Microfluidics-based organism isolation from whole blood - an emerging tool for bloodstream infection diagnosis

Abstract

The diagnosis of bloodstream infections presents numerous challenges, in part, due to the low concentration of pathogens present in the peripheral bloodstream. As an alternative to existing time-consuming, culture-based diagnostic methods for organism identification, microfluidic devices have emerged as rapid, high-throughput and integrated platforms for bacterial and fungal enrichment, detection, and characterization. This focused review serves to highlight and compare the emerging microfluidic platforms designed for the isolation of sepsis-causing pathogens from blood and suggest important areas for future research.

1.0. Introduction

Sepsis is a life-threatening condition that occurs when infectious organisms are present in the peripheral bloodstream resulting in a host shock response. Despite the severity and prevalence of sepsis, current diagnostic methods are extremely time-consuming. This is, in part, due to the extremely low concentration of infectious organisms in the bloodstream ¹¹⁵, making pathogen detection technically challenging.

Microfluidic platforms have emerged as a popular alternative to traditional macro-scale diagnostic methods ^{30, 34, 55, 74, 90}. These micro-scale systems have demonstrated their ability to isolate and detect rare cells from simple fluids (e.g. water, liquid media) by harnessing a variety of physical and chemical separation methods ^{12, 14, 17, 22, 32, 33, 47, 52, 58, 59, 70, 73, 85, 91, 92, 96, 103, 104, 108}. More specifically, prior work has demonstrated the feasibility of employing microfluidic platforms for bacterial isolation and detection of waterborne ^{4, 8, 60–62, 101, 120} and foodborne ^{1, 9, 24,54, 57, 65, 75, 80, 87, 117, 121} pathogens. Additionally, a smaller subset of studies has evaluated bacterial capture and detection in urine ^{83, 116} and joint synovial fluid ¹⁶. That said, most studies in this space circumvent a pressing technical challenge: the use of micro-scale systems for the isolation, detection, and characterization of sepsis-causing organisms in complex biological mediums, such as blood. Blood contains a wide variety of biological material, including red blood cells, white blood cells, platelets, and other small molecules ³⁸, making targeted isolation and characterization of sepsis-causing organisms a more challenging technical problem.

Our discussion begins by providing the necessary background on the epidemiology and etiology of bloodstream infections, as well as highlighting the existing diagnostic workflow. Next, we will summarize and review the primary literature that describes the use of microfluidic tools and approaches for the enrichment of bacteria and fungi from blood. In addition, some of these studies employ novel and integrated pathogen detection strategies, which will also be discussed. We will also briefly highlight some examples of emerging microfluidic platforms for pathogen identification and characterization, which could potentially be employed for downstream analysis. Finally, we will discuss key areas for future work that are needed to help address this important diagnostic need.

2.0. Epidemiology and Etiology of Bloodstream Infections

Sepsis ranks among the top seven causes of death in North America and Europe ³⁵. In these regions, the average annual mortality rate averages 29 per 100,000 individuals, with a total case fatality rate ranging from 13% to 20% ³⁵. Further, among all patients admitted to the intensive care unit (ICU), approximately 7.7% will develop a bloodstream infection (Figure 1) ¹¹⁰. Decreasing time-to-diagnosis is uniquely critical for bloodstream infections. Specifically, it is estimated that mortality from sepsis increases by 7.6% for every hour that treatment is delayed ³¹. Therefore, decreasing time-to-diagnosis is critical to improving overall patient outcomes ⁶⁶.

Figure 1.

Classification of ICU patients as a function of infection and infection type¹¹⁰.

Sepsis can be caused by a wide-range of pathogens (Table 1) ⁷⁹. As is listed in Table 1, gram-positive bacteria are identified in approximately 46.8% of infections, gram-negative bacteria are identified in approximately 62.2% of infections, and fungi are identified in approximately 19.4% of infections. Of note is the relatively high frequency of methicillin-resistant Staphylococcus aureus (MRSA) observed in bloodstream infections. Additionally, in approximately 17% of cases, there is more than one pathogenic species causing an infection, resulting in a co-infection ⁷⁹. As will be discussed in detail below, both rapid species identification and antibiotic susceptibility profiling of the sepsis-causing organism(s) are critical variables in informing the appropriate antibiotic treatment ^{23, 67, 69, 81}.

Table 1:

Observed frequency of sepsis-causing pathogens. Table adapted from Mayr et al. ⁷⁹.

