, of assessing their culturability).

29The vast majority of microbial species remain uncultivated and, until recently, about half of all known bacterial phyla were identified only from their 16S ribosomal RNA gene sequence (Lasken and McLean ). Also, single-cell genomics is a powerful tool for accessing genetic information from uncultivated microorganisms (Lasken; Rinke et al.; Cavanagh et al.; Clingenpeel et al. ). Bacterial single-cell genome sequencing and bioinformatics are, however, challenging (Pallen, Loman and Penn; Didelot et al.; Loman et al.; Fricke and Rasko ).

30The development of sequence-based methods for microbes (and especially non-eukaryotes) owes much to the pioneering work of Carl Woese and colleagues, who recognized the utility of small subunit ribosomal RNA (based on both its essentiality and the small but significant sequence variations) and applied it with great effect in molecular phylogenetics (Woese and Fox; Woese, Kandler and Wheelis ). Notwithstanding modern reinterpretations of the taxonomic details derived therefrom (e.g. Williams et al. ), there can be little doubt that this work drew the attention of microbiologists to the potential of sequence-based methods for detecting microbes that were then invisible to methods based solely on culture, e.g. in clinical microbiology (Didelot et al.; Loman et al.; Proal et al.; Fricke and Rasko ). rRNA remains a widely used strategy for detecting specific microbes. This has of course led to metagenomics, the large-scale sequencing of macromolecules and indeed (statistically) entire genomes from complex (non-axenic) environments, increasing the requirement for a full set of complete reference sequences (Kyrpides et al. ) and not just those of 16S rRNA (Yarza et al. ). Even the coupling of sequences to activities has now become possible (e.g. Radajewski et al.; Wang et al. ).

31More recently, gut metagenomics has been systematized with NIH's Human Microbiome project (HMP) and the European MetaHIT project aiming to deciphering the structure and function of the human gut microbiota (Fredricks; Robles-Alonso and Guarner ). The HMP has developed a reference collection of 16S ribosomal RNA gene sequences collected from sites across the human body (Koren et al.; Ding and Schloss ). This information can be used to associate changes in the microbiome with changes in health, and particularly also the blood microbiome. The Integrative Human Microbiome Project (iHMP, http://hmp2.org), the second phase of the NIH HMP, aims to study the interactions by analyzing microbiome and host activities in longitudinal studies of disease-specific cohorts and by creating integrated data sets of microbiome and host functional properties (The Integrative HMP (iHMP) Research Network Consortium ), ultimately allowing us to analyze host and microbial DNA (genome) and RNA (transcriptome) sequences (Morgan and Huttenhower ). However, in the HMP study, the main anatomic sites where samples are collected are skin, mouth, nose, colon and vagina (ElRakaiby et al. ). So far as we are aware, these projects do not focus on the blood microbiome (which is probably unsurprising when most commentators assume that it does not exist).

32The gut microbiome is by far the largest numerically, and our purpose here is not to review it in any detail, since this has been done very well in terms of its constitution (Lozupone et al.; Weinstock ),temporal variation (Caporaso et al.; Flores et al.; Thaiss et al. ),changes associated with diet (Muegge et al. ),obesity (Turnbaugh et al., ),age and geography (Delzenne and Cani; Delzenne et al.; Yatsunenko et al. ),inflammation (Cani et al., ),the immune system (Kau et al.; McDermott and Huffnagle )and various pathologies (Pflughoeft and Versalovic; Schulz et al. ).

33its constitution (Lozupone et al.; Weinstock ),

34temporal variation (Caporaso et al.; Flores et al.; Thaiss et al. ),

35changes associated with diet (Muegge et al. ),

36obesity (Turnbaugh et al., ),

37age and geography (Delzenne and Cani; Delzenne et al.; Yatsunenko et al. ),

38inflammation (Cani et al., ),

39the immune system (Kau et al.; McDermott and Huffnagle )

40and various pathologies (Pflughoeft and Versalovic; Schulz et al. ).

