Author’s Accepted Manuscript Hybrid integration of scalable mechanical and magnetophoretic focusing for magnetic flow cytometry Mathias Reisbeck, Lukas Richter, Michael Johannes Helou, Stephan Arlinghaus, Birgit Anton, Ignas van Dommelen, Mario Nitzsche, Michael Baßler, Barbara Kappes, Oliver Friedrich, Oliver Hayden
PII: DOI: Reference:
www.elsevier.com/locate/bios
S0956-5663(18)30140-4 https://doi.org/10.1016/j.bios.2018.02.046 BIOS10308
To appear in: Biosensors and Bioelectronic Received date: 22 September 2017 Revised date: 19 February 2018 Accepted date: 20 February 2018 Cite this article as: Mathias Reisbeck, Lukas Richter, Michael Johannes Helou, Stephan Arlinghaus, Birgit Anton, Ignas van Dommelen, Mario Nitzsche, Michael Baßler, Barbara Kappes, Oliver Friedrich and Oliver Hayden, Hybrid integration of scalable mechanical and magnetophoretic focusing for magnetic flow cytometry, Biosensors and Bioelectronic, https://doi.org/10.1016/j.bios.2018.02.046 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Hybrid integration of scalable mechanical and magnetophoretic focusing for magnetic flow cytometry
Mathias Reisbecka,f*, Lukas Richtera, Michael Johannes Heloua, Stephan Arlinghausb, Birgit Antonc, Ignas van Dommelend, Mario Nitzschee, Michael Baßlerb, Barbara Kappesf, Oliver Friedrichf, Oliver Haydena*
a
In-Vitro DX & Bioscience, Department of Strategy and Innovation, Siemens Healthcare GmbH, GüntherScharowsky-Str. 1, 91058 Erlangen, Germany. b
c
microfluidic ChipShop GmbH, Stockholmer Str. 20, 07747 Jena, Germany.
d
e
Analysesysteme und Sensorik, Fraunhofer ICT-IMM, Carl-Zeiss-Str. 18-20, 55129 Mainz, Germany.
Sencio, Transistorweg 7, 6534 AT Nijmegen, Netherlands.
M2 Automation, Bessemerstr. 16, 12103 Berlin, Germany.
f
Institute of Medical Biotechnology, Department of Chemical and Biological Engineering, Friedrich-AlexanderUniversity Erlangen-Nürnberg (FAU), Paul-Gordan-Str. 3, 91052 Erlangen, Germany.
[email protected] [email protected]
*
Corresponding authors. Tel.: +49 151 40197121 (M. Reisbeck)
Abstract Time-of-flight (TOF) magnetic sensing of rolling immunomagnetically-labeled cells offers great potential for single cell function analysis at the bedside in even optically opaque media, such as whole blood. However, due to the spatial resolution of the sensor and the low flow rate regime required to observe the behavior of rolling cells, the concentration range of such a workflow is limited. Potential clinical applications, such as testing of leukocyte function, require a cytometer which can cover a cell concentration range of several orders of magnitude. This is a challenging task for an integrated dilutionfree workflow, as for high cell concentrations coincidences need to be avoided, while for low cell concentrations sufficient statistics should be provided in a reasonable time-to-result. Here, we extend the 1
spatial bandwidth of a magnetoresistive sensor with an adaptive and integratable workflow concept combining mechanical and magnetophoretic guiding of magnetically labeled targets for in-situ enrichment over a dynamic concentration range of 3 orders of magnitude. We achieve hybrid integration of the enrichment strategy in a cartridge mold and a giant-magnetoresistance (GMR) sensor in a functionalized Quad Flat No-Lead (QFN) package, which allows for miniaturization of the Si footprint for potential lowcost bedside testing. The enrichment results demonstrate that TOF magnetic flow cytometry with adaptive particle focusing can match the clinical requirements for a point-of-care (POC) cytometer and can potentially be of interest for other sheath-less methodologies requiring workflow integration.
Keywords:
Giant
magnetoresistance;
Sensor
integration;
Point-of-Care;
Flow
cytometry;
Magnetophoresis.
1. Introduction Functional cell diagnostics at the bedside requires an integrated and effortless workflow for results within a few minutes. However, the concentration of e.g. immune cells in a whole blood sample can vary over orders of magnitude, which is a significant challenge for point-of-care diagnostics (POC). With fluorescence flow cytometers, high-throughput is usually achieved with a sheath-flow based method and several preanalytical steps to minimize coincidences and background signal. The step towards POC, however, remains challenging, which is mainly due to the optical workflow, which requires the lysis of erythrocytes to bypass the optical opacity of whole blood and the removal of fluorescence excess markers by trained users (Shapiro, 2003; Greve et al., 2006; Einwallner et al., 2013; Robinson and Roederer, 2015). To overcome the resulting complexity of an optical read-out, alternative magnetic methods with non-optical detection of biological targets based on magnetoresistive sensing mechanisms with either static (Osterfeld et al., 2008; Schotter et al., 2009; Gaster et al., 2011; Lin et al., 2014; Bechstein et al., 2015; Huang et al., 2017) or dynamic approaches (Loureiro et al., 2011; Issadore et al., 2012; Melzer et al., 2012; Helou et al., 2013; Fernandes et al., 2014; Kim et al., 2015; Reisbeck et al., 2016) have emerged (Lin et al., 2017). Herein, immunologically
2
functionalized magnetic particles at the nanoscale replace fluorescent labels used in optical flow cytometry (Lee et al., 2015). Thus, magnetic flow cytometers for single cell detection overcome the restrictions of optical flow cytometers, such as absorption of the scattered light by the cellular background or the need for highly diluted samples, due to the inherently low magnetic moment of biological samples (Issadore et al., 2014). Despite the great potential for miniaturization and integration of highly sensitive magnetoresistive elements, the step towards an integrated device for quantitative POC applications is challenged by two aspects. First, the integration of the fluidic transport and particle enrichment onto a cm² Si chip containing the sensing elements prevents low-cost, disposable bedside testing (Loureiro et al., 2011; Issadore et al., 2012; Melzer et al., 2012; Helou et al., 2013; Fernandes et al., 2014; Kim et al., 2015; Reisbeck et al., 2016). Second, similar to fluorescence flow cytometry (Keji et al., 1991; Wersto et al., 2001) and impedance sensing (Wynn and Hounslow, 1997; Kammel et al., 2012), quantitative magnetic information on a single particle level is masked by coincidences occurring at particle concentrations exceeding an instrument specific threshold. Thus, without prior knowledge of the particle concentration, an additional dilution step of the sample becomes necessary to avoid coincidences, while sufficient statistics should be provided for analysis of samples with low cell concentration. In previous proof-of-concept studies the fluidic interconnection of mm²-sized sensor chips to peripheral fluidic components has been presented using rather complex fabrication and alignment strategies of multiple Polydimethylsiloxane (PDMS) layers (Wu et al., 2010; Muluneh and Issadore, 2014). However, effortless hybrid integration of an all-magnetic workflow incorporating the singulation of magnetic targets in a clinically-relevant concentration range covering various log-scales into an injection-molded low-cost fluidic platform for quantitative POC testing has not been explored yet. In our previous work we discussed a quantitative magnetic flow cytometry approach, which utilizes cell rolling for precise volumetric measurements and analysis of immunomagnetic binding capacity on a single particle level (Reisbeck et al., 2016). However, quantitative analysis of the magnetic fingerprint of single particles is limited due to the spatial resolution of the sensor and thus coincidences at high particle concentrations. At low particle concentrations, the time-to-result for sufficient statistics impedes bedside application. Here, we extend our previously reported magnetic flow cytometry workflow by an adaptive in-situ enrichment and present effortless hybrid integration of an all-magnetic workflow. We combine in-situ enrichment by means of mechanical chevrons and magnetophoretic forces to cover variable particle concentrations without the need for sample dilution steps and
3
thus buffer reservoirs on a cartridge. We designed and integrated a functional package housing a magnetoresistive half-bridge sensor in a credit card-sized injection molded device, which comprises the mechanical focusing structures for adaptive in-situ enrichment. Furthermore, we show that with time-of-flight (TOF) probing of the magnetic fingerprint, accurate quantitative information on a single particle level can be obtained. Once mechanical and magnetophoretic focusing and TOF analysis is performed, magnetic particles are captured in an integrated cavity and can be extracted with diminished background for subsequent analysis.