	Frequency %	Odds Ratio
Gram-positive	46.8	-
Staphylococcus aureus	20.5	0.8
Methicillin-resistant Staphylococcus aureus	10.2	1.3
Enterococcus	10.9	1.6
Staphylococcus epidermis	10.8	0.9
Streptococcus pneumoniae	4.1	0.8
Other	6.4	0.9
Gram-negative	62.2	-
Pseudomonas spp.	19.9	1.4
Escherichia coli	16.0	0.9
Acinetobacter spp.	8.8	1.0
Enterobacter	7.0	1.5
Other	17.0	1.2
Anaerobes	4.5	0.9
Other bacteria	1.5	1.1
Fungi	19.4	-
Parasites	0.7	-
Other organisms	3.9	-

3.0. Current Diagnostic Workflow for Bloodstream Infections

The diagnosis of sepsis involves three primary benchmarks: 1) pathogen presence, 2) species identification, and 3) antibiotic susceptibility profiling (Figure 2) ^{23, 67, 69, 81}. First, a blood culture is conducted in order to confirm the presence of the pathogen. To do this, a primary patient blood specimen is cultured in a nutrient rich media. As the pathogen proliferates, carbon dioxide is emitted, and generally, detected via a carbon dioxide sensor that is integrated into the blood culture bottle. The time to blood culture positivity is usually greater than twelve hours, and can take upwards of five days, depending on the specific growth rate of the sepsis-causing organism ^{37, 86}. After establishing the presence of a pathogen through a blood culture, additional tests are then needed to identify the specific pathogen causing the infection, as well as characterize the antibiotic susceptibility of the pathogen ^{5, 29, 53, 88}. Traditional culture-based characterization methods require at least 24 hours ⁶⁷. Newer, molecular methods can perform rapid species-level identification and characterize antibiotic resistance profiles for select organisms in as little as 2 – 7.5 hours following blood culture positivity ^{5, 76, 99}.

Figure 2.

Overview schematic of current bloodstream infection diagnostic workflow. First, a 5 – 10 mL patient blood sample is cultured in 25 mL of nutrient rich media to determine if a pathogen is present¹³(1). Next, either culture-based or molecular methods are used to identify the pathogen(s) causing the infection (2). Above, we show a rapid (^~2 h), molecular testing system for pathogen identification, the BioFire® FilmArray® Blood Culture Identification Panel (BCID). Reprinted under permission of the Creative Commons Attribution License⁹³. Lastly, the organism’s susceptibility to antibiotics is determined (3). Here, we highlight the minimum inhibitory concentration (MIC) plate, which is a widely employed culture-based methodology.

Given the strong correlation between shortened time-to-diagnosis and deceased mortality rates, more rapid diagnostic methods are needed to identify and characterize the sepsis-causing organism(s)⁶⁶. By pinpointing the etiology of the organism(s) at an earlier time-point, therapy can more rapidly transition from a broad-spectrum to a targeted antibiotic. In effect, this reduces the likelihood of super infection, lowers healthcare costs, and minimizes contributions to antibiotic resistance ²³. Bacterial enrichment and detection using microfluidics offers the potential to eliminate the current universal need for blood culture. Further, use of an integrated microfluidic platform to 1) diagnose sepsis and 2) characterize the sepsis-causing organism(s) directly from whole blood could serve to optimize and refine the current diagnostic workflow, and contribute dramatically to improved patient outcomes.

4.0. Field Overview: Microfluidic Platforms as a Diagnostic Tool for Bloodstream Infections

In recent years, microfluidic devices have emerged as promising platforms for the diagnosis of bacteremia (sepsis). Microfluidic systems enable extremely precise fluid control and manipulation. This becomes especially relevant when processing samples in complex sample mediums, such as whole blood. Additionally, these high-throughput systems can be readily multiplexed, enabling the simultaneous isolation and detection of multiple bacterial targets. Table 2 outlines the major blood components and some of the physical characteristics that are relevant to consider in the design of microfluidic platforms. As shown, many blood components have similar physical characteristics to bacteria (E. coli). Further, the concentration of sepsis-causing organisms in blood is significantly less than the concentration of other cellular blood components. In combination, these two factors make microfluidic isolation of pathogenic organisms from whole blood a much more challenging process than microfluidic isolation of pathogens from simple fluids.

Table 2:

Physical characteristics of main blood components relative to example sepsis-causing pathogen (E. coli).

Component	Size	Concentration (cells/mL)	Density (g/mL)	References
Red blood cells (RBCs)	7.5 – 8.7 µm (diameter)	3.7 – 5.8 × 109	1.086 – 1.122	25, 38, 90
White blood cells (WBCs)	7 – 20 µm (diameter)	3.0 – 11.7 × 106	1.057 – 1.092	38, 90, 95
Platelets	1.5 – 3 µm (diameter)	2.0 – 4.0 × 108	1.072 – 1.077	38, 90, 107
Plasma	-	-	1.024	90
E. coli	1 – 3 µm (length)	< 10	1.105	78, 98, 115

Table 3 summarizes studies that have employed microfluidic approaches to isolate sepsis-causing pathogens from whole blood. To the best of our knowledge, this matrix captures the current state of the field. As shown, a variety of microfluidic approaches have been employed for rapid pathogen isolation, and include acoustophoresis, dielectrophoresis, immunoaffinity-based methods, inertial fractionation, and adhesion-based separation ^{2, 6, 10, 15, 18, 20, 21, 27, 28, 40, 42–44, 49, 50, 56, 63, 72, 82, 97, 111–114, 118, 119}. Figure 3A highlights the observed frequency in the primary literature of various physical and chemical strategies for pathogen isolation. Notably, only a limited number of studies evaluate bacterial isolation at clinically relevant bacterial concentrations (<10 CFU/mL), with most studies assessing separation performance at bacterial loads greater than 1000 CFU/mL. Additionally, the majority of existing studies limit their analysis to only a single pathogen: Escherichia coli, and only one study tackles the need for rapid antibiotic susceptibility profiling ⁴⁰. Figure 3B summarizes the observed frequency of the various strategies for pathogen detection described in the primary literature. To date, most studies (>85%) fail to integrate on-chip detection. Alternatively, these studies rely on well-established off-chip detection and enumeration strategies, including flow cytometry, microscopy techniques, and molecular methods (Figure 3B) ^{2, 6, 10, 15, 18, 20, 21, 27, 28, 40, 42–44, 49, 50, 56, 63, 72, 82, 97, 111–114, 118, 119}.