41It was implied that a better understanding of microbiome-encoded pathways for xenobiotic metabolism might also have implications for improving the efficacy of pharmacologic interventions with neuromodulatory agents (Gonzalez et al. ), and that the exploration of microbiome and metagenome might give us insightful new perspectives regarding human genetics and how the microbiota contribute to immunity, as well as to metabolic and inflammatory diseases (Cho and Blaser; Blaser et al.; Blaser; Leslie and Young ). This is because it is assumed in such studies that it is the small-molecule products of the gut microbiome that can appear in the human serum metabolome, and thus influence the rest of the human body (e.g. Wikoff et al.; Holmes et al.; Le Chatelier et al., and see Table 2). Here we also need to mention lipopolysaccharide (LPS), a main constituent of the Gram-negative outer membrane that induces the production of cytokines and/or chemokines, which in turn regulate inflammatory and innate and subsequent adaptive immune responses (Glaros et al.; Rhee; Ronco ). The release of LPS may therefore change gut homeostasis, may play a role in e.g. inflammatory bowel disease and necrotizing enterocolitis (Rhee ), and may certainly act as an acute phase protein in sepsis (Ding and Jin ).

42By contrast, our theme here is that it is additionally the microbes themselves that can pass from the gut (and other ‘external’ surfaces) into the human body, a phenomenon sometimes known as ‘dysbiosis’, albeit this term is more commonly used with another meaning. We here need to discriminate a changed (pathologic) microbiota in the place of origin from the results of a translocation of microbiota to other areas of the body. In the following sections, we use the term dysbiosis to describe changes in a microbiome in its main origin (typically the gut), and we coin the term ‘atopobiosis’ to describe microbes that appear in places other than where they should be.

43Dysbiosis, also known as dysbacteriosis, particularly referring to microbial imbalance in the digestive tract, has been widely discussed (e.g. Scher and Abramson; Scanlan et al.; Amar et al.; Bested, Logan and Selhub; Duytschaever et al.; Vaarala ). Core to this literature is the idea that factors that lead to significant changes in the gut microbiota composition (dysbiosis) ultimately result in pathology (Larsen et al.; Amar et al.,; Bested, Logan and Selhub; Burcelin et al.; De Angelis et al.; Fremont et al.; Lanter, Sauer and Davies; Petriz et al.; Power et al.; Tojo et al. ). Table 3 gives a list of diseases, largely inflammatory diseases, which have been associated with gut dysbiosis.

44In addition, we argue here that as well as gut dysbiosis, a derangement of the gut microbiome, what we are seeing here, often called ‘translocation’ in the context of surgery (Swank and Deitch; MacFie ) and various diseases (Berg ) (see Table 4 that lists diseases and conditions where bacterial translocation is specifically implicated), is what might better be called atopobiosis (Greek ατoπoς or atopos, in the wrong place), i.e. an appearance of members of the gut (or other) microbiome in the wrong place. Bacterial translocation is therefore discussed in the context of the movement of gut origin microbes [that changed from normal (dysbiosis)] that moved across the ‘intact’ gastrointestinal tract into normally sterile tissues, including blood, where the organisms may then directly cause infection or inflammation leading to tissue injury, organ failure, etc. (Steinberg; Wiest and Rath; Balzan et al. ). We stress that they may be found in both infectious and non-infectious diseases as well as being translocated during surgery, and that atopobiosis of bacteria originating in the oral cavity, e.g. in periodontal disease, may also be significant in rheumatoid arthritis, for instance (see below). Fig. 3 provides a schematic representation of dysbiosis, bacterial translocation and atopobiosis.

45If the gut microbiome is seen as the main source of the blood microbiome, it is necessary to establish which kinds of conditions might permit this in the absence of real physical damage (as may, for instance, be caused by surgery) leading to microbial translocation. Wiest, Lawson and Geuking mention three possible points of entrance for bacteria into the surrounding (sterile) tissue: by dendritic cells via processes between epithelial cells, not affecting tight junction function,via injured/inflamed epithelium with dysfunctional epithelial barrier,and via M cells overlying Peyer's patches as specialized cells providing access of microbial products to antigen-presenting cells.

46by dendritic cells via processes between epithelial cells, not affecting tight junction function,

47via injured/inflamed epithelium with dysfunctional epithelial barrier,

48and via M cells overlying Peyer's patches as specialized cells providing access of microbial products to antigen-presenting cells.

49We discuss bacterial translocation in this context in the following sections.