a)
b)
Ffluid GMR halfbridge 8 µm / 16 µm / 36 µm
Time (s)
8 µm / 16 µm / 24 µm
120
8 µm / 16 µm / 16 µm
180
8 µm / 16 µm / 8 µm
120
80
80
40
40
60
0
0
0
0
-40
-40
-60
-40
-80
-80
-120
-120
Sensor signal (µV)
Sensor signal (µV)
120
80
120
40
-120
-80
-180 -40 -20
0
20 40 60 80
-40 -20
0
20 40 60 80
-120 -40 -20
0
20 40 60 80
-40 -20
0
20 40 60 80
x (µm)
Fig 1 | Detection and workflow schematics of an integrated magnetic flow cytometer with adaptive particle focusing. (a) Detection of magnetic targets in an external magnetic field with a Wheatstone half-bridge comprising two GMR sensors. Single particle information is derived from the characteristic four-peak magnetic fingerprint originating from a single magnetic particle (black) by TOF analysis. (b) Magnetic sensor signals are simulated for two 8 µm particles passing the GMR resistors sequentially. For 16 µm between the GMR elements, signal interference is observed for a particle-particle distance of 8 µm (purple), 16 µm (green), and 24 µm (blue). However, for 36 µm (red), two distinct characteristic four-peak signal patterns (surrounded by dashed boxes) enabling TOF analysis are observed.
2. Experimental 2.1 Detection mechanism and spatial resolution of the quantitative magnetic flow cytometer. The schematic in Fig. 1a illustrates the detection principle of our magnetic flow cytometry approach, allowing for quantitative non-optical probing of the magnet stray field of single magnetic particles in a highly reproducible manner. The magnetoresistive sensor is incorporated into a microfluidic channel and consists of two 2 × 30 µm² giant magnetoresistance (GMR) sensors arranged transversely to the laminar flow direction of the magnetic particles 4
in a Wheatstone half-bridge configuration. The sensor matrix is positioned precisely over the center of a permanent magnet, which pulls the magnetic particles towards the substrate surface and magnetizes the target particles orthogonal to the sensor plane (Hayden et al., 2016). For a single magnetic particle rolling over the sensor halfbridge at effectively zero distance a characteristic four-peak pattern is observed. By analyzing the normalized integral of the sensor signal, given by the sum of the absolute integral of each half-wave, we derive quantitative single particle information, such as volumetric information and immunomagnetic binding capacity. For TOF calculations, we analyze the characteristic sensor signal within its margins, defined as the starting and end positions where the signal increases above its baseline value by 25 % and then drops to 25 % of its maximum amplitude. Hence, the spatial resolution of the sensor is defined by the dimensions of the signal pattern arising from a single magnetic particle. We evaluated this quantitative approach with respect to cross-sensitivities by numerical simulation of the signal sequence arising from two magnetic particles with a hydrodynamic diameter of 8 µm passing two GMR resistors separated by dGMR=16 µm for varying distances between the particles (Fig. 1b). The sensor signal for a single magnetic particle can be described by the mean dipole field component B in the sensor plane detected by the resistors
(
)
∫
∫
(
| |
)
(1)
where AGMR is the area of a single resistor, mz is the magnetic moment of the magnetic particle magnetized in the zdirection and r(x,y,z) is the position vector of the particle center (Jackson, 1999). For a distance between the particles of 36 µm (highlighted in red in Fig. 1b), two completely separated characteristic four-peak signal patterns are detected by the sensor. The normalized integral can then be calculated for each signal pattern to perform precise TOF analysis on a single particle level. A decrease of the distance between the magnetic targets to 24 µm (highlighted in blue in Fig. 1b), 16 µm (green), and 8 µm (purple) results in six- or eight-peak signal patterns, which is due to the particle-particle distance falling below the spatial resolution of the sensor half-bridge. Thus, the interference of two adjacent signal patterns does not allow to derive quantitative single particle information.