Table 3:

Comparison matrix for primary literature reporting pathogen isolation from whole blood ^{2, 6, 10, 15, 18, 20, 21, 27, 28, 40, 42–44, 49, 50, 56, 63, 72, 82, 97, 111–114, 118, 119}.

AUTHOR, PUBLICATION YEAR	SEPARATION MECHANISM	MINIMUM CELL LOAD (CELLS/ML)*	CAPTURE EFFICIENCY*	PATHOGENS EVALUATED	SAMPLE FLOW RATE	ON-CHIP DETECTION?	SPECIES ID?	AST?
AI ET AL., 20032	Acoustophoresis	3 × 106	95.65% (purity)	E. coli	0.5 µL/min	●	●	●
ARAZ ET AL., 20036	Acoustophoresis	Not specified	Not specified	E. coli	n/a	●	●	●
BISCEGLIA ET AL., 201510	Dielectrophoresis	1 × 104	97% (E. coli)	E. coli, S. epidermis, C. albicans	10 – 20 µL/h	●	●	●
CAI ET AL., 201415	Dielectrophoresis	1 × 10^3	70.9–94.8%	E. coli	1µL/min	●	● (On-Chip PCR)	●
CHO ET AL., 200718	Immunomagnetic	1 × 105	93.40%	E. coli	12 min (Total analysis time)	●	● (Integrated in separation)	●
COOPER ET AL., 201420	Immunomagnetic	1 × 100–1	98%	C. albicans	10 mL/h	●	●	●
D’AMICO ET AL., 201721	Dielectrophoresis	1 × 103	78–79%	E. coli, S. aureus	10 µL/min	●	● (Off-Chip PCR)	●
DOW ET AL., 201827	Acoustophoresis	1.2 × 102	68% (P. aeruginosa)	E. coli, P. aeruginosa, S. aureus	50 µL/min	●	● (Off-Chip Luminescence Assay)	●
FARIDI ET AL., 201728	Inertial	1 × 106	76%	E. coli	30 – 60 µL/h	●	●	●
HOU ET AL., 201540	Inertial	1 × 101	>10%	E. coli, P. aeruginosa, K. pneumoniae, S. aureus, E. faecalis	50 µL/min	●	● (Off-Chip PCR)	●
HUANG ET AL., 200342	Dielectrophoresis	4.6 × 103	89%	E. coli, B. cereus, L. monocytogenes	n/a	●	● (Off-Chip PCR)	●
HWANG ET AL., 200843	Adhesion	1 × 107	40%	E. coli	200 µL/min	●	● (Off-Chip PCR)	●
HWANG ET AL., 201144	Adhesion	1 × 104	40%	E. coli	200 µL/min	●	● (Off-Chip PCR)	●
KANG ET AL., 201449	Nanodroplets	1 × 100	77%	E. coli	0.5 – 3 µL/min	●	● (Nature of Target Complex)	●
KANG ET AL., 201450	Immunomagnetic	1 × 104	>90%	S. aureus, E. coli, C. albicans	10 mL/h	●	●	●
KUCXSENSKI ET AL., 201156	Dielectrophoresis	1 × 106	30%	E. coli	35 µL/h	●	●	●
LEE ET AL., 201363	Immunomagnetic	5 × 106	>95%	E. coli	60 mL/h	●	●	●
MACH ET AL., 201072	Inertial	1 × 108	>80%	E. coli	200 µL/min	●	●	●
OHLSSON ET AL., 201682	Acoustophoresis	1 × 103	10%	P. putida	80 µL/min	●	● (Off-Chip PCR)	●
RAUB ET AL., 201497	Size-exclusion	2.2 × 106	30%	E. coli	~100 µL/min	●	●	●
WANG ET AL., 2012111	Immuno-affinity	5 × 101	70.7%	E. coli	2 µL/min	●	● (Integrated in separation)	●
WEI-HOU ET AL., 2012112	Inertial	1 × 106	80–90%	E. coli, S. cerevisiae	15 µL/min	●	●	●
WU ET AL., 2009113	Inertial	1 × 107.2	62%	E. coli	2 – 18 µL/min	●	●	●
XIA ET AL., 2006114	Immunomagnetic	5 × 106	78%	E. coli	25 µL/h	●	● (Integrated in separation)	●
YUNG ET AL., 2009118	Immunomagnetic	1 × 106	80%	C. albicans	20 mL/h	●	● (Integrated in separation)	●
ZELENIN ET AL., 2014119	Targeted cell lysis	1 × 107	100%	E. coli, M. luteus	48 µL/min	●	●	●

^*Data for lowest reported cell concentration evaluated with statistically different results from control. Pathogens evaluated include only those where separation from whole blood was reported. ID = identification; AST = antibiotic susceptibility testing; ● = no; ● = yes.