50While the gut epithelium represents the largest mucosal tissue, the mechanisms underlying the interaction between the microbiome and the epithelial cells remain poorly understood (Mathias et al. ). Although this is a vast and complex field that warrants a review of its own, we briefly argue that gut dysbiosis results in an atypical interaction of both the microbiota, as well as their secretory products, with the gut epithelial layer. This results in an altered barrier function, which may also lead to changed mucosal immunity and ultimately to atopobiosis. The gut epithelium is necessarily normally quite impermeable to microbes, but there is increasing evidence that direct chemical communication between the microbiota and the epithelial cells regulates mucosal integrity (Venkatesh et al. ). A possible point of entry is by direct cellular uptake, and there is one type of cell that can take up microbes, and these are the M cells overlaying the Peyer's patches (Kernéis et al.; Jepson and Clark; Clark and Jepson; Corr, Gahan and Hill; Lelouard et al.; Fukuda, Hase and Ohno ). Peyer's patches are seen as the ‘immune sensors’ of the gut epithelium. Considerable evidence exists that they provide a primary route for the limited translocation of microbes between the gut epithelium and the blood system (Jung, Hugot and Barreau ). These interactions with the cells of the gut may suggest that changes in the intestinal microbiota also influence mucosal immunity (Sato, Kiyono and Fujihashi et al. ). This is indeed the case, and gut dysbiosis has been shown to play a significant role in the development of autoimmune diseases, in particular inflammatory bowel diseases (Clemente et al.; Morgan et al.; Hold et al.; Kostic, Xavier and Gevers; Owyang and Wu; Ma et al. ). It was also suggested that a changed gut microbiota represents the initial site of autoimmunity generation, and might be a critical epigenetic factor in autoimmune diseases such as rheumatoid arthritis (Scher and Abramson; Luckey et al.; Brusca, Abramson and Scher; Catrina, Deane and Scher; Cénit et al.; Taneja ). There is also evidence that regulatory T cells in the gut are influenced by microbial factors, and that a changed microbiota (dysbiosis) may influence the induction and suppressor functions of these cells, in turn leading to a changed gut mucosal immunity (Kinoshita and Takeda ).

51We have earlier reviewed the literature that suggests that dysbiosis can cause gut epithelial barrier dysfunction, and thereby provide a point of entry into the body, including the blood, resulting in atopobiosis. This is supported by recent research that has suggested that blood microbiota might be implicated in various (cardiovascular and other) diseases. Sequence-based techniques provided evidence for the presence of such a blood microbiome. The question now arises as to whether such a microbiome's presence can be directly measured by e.g. ultrastructural (microscopic) methods, since a consequence of any translocation of microbes between the gut microbiome and blood is that they should then be observable in blood. The next sections will provide visual evidence of the presence of such a microbiota in Alzheimer's disease (AD) and PD. As shown in Table 3, these conditions are known to be associated with the presence of dysbiosis.

52Direct measurement by ultrastructural (microscopic) methods of analysis shows that microbes are in fact common constituents of blood in inflammatory diseases [previously seen in PD—Fig. 8 in (Pretorius et al. and in AD—Fig. 2 in (Lipinski and Pretorius ). We show and annotate selected micrographs from these papers in Fig. 4]. An important concern that needs to be addressed, as is also the case with sequence-based methods, is whether the presence of microbiota in whole blood is indeed not the result of introduced external contamination. There is in fact considerable evidence in the literature that bacteria as well as other microorganisms can reside inside RBCs (e.g. Minasyan ), and thus able to cross the RBC membrane somehow (see Table 5). Transmission electron microscopy (TEM) analysis showing bacteria inside cells would also tend to imply that the bacteria were not externally introduced artefactually during the preparation of the samples.

53For the current paper, we have revisited our AD and PD samples and figures from Pretorius et al. and Lipinski and Pretorius and noted the prevalence of bacteria in almost all of the AD and PD samples, in numbers much in excess of those seen in our database of thousands micrographs from healthy individuals. Here we show additional micrographs from the previously published samples (see Figs 5 and 6). In both conditions (see Figs 5AD and 6PD), microbes were noted in close proximity to RBCs, and in some cases RBCs extended pseudopodia-like projections towards the microbiota. SEM analysis of AD whole blood (Fig. 5) shows that mostly coccus-shaped bacteria are present. White blood cells are seen in close proximity to these bacteria in AD patients (see Fig. 5A–C). SEM analyses of PD patients (Fig. 6) show both coccus- and bacillus-shaped bacteria in close proximity to RBCs. We also observed that RBCs extend pseudopodia towards these bacteria and this might be part of the mechanism by which the bacteria enter the RBCs (see Fig. 6C–F). We also note possibly dividing coccus-shaped bacteria in both these conditions, indicated with blue arrows on Fig. 5A (AD patient) and Fig. 6D (PD patient). This might suggest that these bacteria may be(come) culturable under appropriate conditions (see also Soina et al.; Epstein ).