5
a)
t0 Side view
t0 Section 4 (L4,h4,A4) Section 3 (L3,h3,A3)
t0 Top view
z
Q0 >> 20 µl s-1
Section 2 (L2,h2,A2) Section 1 (L1,h1,A1)
y
x
GMR sensor
t1 Side view
t1
Fmag
Section 4 (L4,h4,A4)
t1 Top view
Section 3 (L3,h3,A3) Section 2 (L2,h2,A2) Section 1 (L1,h1,A1) GMR sensor
t2 Side view
t2 Section 4 (L4,h4,A4)
t2 Top view
Section 3 (L3,h3,A3)
Q2 ~ 1.5 µl s-1
Section 2 (L2,h2,A2) Section 1 (L1,h1,A1) GMR sensor
Detection rate (s-1)
b)
t0
t1
t2
Time (s)
Fig 2 | All magnetic workflow for adaptive focusing by magnetophoretic and mechanical means. (a) The spatial bandwidth of the sensor can be overcome by adaptive focusing of the particles to ensure sufficient particle-particle distance for TOF analysis by varying channel height and substrate coverage with chevron patterns in a microfluidic channel guiding the particles to the GMR sensor (same color code as in Fig. 1). A permanent magnet placed underneath the microfluidic channel enables an all-magnetic integrated workflow. First, the microfluidic channel is filled with the particle suspension at a high flow rate (Q0, t0). Second, the particles are pulled onto the substrate surface by the magnetophoretic force of the permanent magnet at zero flow rate (Q1, t1). Last, the in-situ focusing is performed by balancing magnetic and fluidic drag forces (Q2, t2). (b) Increasing detection rates with increasing surface fraction patterned with chevrons for particle enrichment. For TOF signal analysis, the section providing sufficient particleparticle distance is evaluated.
6
2.2 Methodology and workflow for adaptive particle focusing by mechanical and magnetophoretic means. To overcome the limitation due to the interference of the stray fields of two adjacent magnetic particles without preanalytical dilution steps, we developed an adaptive microfluidic enrichment and focusing concept. To ensure sufficient particle-particle distance we integrate the GMR sensor into a three-dimensional (3D) patterned microfluidic channel, which guides the magnetic targets to the GMR sensor (Fig. 2a). We use mechanical chevron structures which cover a defined fraction A of the substrate surface for mechanical focusing of the rolling magnetic particles to the detection area. We further divide the microfluidic channel into N=4 enrichment sections with varying height hn, length Ln, and surface fraction An to control particle number and particle-particle distance. For our all-magnetic workflow, the microfluidic channel is positioned over a permanent magnet covering the entire channel area. First, the microfluidic channel is filled with the sample at a flow rate Q0 >> 20 µl s-1 (t0 in Fig. 2a). Second, the flow rate is set to zero and the magnetic particles, which are stochastically distributed in the channel volume, are pulled onto the substrate surface of the channel by the magnetophoretic force Fmag exerted by the permanent magnet positioned underneath the channel (t1 in Fig 2a). For a continuous stream of enriched particles rolling over the GMR sensor, the flow rate is set to Q2 to balance magnetophoretic and fluidic forces and to ensure sufficient drag in all four enrichment sections (t2 in Fig. 2a). In this way, the rolling of magnetic particles from predefined fractions of the substrate surface leads to continuously aligned particles like pearls on a chain in the center of the microfluidic channel. With the fraction of the substrate surface area An used for enrichment only a pre-defined percentage of the magnetic particles is focused towards the GMR sensor, which allows to control particle-particle distance for single particle TOF magnetic sensing. The magnetic particles enriched on the substrate surface outside An travel past the sensor without being detected (Suppl. Vid. 1†). Furthermore, for each section, the total number of particles Mn to be focused to the GMR sensor can be controlled by the height hn and the length Ln of the section. For a continuous measurement, the sections are analyzed in ascending order of An. Fig. 2b qualitatively shows the expected increase in the detection rate of the GMR sensor with the expansion of the focusing chevrons.
7
a)
b) 4 1 3 3
4
2
GMR sensor
1
2
c)
d) dparticle,TOF
# Section
5 7 µl-1 - 92 µl-1
4
40 µl-1 - 230 µl-1
3 2 182 µl-1 - 1,100 µl-1
610 µl-1 - 3,450 µl-1
1
1E+01
1E+02
1E+03
1E+04
Particle concentration (µl-1)
e)
dhydrodynamic
5 7 µl-1 - 276 µl-1
# Section
4
40 µl-1 - 690 µl-1
3 2 182 µl-1- 3,300 µl-1 1
610 µl-1 - 10,350 µl-1
1E+01
1E+02
1E+03
1E+04
Particle concentration (µl-1)
Fig 3 | Microfluidic guiding and enrichment in a cartridge-based format. (a) Integration of the focusing and detection parts (red rectangle) into a credit-card sized format. The cartridge additionally houses a female Luer interface for the connection of a syringe (1), blister socket (2), incubation chamber for immunomagnetic labelling (3), and waste compartment (4). For the presented allmagnetic workflow, the cartridge is arranged over a NdFeB permanent magnet, indicated by dashed brown lines, with the integrated GMR sensor (blue rectangle) over the center of the magnet. (b) The microfluidic channel is designed as a meander and separated into four enrichment sections covering different particle concentrations. The different sections are color-coded for visualization. (c) Trajectory analysis of the focusing mechanism on the chevron structures with 6 µm magnetic beads. Scale bars are 500 µm. Particle concentrations covered by the different microfluidic sections for (d) TOF analysis and (e) sole detection and counting of magnetic particles.