Figure 3.

Breakdown of strategies for pathogen isolation and detection, and their observed frequency in primary literature. A. Strategies for pathogen isolation from whole blood ^{2, 6, 10, 15, 18, 20, 21, 27, 28, 40, 42–44, 49, 50, 56, 63, 72, 82, 97, 111–114, 118, 119}. Each dot represents one study in the primary literature. “Other” category includes 1) isolation using nanodroplets ⁴⁹ and 2) selective lysis of red blood cells ¹¹⁹. B. Strategies for pathogen detection following isolation^{2, 6, 10, 15, 18, 20, 21, 27, 28, 40, 42–44, 49, 50, 56, 63, 72, 82, 97, 111–114, 118, 119}. Primary detection method is indicated. Green ×’s indicates on-chip detection. In some cases, (n=2), two methods are specified. “Culture” is only listed as the detection method if no other method was employed. (In many cases, culture-base methods are used as a control and/or gold standard; these cases are not specified above if more rapid methods were employed).

Below, we will discuss the primary literature in more detail. First, we provide an overview of notable strategies for on-chip pathogen isolation and enrichment. Next, we describe reported integrated methods for on-chip pathogen isolation and detection. Lastly, we highlight our conclusions and suggest areas for future research and development.

5.0. Microfluidic Methods for Pathogen Isolation and Enrichment

5.1. Acoustophoresis

Acoustophoresis is a separation method that employs high-intensity sound waves to enable contact-free migration of target cells towards low pressure nodes. Contact-free manipulation of cells limits cell stress and preserves cell viability making acoustophoresis an attractive option for whole-cell pathogen separation ¹⁰⁹. To date, a few studies have used this separation tool for the purpose of separating bacteria from whole blood. Ai et al. separated E. coli from peripheral blood mononuclear cells with a resulting sample purity of 95.65%. Following on-chip separation using two identical surface acoustic waves (Figure 4A), the separation efficiency was characterized off-chip using flow cytometry ². Although this study provides an initial proof-of-concept for the use of acoustic waves for the separation of bacteria from whole blood, the bacterial concentration used for this analysis was on the order of 10⁶ cell/mL, which is significantly higher than the concentration of bacteria found in a primary human blood sample ².

Figure 4.

Acoustic separation platforms for the isolation of bacteria from blood. A. Overview of Ai et al. platform, which employed two parallel acoustic waves to focus bacteria in the center of the microchannel ². Reprinted with permission from https://pubs.acs.org/doi/10.1021/ac4017715. Copyright 2013 American Chemical Society. Further permissions related to the material excerpted should be directed to the ACS. B. Overview of Ohlsson et al. workflow. First, RBCs are focused in the center of the channel. Next, bacteria are acoustically trapped on a seeded particle and profiled using PCR ⁸². Reprinted with permission from https://pubs.acs.org/doi/abs/10.1021%2Facs.analchem.6b00323. Copyright 2016 American Chemical Society. Further permissions related to the material excerpted should be directed to the ACS. C. Methodology proposed by Dow et al. First, the bacterial sample is enriched on a plastic microdevice. Enriched sample is exposed to bacteriophage luminescence assay ²⁷. Republished with permission of Royal Society of Chemistry from, Acoustic separation in plastic microfluidics for rapid detection of bacteria in blood using engineered bacteriophage, Dow et al., 18, 2018; permission conveyed through Copyright Clearance Center, Inc.

Ohlsson et al. conducted a very comprehensive study using acostophoretic separation. Experiments from this study included the separation of bacteria from plasma, the separation of bacteria from whole blood, and the separation and identification of bacteria from ex vivo clinical blood cultures ⁸². Specifically, for their work involving the separation of bacteria from whole blood, Ohlsson et al. reported the successful detection of 1000 bacteria/mL using their microchip coupled with an external thermocycler ⁸². First, Ohlsson et al. acoustically focused red blood cells in the center of the microchannel to be removed. Following this, bacteria (Pseudomonas putida) were trapped on 12 μm polystyrene seeding particles in a sequential capillary channel (Figure 4B) ⁸². These acoustically trapped bacteria-seeding particle complexes were then washed and released into a secondary polymer microchip contain dry PCR reagents. The solution was then transferred to a thermocycler to continue the molecular analysis ⁸². Although the detection limit of their first-generation system is not yet capable of detecting low-grade sepsis, the results presented in this study are extremely promising, and suggest the feasibility of eliminating the need for the time-consuming blood culture.

Most recently, Dow et al. coupled acoustic separation in a plastic micro-device to a bacteriophage-based luminescence assay (Figure 4C) ²⁷. At a clinically relevant bacterial input concentration of 120 cells/mL, their acoustic separation device achieved a 68% bacteria yield (P. aeruginosa). As a result of this enrichment step, they observed a 33-fold improvement on the limit of detection for their novel bacteriophage-based luminescence assay ²⁷. Notably, the host-phage specificity required for the functionality of the luminescence assay allows for the potential incorporation of rapid pathogen species identification into the proposed workflow. Dow et al. also evaluated the separation efficiency of E. coli and S. aureus from whole blood to demonstrate the versatility of a non-specific separation platform in isolating both Gram-positive and Gram-negative organisms ²⁷.