54TEM analysis of the samples from Lipinski and Pretorius and Pretorius et al. showed the presence inside RBCs of cells that appeared to be microbial in nature (unpublished data). These internalized cells further provide evidence for a sustained presence of such a blood microbiota (and one hardly explained by contamination) (see Fig. 7A and B: AD and C and D: PD). Bacteria are shown with arrows in the micrographs. No bacterial membrane was noted; therefore, the bacteria may be L-forms. There seems to be bacterial species selectivity for a given disease, as our preliminary observations suggest a prevalence for bacillus-type bacteria in AD, but both coccus- and bacillus-shaped bacteria in PD patients.

55Our observations suggest that the presence of bacteria in these two diseases occurs in only a small fraction of the RBC population, which is why we had not really noted them in our previous studies (e.g. Bester et al.; Pretorius et al.,,; Pretorius and Kell ), and SEM and TEM analysis confirms this observation. We have never (or not yet) found bacteria inside RBCs from healthy controls (these without overt, diagnosed diseases) when studying blood smears using TEM analysis. The microscopy preparation methods involve a washing process, and this may wash away some of the bacteria, or RBCs and white blood cells associated with bacteria. Therefore, the actual quantification of the bacteria can only be done by other means; however, dormancy and viability versus non-viability issues pertain (as discussed above).

56We found a definite association between RBCs and bacteria, with RBCs (see Figs 6 and 7) forming pseudopodia-like extension, as if in the process of engulfing bacteria. Both coccoid (round) and bacillary (elongated) bacteria were found in PD whole blood SEM micrographs, but only coccoid forms in AD whole blood SEM micrographs. Samples from 25 diagnosed AD patients were studied and bacteria were detected in 14 individuals from this AD sample, while samples from 30 PD patients were studied, in 21 of whom we detected bacteria. Obviously, the type of bacteria cannot be identified from ultrastructural observations. As with the timeline of established cases such as the role of H. pylori in ulcers and colon cancer, the next tasks are to bring these microscopically observed bacteria into culture and to carry out sequence-based studies to establish their role (if any) in non-communicable diseases. However, to illustrate that the bacteria may indeed be engulfed by the RBCs, and to confirm that the phenomenon is not due to external contamination, we show TEM micrographs from both of the studied diseases (see Fig. 7, AD and PD).

57‘Non-culturable’ (which should be called ‘not-easily-culturable’ or ‘not-yet-cultured’) microbes are commonplace in the ‘environmental microbiology’ of soil and water, and the blood certainly represents an ‘environment’. As we show here, there is a large and scattered literature, increasing in size, to the effect that there might be a (mainly dormant) microbial component in a variety of chronic diseases that are normally considered to be non-microbial or non-communicable in nature, even when microbes appear absent by culturability criteria. Our previous work (e.g. Bester et al.; Pretorius et al.,; Kell and Pretorius,; Pretorius and Kell ) has implied iron dysregulation as a regular accompaniment to, and probable contributory factor for, a variety of similar diseases, all of which have an inflammatory component. We argue here that there is also a microbial contribution to this in the blood, and it is not unreasonable that the microbial requirement for iron means that, despite the oxidative stress it can entail (Touati; Kell, ), microbes may be anticipated to increase in prevalence when iron is free (e.g. Ratledge; Clifton, Corrent and Strong; Sia, Allred and Raymond; Chu et al. ) and available (D'Onofrio et al. ), probably behaving in a social manner (Kell, Kaprelyants and Grafen; West and Buckling; Diggle et al.; Harrison and Buckling ).

58We have here pointed up the likelihood of a steady crop of effectively dormant microbes being a feature of blood biology in chronically diseased humans, including those with non-communicable diseases. As with any complex system, the magnitude of any component is affected by the kinetics of every relevant step; while the precise nature of all the interactions is uncertain, Fig. 8 describes the general network—the first step in any systems analysis (Kell; Kell and Knowles ).

59Consequently, we recognize that the analysis above has largely been qualitative (the ‘presence’ of a microbial component in a specific disease is a qualitative statement). However, chronic, non-communicable diseases are very far from being static (and thousands of human genes change their expression at least 2-fold even on a diurnal basis; Zhang et al. ). Thus, a clear further issue is to seek to understand how the blood microbiome may co-vary with the day-to-day dynamics of chronic diseases. For example, rheumatoid arthritis has circadian rhythms (Straub and Cutolo ) and is well known to provide significant variations (‘flares’; Flurey et al. ) in severity at different times. A reasonable strategy is thus to look for changes in a detectable blood microbiome in this and other diseases that show such flares. As with H. pylori and stomach ulcers (and cancer), the simple prediction is that bactericidal antibiotics should be of value in the treatment of such supposedly non-communicable diseases. Indeed, this prediction is borne out for diseases such as rheumatoid arthritis (Ogrendik,; Kwiatkowska and Maślińska ) and multiple sclerosis (Ochoa-Repáraz et al.; ), while antipneumococcal vaccination has shown efficacy in preventing stroke (Vila-Corcoles et al. ). Of course, events such as heart attacks and strokes (and see Table 4) may also be seen as sudden increases in severity of an underlying condition, and in some cases (such as the much increased likelihood of strokes after subarachnoid haemorrhages; McMahon et al. ), analysis of changes in the blood microbiome might prove predictive.