2.3 Implementation of the mechanical focusing concept in a credit card-sized cartridge format. For our integrated magnetic flow cytometer with adaptive particle focusing, we chose an injection-molded cartridge of a standardized credit-card format with dimensions of 85 x 55 mm² (Fig. 3a). For POC applications, the device additionally houses a female Luer interface being compatible with standard syringes commonly used for blood gas analyzers, a socket for a blister containing functionalized nanoparticles, an incubation chamber for an integrated
8
immunomagnetic labelling step, a cavity for subsequent magnetophoretic sorting of the magnetic targets, and a waste compartment for effortless handling of potentially infectious samples. In a first step, we defined the position of the GMR half-bridge at the center of a Neodymium Iron Boron (NdFeB) permanent magnet (IBS magnet) with dimensions of 75 × 50 × 10 mm 3 to avoid interfering in-plane field components for highest sensitivity of the device (Hayden and Rührig, 2012). To implement the adaptive enrichment concept illustrated in Fig. 2a we conceptualized the cartridge as a twopiece injection-molded device using Cyclo Olefin Polymer (Zeonor). In detail, the top mold defines the microfluidic channel, while the bottom mold comprises the chevron structures for mechanical guiding of the magnetic particles (Suppl. Fig. 1†). The 3D patterned microfluidic channel is divided into N=4 enrichment sections (n=1…4) with a total length of ~ 29 cm over the 3 × 3 cm² center area of the permanent magnet (Fig. 3b). We tune the particle-particle distance for magnetic TOF analysis by varying height hn and channel length Ln of each enrichment section, as well as the surface fraction An patterned with focusing chevrons. We define these layout parameters to cover a wide dynamic concentration range to match a typical range of immune cell concentrations in chemotherapy or healthy patients. In detail, the lower limit of the particle concentration displayed by section n is defined by the minimum number of particles to be detected by the sensor, which we define to be 500 for significant statistics. The upper bound of the particle concentration in each section is given by the length of the chain of the focused particles Lchain,n, which may not exceed the length Ln of the respective section. Lchain,n can be calculated for each section by multiplying the number of particles Mn contained in An by the minimum particle-particle offset dparticle,TOF required for quantitative TOF analysis, which is defined by
(2)
where dGMR,min is the minimum distance of the GMR resistors forming a characteristic four-peak pattern and ddipole,max is the dimension of the dipole field to be evaluated for quantitative single particle information. First, we define the number of particles to be focused to the sensor in each section by varying channel height hn and channel length Ln. Therefore, we design section 1 (highlighted in red in Fig. 3b) with L1 = 68.5 mm and section 2 (blue) with L2 = 51.2 mm, and a height of the channel with h1 = 150 µm and h2 = 200 µm, respectively. For
9
sections 3 and 4 (green and black) with L3 = 53.8 mm and L4 = 117.9 mm we increase the channel height to h3 = h4 = 300 µm. The width w of the channel is kept constant with 1,500 µm. Second, we further control particle-particle distance by varying the surface faction An patterned with mechanical guiding structures. First, section 1 exhibits the lowest surface area covered by chevron structures with A1=0.053 focusing the lowest particle fraction to the sensor, whereas in section 2 12.6 % of the surface are occupied by chevron patterns (A2=0.126). Second, we further increase the surface area to be focused to the sensor in section 3 to A3=0.4. Last, we set A4=1 resulting in all particles contained in section 1 being focused to the GMR half-bridge sensor. The minimum height of the chevron structures hchevron required for a reproducible mechanical guiding of the magnetic particles on the substrate surface was evaluated to be 40 µm by chevron patterns with different heights structured by standard photolithography (Suppl. Fig. 2†). The width of the single guiding structures was set to 50 µm which correlates to the resolution of the manufacturing process of the master for the injection mold. The guiding structures in sections 2-4 were positioned with an angle of 15° relative to the laminar flow direction of the magnetic targets and arranged as pairs of two opposing structures, whereas in section 1 the angle of the chevrons was lowered to 12° and patterned sequentially on the substrate surface. A precise focusing and continuous rolling of the magnetic particles to the sensor matrix along the chevrons was optically evaluated using Diatrack® trajectory analysis (Fig. 3c) (Valloton and Olivier, 2013). For the applied flow rates in the µl s-1 regime we observe smooth rolling of the particles around the edges of the mechanical chevrons and thus have no indication of turbulent flow in our experimental settings. Fig. 3d shows the particle concentrations for each enrichment section for the presented layout of the microfluidic channel and magnetic particles with a hydrodynamic diameter of 8 µm with ddipole,max=9 µm and dGMR,min=6 µm, both derived from numerical simulation using eqn. (1). The overlap of the concentration range of two adjacent sections fulfils the requirement for a continuous sampling of the particle concentration in a range between 7 particles µl-1 and 3,450 particles µl-1. In this simulated experiment the minimum particle-particle distance is greater than dparticle,TOF and thus enables quantitative probing of the magnetic fingerprint of each single particle (cf. Fig. 1b). We also mapped the concentration range of our adaptive focusing concept for applications which solely require counting of the magnetic particles rolling in direct contact over the sensor, where the upper limit of the concentration can be increased to 10,350 particles µl-1 (Fig. 3e). Here, the minimum particle-particle distance is defined as the
10
hydrodynamic diameter dhydrodynamic of the particles, which prevents TOF analysis due to coincidences detected by the sensor, but still would enable counting of the particles (cf. Fig. 1b). Hence, the lower bound of the simulated concentration range for each section is defined by the minimum number of particles to be detected by the sensor and is thus constant in Fig. 3d and Fig. 3e. However, the upper bound of the concentration range of each section is limited by the particle-particle distance and can be increased by a factor of ~3 due to dparticle,TOF ~ 3 × dhydrodynamic for applications requiring only detection and counting of the particles. We designed our device in a way, that for TOF analysis and particle concentration measurements not all enrichment sections but only K enrichment sections, which provide sufficient particle-particle distance will be taken into account. Specifically, this means to exclude coincidences (cf. Fig. 1b), by keeping the particle-particle distance either > dparticle,TOF for applications requiring TOF analysis (Fig. 3d) or > dhydrodynamic for sole detection (Fig. 3e), respectively. The final particle concentration c can then be derived from
∑
(3)
where Mtotal is the number of particles to be focused to the sensor from the K sections providing sufficient particleparticle distance (color-coded particles in Fig. 2a). An × Ln × hn × w is the volume fraction to be focused to the sensor in the respective section.
11
a)
b)
c)
e)
Sensor Die attach Lead frame Adhesive layer Cartridge bottom mould
10
384
8
378
6
372
S0 = 1.2 %/mT MR = 9.84 % R0 = 367.7 2
366
4
GMR sensor Packaged sensor Cartridge prober Cartridge reader
0 -10
-5
0
5
360 354
Resistance ()
Cartridge top mould
GMR effect (%)
d)
348
10
Magnetic flux density (mT)
Fig 4 | Hybrid integration and characterization of a Wheatstone half-bridge GMR sensor with magnetophoretic focusing. (a) The Si die containing the GMR sensor (highlighted in yellow) is housed in a QFN package and positioned next to the microfluidic channel. (b) Additional enrichment structures (surrounded by red rectangle) patterned on the package guide the magnetic targets to the sensor die. Scale bar is 2 mm. (c) Ferromagnetic structures on the Si chip for magnetophoretic cell focusing to the GMR halfbridge. Optimally focused magnetic particles are false-colored in red for visualization. Scale bar is 300 µm. (d) The sensor package and top mold of the cartridge form the microfluidic channel for magnetic detection under laminar flow conditions. (e) Hysteresis curves of the GMR sensor at different points of the integration process. Sensor sensitivity is evaluated at the operating point of the transfer curve (inset).