5.2. Dielectrophoresis

Dielectrophoresis is a technique that uses an electric field gradient to isolate target cells as a function of their dielectric characteristics ⁸⁹. Kuczenski et al. presented the critical hydrodynamic and electrokinetic theories for effective sorting of spheroid cells. The team then applied this theory to the design of an electrophoretic microfluidic device ⁵⁶. With an initial bacterial load (E. coli) of 10⁶ cell/mL, their device had a capture efficiency of 30%. Further improving upon these results, Bisceglia et al. demonstrated a 97% capture rate of E. coli using dielectrophoresis with an initial bacterial load of 1 × 10⁴ cells/mL ⁵⁶. Bisceglia et al. also demonstrated the feasibility of this methodology in isolating Gram-positive (S. aureus) and fungal (C. albicans) pathogens ¹⁰. Additionally, this work assessed the relative effects of isotonic versus hypotonic buffer conditions on the conductivity and relative permeability of the sample components, in addition to the overall efficacy of the separation ¹⁰.

More recently, D’Amico et al. reported on a microfluidic system that used dielectrophoresis to isolate E. coli and S. aureus from whole blood ²¹. At bacterial concentrations as low as 1000 cells/mL, capture efficiencies were 79% and 78%, respectively ²¹. Specifically, they reported on an integrated microfluidic dialysis-dielectrophoresis isolation system (Figure 5A) ²¹. This system involves two membraneless microdialysis devices (MMDs) arranged in series, followed by the active dielectrophoresis microchip. The MMD devices are necessary to help reduce the electrical conductivity of the blood sample via rapid diffusion prior to the sample entering the dielectrophoresis microchip. In the dielectrophoresis microchip, bacteria are drawn towards the electrodes, while other blood components are repelled and washed away (Figure 5B) ²¹. Following separation, the dielectric field is turned off, the captured sample washed out of the microchip, and 16S PCR is conducted for species identification and characterization (Figure 5C) ²¹. This study by D’Amico et al. demonstrates the potential feasibility of dielectrophoresis as a method for rapid bloodstream infection diagnosis.

Figure 5.

Dielectrophoretic platform for isolation of bacteria from blood proposed by D’Amico et al 21. A. Schematic of the integrated microfluidic dialysis-dielectrophoresis isolation system. Two MMDs are connected in series, followed by the active dielectrophoresis device. B. Sample flow through separation microchip. Bacteria (green) are attracted to electrodes, while other blood components are repelled and washed through the device. C. Overview of proposed workflow. Following separation, sample is eluted from device for molecular analysis. Republished with permission of Royal Society of Chemistry from, Isolation and concentration of bacteria from blood using microfluidic membraneless dialysis and dielectrophoresis, D’Amico et al., 17, 2017; permission conveyed through Copyright Clearance Center, Inc.

5.3. Immunoaffinity-Based Methods

Immunoaffinity-based methods use recognition antibodies to specifically bind a target antigen on a cell surface. This method is desirable due to its unparalleled specificity. However, major drawbacks include high cost and the relative instability of antibodies. One example is described in a study by Wang et al., which reports the selective capture of E. coli using immobilized capture antibodies on the surface of the microchip ¹¹¹. By employing a biotinylated anti-lipopolysaccharide binding protein (anti-LBP) and NeutrAvidin-based surface chemistry, Wang et al. reports a limit of detection of 50 cells/mL, with a mean capture efficiency of 70.7% ¹¹¹. Following capture and wash steps, detection was carried out on chip via fluorescent tagging and imaging of E. coli ¹¹¹.

Immunomagnetic separation is one of the most commonly reported immunoaffinity-based separation methods used to isolate pathogens from whole blood. In this approach, magnetic particles are functionalized with antibodies to enable highly specific binding to a target antigen on the cell surface. The primary sample is incubated with these functionalized magnetic nanoparticles, and target cells (e.g. pathogens) are labeled and retained within the microfluidic chip (positive selection) by an external magnet. In one example, Cho et al. used immunomagnetic separation to isolate E. coli from whole blood, enabling on-chip DNA extraction using a centrifugal microfluidic device on a compact disc (CD) platform (Figure 6) ¹⁸. Specifically, primary biotinylated anti-E. coli antibodies were bound to 1 μm streptavidin-coated magnetic beads for E. coli isolation. Of note, incubation of the sample and functionalized magnetic particles was performed on-chip. At a starting pathogen concentration of 1 × 10⁵ cells/mL, capture efficiency was measured to be 93.4% ¹⁸. Following capture, the magnetic particles served a second function as micro-scale heaters. The enriched sample was exposed to an 808 nm laser. The conductive nature of the magnetic particle enabled rapid on-chip cell lysis, allowing for downstream (off-chip) molecular analysis ¹⁸. This entire sample enrichment and preparation process was reported to take less than 12 min ¹⁸.

Figure 6.