60The obvious next tasks are thus to relate the number and nature of blood microbes observed in cases such as the above to microbial sequences and antigens that can be detected in aliquots of the same samples (e.g. Salipante et al., ), to determine the physiological state of the various microbes (including e.g. whether they are L-forms), and to establish methods to bring them (back) into culture. Since microbes, inflammation and various syndromes are such common co-occurrences (as are coagulopathies; Kell and Pretorius ), longitudinal studies will have a specially important role, as they will both show the dynamics and be able to help discriminate cause and effect during the time evolution of chronic, non-communicable diseases in ageing populations. The immunogenicity of persisters, and their ability to induce various kinds of inflammation, must be rather different from that of replicating organisms, and this must be investigated. Armed with such collective knowledge, we might be better placed to develop therapeutics such as pre- and probiotics and bactericidal antibiotics for use in such cases previously thought to lack a microbial contribution.

61We thank the Biotechnology and Biological Sciences Research Council (grant BB/L025752/1) as well as the National Research Foundation (NRF) of South Africa for supporting this collaboration.

62Conflict of interest. None declared.

6316S ribosomal RNA: a component of the 30S small subunit of prokaryotic ribosomes. The 16S rRNA gene is found in all bacteria and archaea and consists of nine short hypervariable regions that may be used to distinguish bacterial taxa.

64Anabiosis: when an organism is in a state of very low metabolic activity to the extent where it is hardly measurable and in some cases come to a standstill. The physiological and biochemical processes are arrested for different periods of time but can be reversed.

65Atopobiosis (Greek or atopos) appearance of the gut or other microbiome in the wrong place.

66Bacterial translocation: the passage of viable resident bacteria from the gastrointestinal tract to normally sterile tissues such as the mesenteric lymph nodes and the other internal organs.

67Cryptobiosis: refers to latent life or a state where an organism lacks any visible signs of life but is not dead in that it may revert to a state of aliveness as usually defined. Its metabolic activity becomes hardly measurable, or comes reversibly to a standstill.

68Culturability: the ability of a cell to reproduce.

69Direct viable count: the original method comprises incubation of samples with nutrients (yeast extract) and a single antimicrobial agent that specifically inhibits DNA synthesis but not RNA synthesis (nalidixic acid). Cell division ceases as a result of DNA synthesis inhibition but other cellular metabolic activities remain unaffected and therefore cells continue to metabolize nutrients and grow in size, which allows their detection microscopically in situ.

70Dormant: not viable in the sense of not being more or less immediately culturable, but may be returned to a state of viability or culturability by preincubation under suitable conditions.

71Dysbiosis: derangement of the species distribution in the normal microbiome.

72L-forms: these bacteria are cell wall-deficient forms of normal bacteria. They are able to proliferate as sphaeroplasts or protoplasts under certain conditions.

73Metagenomics: direct genetic analysis of a collection of genomes contained in an environmental sample.

74Microbiome: the genetic sum of the ecological community of commensal, symbiotic and pathogenic microorganisms that lives on and inside our bodies.

75‘Most Probable Number’ technique: is a method used to quantify the concentration of viable microorganisms in a sample. It involves replicate liquid broth growth in 10-fold dilutions. When a dilution lacks growth, it is assumed not to have any organisms. Back-calculation via a Poissonian distribution leads to the ‘most probable number’ in the original sample

76Non-axenic culture: contains more than one species, variety or strain of organism.

77Non-viable: incapable of observable replication by any stated means normally capable of effecting replication in the relevant organism.

78Phylogenetics: a discipline of evolutionary biology that studies the relationships between organisms based on how closely similar some of their macromolecular sequences are.

79Pleomorphic: possessing the ability to change shape or size in response to environmental stimuli.

80Resuscitation: induction of apparently non-culturable cells to a state of culturability.

81Sterile: refers to an absence of operationally viable organisms.

82Viable: capable of observable replication, i.e. culturable, by any stated means.