2.4 Hybrid integration of the GMR sensor half-bridge in the injection-molded device. We designed sensor devices with dimensions of 2 × 2 mm² containing ferromagnetic guiding patterns and a single Wheatstone half-bridge with either 18 µm or 22 µm between the 2 × 30 µm² GMR resistors. Passivation of the sensor was achieved with a pinhole-free 70-nm SiN layer. The singulated chips are positioned on a lead frame by a standard die attach process and are wire-bonded to the lead frame for electrical connection to the peripheral read-out system. The sensor is housed in a Quad Flat No-Lead (QFN) package formed via transfer molding. The mold compound was tested with respect to biocompatibility with whole blood samples to avoid cell aggregation and clotting effects. For the transfer molding process, a 1.5 × 1.5 mm² area on the Si die containing the GMR half-bridge
12
and ferromagnetic chevrons for magnetophoretic focusing is masked by a polymer stamp to later expose the sensor device to the sample at minimal sensor-to-analyte offset for highest sensitivity. For hybrid integration of the sensor device, we positioned the QFN package in a pocket implemented in the bottom mold of the cartridge subsequent to the focusing structures (Fig. 4a). Thereafter, the top and bottom mold of the cartridge were aligned and a permanent bond was formed. To compensate for an assembly offset between the package and the chevron patterns for in-situ guiding of the magnetic particles on the cartridge, we functionalized the QFN mold compound with additional mechanical focusing elements at an angle of 12° to the laminar flow direction covering a total width of 300 µm (Fig. 4b). Hence, the package for the GMR sensor has an integrated functional element. We further implemented ferromagnetic NiFe structures on the Si die to deterministically guide the enriched magnetic particles to the center of the 30-µm GMR resistors by magnetophoretic means (Fig. 4c) (Inglis et al., 2006). The fishbone-like structures have a width of 10 µm and are patterned at an angle of 5.7° relative to the laminar flow direction of the magnetic particles. We set the total width of the ferromagnetic chevrons to 150 µm, allowing for compensation of an offset induced by positioning of the sensor chip on the lead frame. By combining mechanical adaptive focusing on the cartridge and magnetophoretic guiding on the sensor chip, we achieve an enrichment factor of up to 3,000 from section 4 on the mold with a channel cross section of 1,500 × 300 µm² (width w × height h4) to a virtual string of aligned magnetic particles in a cross section of 15 × 10 µm² (width × height) at the exit position of the ferromagnetic patterns upstream to the GMR sensor (Suppl. Video 2†). To avoid any barriers for rolling magnetic particles after the assembly, we designed shallow ramps on the package surface (Fig. 4d).
2.5 Peripheral components and sensor characterization. For the electrical connection and characterization of the GMR half-bridge we designed a miniaturized read-out instrument, which comprises the peripheral components for the all magnetic workflow, including a slot for the uptake of the cartridge, the NdFeB permanent magnet, magnetic shielding, contact pins for electrical read-out of the half-bridge, and a pulsation-free syringe system (Suppl. Fig. 3†). For the characterization of the integrated magnetic flow cytometer, we used a customized Helmholtz coil setup providing an alternating magnetic field density between ± 10 mT. The NdFeB permanent magnet implemented in the read-out instrument is positioned relative to the GMR
13
sensor half-bridge in the coil setup (Suppl. Fig. 4†). This way, we minimize magnetic field components of the external permanent magnet in the sensor plane to ensure highest sensitivity for single particle detection. The transfer characteristics of the integrated magnetoresistive sensor on a cartridge-level with a sensitivity of 1.2 % mT -1 and a GMR effect of 9.84 % match the specifications of the bare Si die and confirm the stability of the sensor during the integration process (Fig. 4e).
2.6 Signal read-out and TOF analysis. For TOF measurements the sensor is supplied with an AC modulated signal with a frequency of 10 kHz and a peakto-peak voltage of up to 400 mV. The differential signal of the half-bridge is fed into a single-phase lock-in setup (Abacus Analytical Systems GmbH), which compensates for a DC offset of the differential signal and amplifies the signal 30,000-fold. The signal is then averaged with a time constant of 580 µs with a roll-off of 12 dB/octave and fed into an A/D converter, which digitizes the sensor signal at a sample rate of 20 kHz and a resolution of 16 Bit. Signal identification is performed with a customized state-event-machine-based algorithm smoothing the digitized sensor signal with a FIR filter and probing the output signal stream for the characteristic four-peak sequence originating from a single particle passing the sensor half-bridge (cf. Fig. 1a). TOF analysis of the sensor signal is carried out according to our previous approach (Reisbeck et al. 2016).
2.7 Reference microspheres for proof-of-concept experiments. For validation of the integrated magnetic flow cytometer and magnetophoretic cell sorting, we obtained magnetic monodispersed reference microspheres with an average hydrodynamic diameter of 6 µm as well as 12 µm and a coefficient of variation < 10 % consisting of a polystyrene core with notches in its shell filled with superparamagnetic iron oxide (micromer®-M, micromod Partikeltechnologie GmbH). The magnetization of the particles was characterized with a vibrating sample magnetometer, showing a mean magnetic moment of 7.84 10-13 A m² for a single magnetic particle. The magnetic moment and the hydrodynamic diameter of the reference beads are comparable to the values expected for immunomagnetically-labeled analytes and are therefore chosen for characterization of our device (Suppl. Fig. 5†) (Downey et al., 1990). 14
GMR sensor
Normalized detection rate (a.u.)
a)
18 16 14 12 10 8 6 4 2 0
0
500 1000 1500 2000
Time (s)
b)
Fig 5 | Optical validation of the mechanical enrichment process. (a) Validation of the detection rate by time-lapse of magnetic particles with a mean hydrodynamic diameter of 6 µm on the final enrichment structures (highlighted by yellow box). (b) Dark field microscopy captures the increasing number n of focused beads (pseudo-colored for visualization) with time. The channel part containing the focused magnetic targets is marked by dashed lines. Scale bars are 500 µm.