Immunomagnetic platforms for the isolation of bacteria from blood proposed by Cho et al ¹⁸. A. Image displays dual-functionality of bound magnetic particles (separation & micro-heaters). B. Schematic of on-chip functionality. C. Overview of CD centrifugal platform. Republished with permission of Royal Society of Chemistry from, One-step pathogen specific DNA extraction from whole blood on a centrifugal microfluidic device, Cho et al., 7, 2007; permission conveyed through Copyright Clearance Center, Inc.

A few studies have proposed the use of immunomagnetic separation of pathogens from blood as feasible strategy for sepsis therapy (i.e. blood cleansing). These studies are included because although the application is derivative from the current discussion, the same fundamental technical strategies of bacterial capture and isolation from a primary blood sample are explored. For example, Lee et al. designed a novel functional group that can non-specifically bind both Gram-negative and Gram-positive bacteria ⁶³. By modifying magnetic nanoparticles with a non-specific functional group, they demonstrated >95% clearance of E. coli at a starting concentration of 5 × 10⁶ cells/mL, and a flow rate of 60 mL/h ⁶³. Additionally, Xia et al. attempted the immunomagnetic enrichment of E. coli. In this case, the chosen capture antibody was specific to E. coli, and bound to 125 nm magnetic particles ¹¹⁴. The group observed that increasing magnetic bead concentration relative to pathogen concentration improved the separation efficiency. At 25 μL/h and a starting cell concentration of 5 × 10⁶ cells/mL, separation efficiency of E. coli from blood was approximately 78% ¹¹⁴. Yung et al. explored a similar approach, but with a different pathogenic target: C. albicans ¹¹⁸. Using 1 μm magnetic nanoparticles functionalized with antibodies specific to C. albicans, Yung et al. demonstrated 80% clearance of the pathogen from whole blood at a starting cell concentration of 10⁶ cells/mL ¹¹⁸. Lastly, Kang et al. engineered a broad spectrum human opsonin (mannose-binding lectin) to enable the non-specific capture of a Gram-positive bacteria, Gram-negative bacteria, and fungi. The in vitro results demonstrate >90% clearance of multiple pathogens (e.g. E. coli, S. aureus, C. albicans) ⁵⁰.

5.4. Inertial Focusing

Inertial focusing relies on three fundamental forces that exist in the microchannel environment: 1) channel wall interactions, 2) fluid drag, and 3) and lift ⁷⁷. These forces, and how they relate to one another, can be tuned by a variety of experimental factors including microchannel design, flowrate, and the physical and chemical characteristics of the liquid medium. With appropriate tuning, bacterial cells can be separated from other blood components. For example, Faridi et al. reported the inertial separation of E. coli (1 × 10⁶ cells/mL) from blood with 76% separation efficiency ²⁸, and Mach et al. demonstrated >80% margination of E. coli with a starting bacterial load of 1 × 10⁸ cells/mL ⁷². More specifically, Wu et al. explored an alternative approach that combines an asymmetrical sheath flow with a specifically designed channel geometry ¹¹³. This approach resulted in the generation of a soft inertial force proportional to the fluid Reynolds number, and a particle deflection distance proportion to the particle Reynolds number. In other words, larger particles were deflected away from the sample streamline, while smaller particles were retained near the original flow line. Using this design, Wu et al. reported a 62% separation recovery of E. coli ¹¹³.

Of note, is a study conducted by Hou et al. In this work, use of a spiral microchannel creates a lateral drag force due to Dean Flow fractionation, resulting in the separation of bacteria (outer channel wall) from red blood cells and white blood cells (inner channel wall) (Figure 7A)⁴⁰. Hou et al. evaluated this platform on four different bacterial species (E. coli, S. aureus, P. aeruginosa, Enterococcus faecalis) in order to adequately represent the different shapes and surface characteristics of sepsis-causing organisms⁴⁰. At bacterial concentrations as low as 10 cells/mL, bacteria were successfully recovered. Following bacterial isolation, the team reported on a detailed methodology for the downstream molecular profiling of the isolated pathogens, including antibiotic susceptibility testing (Figure 7B)⁴⁰. This study is notable due to its success in isolating a variety of representative sepsis-causing pathogens at clinically relevant concentrations, as well as its holistic consideration of the entire diagnostic workflow.

Figure 7.

Inertial separation of spiral microchannel as described by Hou et al⁴⁰. A. Spiral microchannel (left) and cross-sectional view of the sample at indicated points along the microchannel (right). B. Overview of proposed workflow for pathogen isolation, species identification, and antibiotic susceptibility profiling. Republished with permission of Royal Society of Chemistry from, Direct detection and drug-resistance profiling of bacteremias using inertial microfluidics, Hou et al., 15, 2015; permission conveyed through Copyright Clearance Center, Inc.

5.5. Adhesion-Based Methods

One of the more unique approaches to bacterial enrichment from blood has been pioneered by Hwang et al. ^{43, 44}. Their approach involved constructing an array of surface-modified silicon micropillars with surface conditions adapted to optimize bacterial adhesion. Specifically, silane compounds were coated on silicon micropillars. The silicon micropillar structure was employed in effort to maximize surface to volume ratio. The micropillars were 25 μm² with 12 μm pillar spacing, and the microchip had a total internal volume of a 5 μL ⁴⁴. When coupled with an optimized media pH, the capture efficiency of E. coli (1 × 10⁷ cells/mL) was about 40% in a 50% whole blood matrix ⁴⁴. It was also observed that during the washing step, E. coli were retained in the microchip, while remaining RBCs were washed out. Additionally, Hwang et al. demonstrated in situ DNA extraction. Following the capture and wash steps, PCR reagents were injected into the microchip and the mixture was moved to an external thermocycler to continue the analysis ⁴⁴. An overview of the experimental setup and an SEM image of the micropillar array is shown in Figure 8 ⁴⁴.

Figure 8.

Adhesion-based bacterial isolation from whole blood as described in by Hwang et al. ⁴⁴. A. Experimental setup. B. SEM images of functionalized silicon microarray. Reprinted with permission from Hwang K.-Y. et al. Bacterial DNA Sample Preparation from Whole Blood Using Surface-Modified Si Pillar Arrays. Analytical Chemistry 80: 7786–7791, 2008. Copyright 2008 American Chemical Society.

Overall, the method proposed and tested by Hwang et al. is effective in decreasing the blood volume in a highly concentrated E. coli sample. Sample preparation prior to molecular analysis took less than 1 hour, and detection sensitivity was improved by more than 100-fold. That said, given the high concentration of a bacteria in the initial sample, this platform has not yet demonstrated its utility in bacterial isolation and detection from a primary human blood sample. This technology could be a viable platform for rapid sample preparation for organism characterization following blood culture positivity, where concentrations of bacteria range from approximately 10⁷ to 10⁹ cells/mL ¹⁹.

6.0. Integrated On-Chip Pathogen Enrichment and Detection Platforms

To the best of our knowledge, there have been very few reports of integrated pathogen and enrichment and detection platforms. Many studies have come close, but still require some type of off-chip imaging and/or analysis. Below, we will highlight two studies that have successfully integrated on-chip pathogen enrichment and detection. (See Table 1 for additional integrated platforms). First, Cooper et al. designed a micro-device that employs an immunomagnetic isolation strategy, followed by an on-chip optical detection strategy ²⁰. Because the number of magnetic beads required is over 1000-fold the pathogen concentration, it can be challenging to optically detect pathogens captured immunomagnetically, as dense piles of bead-pathogen complexes form in the microchannel. Cooper et al. optimized the magnetic capture field to promote the formation of a homogenous layer of captured cells within the microchannel, allowing for automated optical detection (Figure 9A) ²⁰. Using a generic blood opsonin capture antibody, 98% percent of the fungal pathogens (C. albicans) were captured from initial starting sample concentrations ranging from 10 – 100 cells/mL. Further, by employing the optimized magnetic field concentrator, the optical detection rate increased from 43% to 67% ²⁰.

Figure 9.

Integrated on-chip isolation and detection platform. A. Overview of Cooper et al. Image of microchannel without (top) and with (middle) magnetic field concentrator (MFC). First half of channel is shown (scale bar = 50μm). Fluorescent image displaying captured C. albicans following immunofluorescent staining (bottom) ²⁰. Further, the magnetic field concentrator (MFC) minimizes magnetic bead clumping (middle), allowing for optical detection (bottom)²⁰. Republished with permission of Royal Society of Chemistry from, A microdevice for rapid optical detection of magnetically captured rare blood pathogens, Cooper et al., 14, 2014; permission conveyed through Copyright Clearance Center, Inc. B. Overview of Cai et al. integrated methodology. a) Schematic of workflow. b) relative size of microdevice (left) with zoomed in image of microchannel and preloaded microwells. c-f) cross-sectional view of device operation. Pathogen is retained in grooves, while blood cells (BCG) are washed away. Device is “slipped” to expose liquid PCR reagents to pathogen. Device is returned to original position avoid contaminated surface. Device is thermocycled in situ and imaged ¹⁵. Republished with permission of Royal Society of Chemistry from, An integrated microfluidic device utilizing dielectrophoresis and multiplex array PCR for point-of-care detection of pathogens, Cai et al., 14, 2014; permission conveyed through Copyright Clearance Center, Inc.

Second, Cai et al. developed an integrated microfluidic device that coupled dielectrophoresis with on-chip multiplex array PCR ¹⁵. Using dielectrophoresis as the separation method, E. coli were subjected to a positive dielectrophoretic force and were retained in grooves along the base of the microchip (Figure 9B) ¹⁵. At a starting sample concentration of 1.6 × 10⁷ cells/mL, E. coli capture efficiency ranged from 70.9% (10-fold diluted blood) to 91.5% (100-fold diluted blood) ¹⁵. Following capture, the remaining blood components were flushed out of the channel. The device was “slipped” to mix retained pathogens with pre-loaded PCR reagents. Following mixing, the device was returned to its original position to avoid unnecessary contamination. After in situ thermocycling, pathogen presence and identification was determined via a fluorescent readout ¹⁵. Although the isolation of other pathogens was not reported in the study, the on-chip multiplex array PCR system was also validated with Candida tropicalis, P. aeruginosa, S. aureus, and Streptococcus mutans ¹⁵. Although clinically relevant pathogen concentrations were not assessed in this work, Cai et al. creatively demonstrates the efficacy of integrated dielectrophoretic isolation and species-level pathogen detection.