3. Results 3.1 Optical validation of the adaptive mechanical focusing. To validate the functionality of our integrated magnetic flow cytometer, we explored the adaptive focusing by optical microscopy. To do so, we diluted monodispersed polystyrene particles with a magnetic shell and an average hydrodynamic diameter of 6 µm in 1 % bovine serum albumin (BSA) (w/V) to a final concentration of 200 particles µl-1. The external forces applied to the microfluidic system - flow rate and magnetic field density were set to 1.5 µl s-1 and 130 mT. We defined a region of interest at the final enrichment structures on the cartridge and performed time-lapse microscopy in dark-field mode over a time frame of 1,800 s. (Fig. 5a). Analysis of the increase in the detection rate was carried out by manually counting the number of particles per capture and normalizing the detection rate with the value at the starting time defined at 0 s. With the expansion of the mechanical chevrons the normalized detection rate of the device showed an increase over time by a factor of ~ 15
15
(cf. Fig. 2b). Furthermore, we analyzed dark-field microscopy captures, exemplarily shown in Fig. 5b, with respect to the mean particle-particle distance d, which is found to decrease from ~ 1,100 µm at 0 s to ~ 76 µm at 1,800 s. Despite d being greater than the minimum distance dparticle,TOF between adjacent particles required for magnetic TOF analysis, the dark-field capture at 1,800 s shows a significant number of magnetic particles that have formed aggregates and thus cannot be analyzed on a single particle level by the GMR sensor (cf. Fig. 1b).
a)
Sensor signal (V)
1.0 0.5 0.0 -0.5 -1.0 -1.5 0.00 0.25 0.50 0.75 1.00 1.25 1.50
Normalized frequency (a.u.)
b) 1.5
1.2
0.8 0.6 0.4 0.2 0.0
6
8
10 12 14 16 18 20
Hydrodynamic diameter (µm)
Time (s)
d)
Sensor signal
Detection rate (Particles / 30 s)
c)
Time
e)
110 100 90 80 70 60 50 40 30 20 10 0
0
500
1000 1500 2000 2500
Time (s)
f) 0.5
0.4
0.3
0.2
0
500
1000
Time (s)
1500
2000
Normalized frequency (a.u.)
Signal overlap (a.u.)
Magnetic flow cytometer Impedance measurement
1.0
1,2 Sections 1-2 Sections 1-4
1,0 0,8
538 Events µ=12.99 µm sigma=1.67 µm
0,6 0,4 0,2 0,0
0
5
10 15 20 25 30 35
Hydrodynamic diameter (µm)
Fig 6 | Proof-of-concept of the magnetic TOF analysis and focusing methodology. (a) Reproducible characteristic four-peak patterns of single magnetic reference particles enable TOF analysis. (b) Quantitative volumetric analysis of monodispersed 12-µm particles by TOF analysis compared to a Coulter Counter, the gold standard for volumetric measurements. (c) Stability of the GMR sensor in a continuous signal stream over > 10 min. and (d) magnetic evaluation of the detection rate within the four enrichment sections. (e) Analysis of the signal overlap to quantify the increasing rate of particle aggregates detected by the sensor (surrounded by red oval). (f) Comparison of a quantitative volumetric measurement of 220 particles µl-1 in sections 1-2 and 1-4 highlights the adaptive focusing concept ensuring single particle detection and quantitative volumetric analysis.
3.2 Quantitative volumetric TOF sensing and magnetic verification of the in-situ enrichment.
16
We thus verified the importance of the integrated adaptive focusing concept for a quantitative probing of the magnetic fingerprint of each single particle by magnetic TOF analysis of monodispersed reference beads with a mean hydrodynamic diameter of 12 µm and a magnetic shell in 1 % BSA (w/V) at a final concentration of 220 particles µl-1. The flow rate of our device is tuned with respect to the hydrodynamic diameter and magnetic moment in order to balance fluidic drag and magnetophoretic forces acting on the magnetic particles. With a flow rate of 1.2 µl s-1 and a magnetic field density of 130 mT, we ensure precise focusing of the magnetic particles on the enrichment structures on the cartridge and the ferromagnetic patterns on the integrated Si die. In this way, we further achieve defined rolling of the magnetic particles on the sensor surface and consequently highest signal-to-noise ratio. For the magnetic measurement we chose a sensor half-bridge layout with a distance of 18 µm for the magnetoresistive elements. The reproducibility of the characteristic four-peak signal with a SNR > 5 confirms the highly defined rolling of the particles over the center of the half-bridge allowing for precise TOF analysis of each single magnetic pattern (Fig. 6a). The mean velocity of the magnetic particles passing the sensor was calculated to be ~ 613 µm s-1. For experimental validation of our quantitative approach, we acquired 200 single events and analyzed the normalized integral of the signal pattern to derive the hydrodynamic diameter of the magnetic beads (Fig. 6b). By fitting the frequency distributions with a Gaussian distribution (R² > 0.85), we derived the mean values and standard deviations of the particle populations. As a quantitative means of the accuracy of the measurement, we explored the coefficient of variation with 8.3 %, which corresponds to the manufacturer’s specifications of < 10 %. We furthermore analyzed the same sample with a Coulter instrument (Beckman Z2, Beckman Coulter) as the gold standard for volumetric particle analysis, which was in good correlation with our integrated magnetic methodology. Next, we evaluated the adaptive focusing concept implemented on the cartridge mold with respect to magnetic TOF sensing. From the continuous signal stream depicted in Fig. 6c an increase in the detection rate can be identified. We quantified the functionality of the presented adaptive focusing concept by deriving the detection rate from the measurement stream with a resolution of 30 s (Fig. 6d). The detection rate, which was smoothed with an 11-point moving average filter, shows the expected increase consisting of four plateaus representing the individual enrichment sections implemented on the cartridge. The minimum value of 11 particles per 30 s is observed in the section nearest to the sensor with its focusing area A1 covering 5.3 % of the substrate surface, while the maximum is found to be 92 particles per 30 s for section 4 with A4 covering 100 % of the substrate surface. This increase in the
17
detection rate is found to be 50 % lower than the value obtained from optical validation and is attributed to the detection of coincidences (cf. Fig 5a). We tested this hypothesis by quantifying the contribution of the absolute integral of the outer signal half waves to the absolute integral of the sensor signal. This signal overlap describes the simultaneous detection of the magnetic stray field by the two resistors and consequently increases with the formation of aggregates of magnetic particles. While the major population of the particles exhibits a mean value for the signal overlap of 0.32, an additional minor population is found with a mean of 0.4, indicating greater signal overlap. The time values of the respective signal patterns can be attributed to the sections 3 and 4 focusing the highest number of particles to the sensor (Fig. 6e). In detail, for the sections 1 and 2 the lower surface fractions enriched to the sensor (A1=0.053 and A2=0.126) ensure sufficient particle-particle distance and precise enrichment of single particles for accurate TOF analysis. However, in sections 3 and 4 (A3=0.4 and A4=1) the particle-particle distance exceeds the spatial resolution of the sensor half bridge, which results in the detection of coincidences masking TOF analysis. Thus, for the given particle concentration of 220 particles µl-1, only the magnetic signals originating from sections 1 and 2 nearest the sensor (K=2) are analyzed with respect to volumetric single particle information, while sections 3 and 4 are excluded from TOF analysis due to the detection of coincidences indicated by the increased signal overlap (cf. Fig. 3d). To illustrate the difference between enrichment of rolling single particles in sections 1 and 2 and coincidence detection in sections 3 and 4, we evaluated the hydrodynamic diameter for signals within a time span of 0 s - 650 s, which can be attributed to sections 1 and 2 (cf. Fig. 6d), and compared it to the diameter distribution of the complete signal stream 1-4 (Fig. 6f). While the distribution of the diameter of particles from sections 1 and 2 (Mtotal=538) providing sufficient particle-particle distance for TOF probing due to surface patterning shows the expected Gaussian distribution with a mean of 12.99 µm and a standard deviation of 1.67 µm, the distribution originating from the complete signal stream (sections 1-4) shows additional peaks at 20 µm and 30 µm indicating coincidences. Thus, enrichment sections 3 and 4 are excluded from quantitative TOF analysis. These results emphasize the importance of the adaptive focusing for quantitative volumetric TOF probing of single particles by fractioned surface patterning for controllable particle-particle distance.