7.0. Microfluidic platforms for downstream pathogen characterization

We would like to briefly highlight a few examples of microfluidic platforms that have demonstrated success in 1) pathogen identification, or 2) pathogen antibiotic susceptibility profiling. Following successful organism isolation from whole blood, these systems could have potential downstream utility in enabling comprehensive characterization of the disease-causing organism(s). As discussed, both species-level information and antibiotic susceptibility information are critical data needed to inform a targeted antibiotic therapy regimen, and to improve patient outcomes.

First, to obtain species-level information, molecular methods are widely employed for rapid classification. For example, Jiang et al. designed a rapid, continuous-flow PCR and DNA hybridization microfluidic platform for bacterial identification. The platform demonstrated successful identification of clinically relevant pathogens, such as E. coli, S. aureus, and P. aeruginosa⁴⁶. In another example, Sun et al. proposed a bacterial classification platform that circumvents the use of bacterial DNA. The group developed a microfluidic platform to differentiate bacteria as a function of cell wall structural differences. By employing a tapered channel design, cell stiffness could be determined as function of distance traveled in the microchannel (Figure 10A).

Figure 10.

Strategies for downstream pathogen identification and antibiotic susceptibility profiling. A. Basic bacterial differentiation principle proposed by Sun et al. The distance (d) traveled in the tapered microchannel is inversely proportional to cell stiffness. Force is drive by fluid pressure, p¹⁰⁵. Republished with permission of Royal Society of Chemistry from, A microfluidic platform for profiling biomechanical properties of bacteria, Sun et al., 14, 2014; permission conveyed through Copyright Clearance Center, Inc. B. Plug-based microfluidic platform proposed by Boedicker et al¹¹. Schematic of channel shows formation of nanoliter droplets containing bacteria, drug condition, and viability indicator. Republished with permission of Royal Society of Chemistry from, Detecting bacteria and determining their susceptibility to antibiotics by stochastic confinement in nanoliter droplets using plug-based microfluidics, Boedicker et al., 8, 2008; permission conveyed through Copyright Clearance Center, Inc.

Second, the growing problem of antibiotic resistance has led to increased research efforts surrounding the development of rapid antibiotic susceptibility profiling platforms⁸⁴. Boedicker et al. demonstrated on-chip antibiotic susceptibility testing by employing plug-based microfluidics to form nanoliter droplets containing a single cell, an antibiotic-of-interest, and a viability indicator¹¹. This platform allowed for the rapid determination and quantification of antibiotic efficacy and minimum inhibitory concentration (MIC), respectively (Figure 10B)¹¹. In another example, Kalashnikov et al. described a stress-based microfluidic platform for evaluating antibiotic susceptibility⁴⁸. Further, it was observed that shear flow potentiated the efficacy of antibiotics. In this system, bacteria were covalently-bound to the floor of the microchannel and exposed to the antibiotic-of-interest and shear flow. Bacterial viability was observed via automated fluorescent microscopy. This study reports comprehensive antibiotic susceptibility profiling of clinically-relevant Gram-negative bacteria within 2 h⁴⁸.

8.0. Conclusions and Future Work

In this review, we discussed emerging micro-scale platforms for the isolation of sepsis-causing pathogens from whole blood. The highlighted research represents a relatively new and emerging field, though the basic notion of employing microfluidic platforms for rare-cell capture from whole blood is not entirely new ⁹⁴. A large body of work exists describing rare-cell capture from whole blood for cancer screening, diagnosis, and monitoring. More specifically, the field of liquid biopsy describes numerous strategies for the capture and detection of circulating tumor cells (CTCs) and exosomes ^{3, 7, 26, 36, 39, 41, 45, 51, 63, 64, 68, 71, 100, 102, 106}. The existing and emerging technologies in this space should be considered for their potential translation to whole-cell capture of infectious organisms in whole blood samples.

Upon thorough consideration of the reported work in the area of sepsis-causing organisms’ isolation from whole blood, we would like to offer the following recommendations for future studies conducted in this space. First, clinically relevant bacterial concentrations must be evaluated. Given that the primary technical hurdle to sepsis diagnosis is the relatively low concentration of bacteria in the blood (1 – 10 CFU/mL), micro-scale systems must demonstrate their ability to detect bacteria at these low levels. As at minimum, experimental studies should report a limit of detection. Second, the field needs to move beyond non-specific organism isolation and detection, and towards a platform that can provide a comprehensive etiological characterization of the specific sepsis-causing organism(s). Both species-level information and antibiotic susceptibility information are critical data needed to inform a targeted antibiotic therapy regimen and improve patient outcomes. Creative methods that can provide more a more specific and multi-dimensional diagnostic output should also be explored. Finally, it is extremely important to design with downstream integration in mind. Although this paper focused on microfluidic capture strategies and platforms, some of the most attractive platforms successfully integrate organism capture and detection on a single chip. A holistic analysis of diagnostic workflow is critical for future clinical translation and commercial integration.