18
a)
c)
b) FFdrag fluid Fmagnetophoretic Fmag
d)
e) FFdrag fluid Fmagnetophoretic F mag
f)
g)
Fmagnetophoretic
Fmag
Fig 7 | In-situ magnetophoretic particle capture. (a) Implementation of a fluidic cavity (surrounded by red oval) before the waste compartment for magnetophoretic enrichment of magnetic particles. (b) Magnetized particles (green) are retained by magnetophoresis whereas non-magnetic particles (black) are washed into the waste section. (c) Microscope image of 12 µm magnetic reference beads (highlighted by red circles) spiked into whole blood. (d) An additional washing step ensures elimination of non-magnetic background. (e) Microscope image of the magnetophoretically enriched 12 µm analytes in the cavity. (f) Finally, the magnetic beads are aspirated with a pipette for subsequent analysis with diminished background. (g) Microscope image of 12 µm particles enriched from whole blood. Scale bars are 400 µm.
3.3 In-situ magnetophoretic sorting of the magnetic particles for follow-up diagnostics. For potential subsequent analysis of the magnetically labelled targets we implemented a magnetophoretic cell sorter before the waste compartment on the cartridge (Fig. 7a). Therefore, we designed a cavity with a volume of 20 µl near the edge of the permanent magnet positioned underneath the device in order to collect all magnetic targets due to lowered fluidic drag and the increased magnetic field gradients at the edge of the permanent magnet. For a proof-of-concept of the magnetophoretic sorter functionality, we spiked 12 µm magnetic reference beads into 1:10 diluted whole blood and pumped the sample at a flow rate of 1.5 µl s -1 through the microfluidic channel. The cavity reduces the fluidic drag force acting on the particle due to the higher cross section of the channel, which
19
causes the magnetophoretic force to exceed the fluidic drag and captures the magnetic targets (Fig 7b,c). The remaining cellular background was washed twice from the cavity into the waste compartment with PBS buffer at a higher flow rate of 7 µl s-1 (Fig. 7d,e). The cartridge was then removed from the magnet and the magnetic particles can be aspirated through a port for a standard 100-µl pipette tip (Fig. 7f). By counting the number of erythrocytes per particle in our sample, we found the enrichment factor of our integrated sorter to be 235 allowing for follow-up analysis of the magnetic particles with diminished cellular background (Fig. 7g).
4. Discussion We have developed an adaptive and scalable particle focusing concept for magnetic flow cytometry. Contrary to high throughput flow cytometry, which for particle concentrations exceeding the detection limit of the instrument requires additional dilution steps and gating strategies to exclude coincidences (Keji et al., 1991; Wynn and Hounslow, 1997; Wersto et al., 2001; Kammel et al., 2012), our methodology enables singulated magnetic targets rolling with controllable particle-particle distance. The adaptive focusing covers a dynamic range of 2.7 orders of magnitude based on the fractioned patterning of the substrate surface of a microfluidic injection-molded device with chevron-like barriers. Our proposed all-magnetic workflow is based on balancing magnetophoretic and hydrodynamic forces to eliminate the need for additional sheath-flow focusing (Issadore et al., 2012; Lee et al., 2014; Muluneh and Issadore, 2014). While previous research on magnetoresistive sensor integration utilized complex multilayer PDMS interface between the sensor and fluidics (Wu et al., 2010; Muluneh and Issadore, 2014), we achieve effortless integration of a functionalized QFN package exposing the passivated GMR sensor in an injection-molded cartridge device. In this way, we minimize the Si footprint for a cost-efficient molded cartridge using standard semiconductor postprocessing techniques. Furthermore, an injection-molded device allows for precise and reproducible positioning of the mechanical chevrons with respect to the detection area and low-cost implementation of peripherals. The combination of the adaptive focusing and the highly precise positioning of the magnetic particles rolling over a sensor half-bridge allowed us to derive accurate quantitative volumetric information, even for high particle concentrations exceeding the spatial resolution of the sensor.
20
For subsequent analysis and means of future quality control, we even incorporated a magnetophoretic sorting mechanism before the waste compartment of the device. In contrast to reported devices for magnetophoretic sorting, which require additional features to create high localized magnetic field gradients (Xia et al., 2006; Adams et al., 2008; Issadore et al., 2010), we explored an effortless design, which utilizes the field gradient of the permanent magnet placed underneath the cartridge and the reduced drag force in a cavity within the microfluidic channel. In a proof-of-concept experiment with magnetic reference beads spiked into whole blood we demonstrated a significant reduction in the cellular background by a factor of 235.
Conclusion In brief, we present an integrated and scalable particle focusing method, which can be realized with standard molding methods and increases the dynamic concentration range of a sheath-less magnetic flow cytometer over various orders of magnitude. At the same time, particle enrichment and guiding effort by magnetophoresis on the Si substrate are reduced, which allows to minimize sensor footprint and thus costs. The combination of both adaptive mechanical and magnetophoretic analyte enrichment is not limited to magnetic sensing and could be of interest for other miniaturized cytometry and cell sorting methods with unknown cell concentration.
Acknowledgements The project was supported by the German Federal Ministry of Education and Research under the program ‘Werkstoff-innovationen für Industrie und Gesellschaft – WING’ (grant number 13N12013). We thank R. Gransee for fruitful discussions and assistance in cartridge design. We are grateful to M. Joksch for contribution in software development algorithms and T. Endner, J. Paquet, and A. Rosenberg for technical assistance in device fabrication and experiments. Discussions with S. Krause, J. Bosch, and B. Bock on flow cytometry experiments are greatly acknowledged.
Contributions M.R., L.R., M.J.H., S.A., B.A., I.v.D., M.N., M.B., B.K., O.F., and O.H. conceived the mechanophoretic enrichment methodology for an integrated point-of-care magnetic flow cytometer and designed experiments. O.H., B.K., and O.F. supervised the research and experimental results. M.R., L.R., M.J.H., S.A., and M.B. conceptualized the adaptive focusing methodology and sensor design. M.R. and L.R. conducted
21
experiments and analyzed the data. B.A. designed fluidic components and cartridge assembly. I.v.D. conducted sensor packaging and device integration. M.N. designed and assembled the read-out instrument. M.R. wrote the manuscript. All authors reviewed the experimental results and commented on the manuscript.
Competing interest The authors declare no competing financial interests.
Supplementary Video 1: Enrichment of a pre-defined magnetic particle fraction while deflected particles are passing the sensor outside the detection area on the side walls of the microfluidic channel. Supplementary Video 2: Magnetophoretic focusing of sequentially aligned magnetic particles via NiFe chevrons on the sensor chip.
References Adams, J.D., Kim, U., Soh, H.T., 2008. Proc. Nat. Acad. Sci. 47, 18165-18170. Bechstein, D.J.B., Ng, E., Lee, J.R., Cone, S.G., Gaster, R.S., Osterfeld, S.J., Hall, D.A., Weaver, J.A., Wilson R.J., Wang, S.X., 2015. Lab Chip 15, 4273-4276. Downey, G.P., Doherty, D.E., Schwab, B. 3rd, Elson, E.L., Henson, P.M., Worthen, G.S., 1990. J. Appl. Physiol. 69, 1767-1778. Einwallner, E., Subasic, A., Strasser, A., Augustin, D., Thalhammer, R., Steiner, I., Schwarzinger, I., 2013. J. Immunol. Methods 390, 127-132. Fernandes, A.C., Duarte, C.M., Cardoso, F.A., Bexiga, R., Cardoso, S., Freitas, P.P., 2014. Sensors 14, 15496-15524. Gaster, R., Xu, L., Wilson, R., Hall, D., Osterfeld, S., Yu, H., Wang, S., 2011. Nat. Nanotechnol. 6, 314-320. Greve, B., Beller, C., Cassens, U., Sibrowksi, W., Göhde, W., 2006. Stem cells 24, 793-799. Hayden, O., Rührig, M., 2012. European Patent, 2404155. Hayden, O., Helou, M.J., Reisbeck, M., Tedde, S.F., 2016. European Patent 2641087. Helou, M.J., Reisbeck, M., Tedde, S.F., Richter, L., Bär, L., Bosch, J.J., Stauber, R.H., Quandt, E., Hayden, O., 2013. Lab Chip 13, 1035-1038. Huang, C.C., Zhou, X., Hall, D.A., 2017. Sci. Rep. 7, 45493. Inglis, D.W., Riehn, R., Sturm, J.C., Austin, R.H., 2006. J. Appl. Phys. 99, 08K101. Issadore, D., Shao, H., Chung, J., Newton, A., Pittet, M., Weissleder, R., Lee, H., 2010. Lab Chip 11, 147-151. Issadore, D., Chung, J., Shao, H., Liong, M., Ghazani, A.A., Castro, C.M., Weissleder, R., Lee, H., 2012. Sci. Transl. Med. 4, 141ra92. Issadore, D., Park, Y.I., Shao, H., Min, C., Lee, K., Liong, M., Weissleder, R., Lee, H., 2014. Lab Chip 14, 2385-2397. Jackson, J.D., Classical electrodynamics, 1999, 3rd edition, Wiley. Kammel, M., Kummrow, A., Neukammer, J., 2012. J. Lab. Med., 36, 25-35 Keij, J.F., Van Rotterdam, A., Groenewegen, A.C., Stokdijk, W., Visser, J.W.M., 1991. Cytometry 12, 398-404. Kim, K.W., Reddy, V., Torati, S.R., Hu, X.H., Sandhu, A., Kim, C.G., 2015. Lab Chip 15, 696-703. Lee, C.P., Lai, M.F., Huang, H.T., Lin, C.W., Wei, Z.H., 2014. Biosens. Bioelectron. 57, 48-53. Lee, H., Shin, T.H., Cheon, J., Weissleder, R., 2015. Chem. Rev 115, 10690-10724. Lin, G., Makarov, D., Melzer, M., Si, W. Yan, C., Schmidt, O.G., 2014. Lab Chip 14, 4050-4058. Lin, G., Makarov, D., Schmidt, O.G., 2017. Lab Chip 17, 1884-1912.
22
Loureiro, J., Andrade, P.Z., Cardoso, S., da Silva, C.L., Cabral, J.M., Freitas, P.P., 2011. Lab Chip 11, 2255-2261. Melzer, M., Karnaushenko, D., Makarov, D., Baraban, L., Calvimontes, A., Mönch, I., Kaltofen, R., Mei, Y., Schmidt, O.G., 2012. RSC Adv. 6, 2284-2288. Muluneh, M., Issadore, D., 2014. Lab Chip 14, 4552-4558. Osterfeld, S., Yu, H., Gaster, R., Caramuta, S., Xu, L., Han, S., Hall, D., Wilson, R., Sun, S., White, R., Pourmand, N., Wang, S., 2008. Proc. Natl. Acad. Sci. 105, 20637-20640. Reisbeck, M., Helou, M.J., Richter, L., Kappes, B., Friedrich, O., Hayden, O., 2016. Sci. Rep. 6, 32838. Robinson, J.P., Roederer, M., 2015. Science 350, 739-740. Schotter, J., Shoshi, A., Brueckl, H., 2009. J. of Mag. And Mag. Mat. 321, 1671-1675. Shapiro, H.M., Practical Flow Cytometry, 2003, Wiley. Vallotton, P., Olivier, S., 2013. Microsc. Microanal 19, 451-460. Wersto, R., Chrest, F., Leary, J., Morris, C., Stetler-Stevenson, M., Gabrielson, E., 2001. Cytometry 46, 296-306. Wu, A., Wang, L., Jensen, E., Mathies, R., Boser, B., 2010. Lab Chip 10, 519-521. Wynn, E., Hounslow, M., 1997. Powder Technol. 93, 163-175. Xia, N., Hunt, T.P., Mayers, B.T., Alsberg, E., Whitesides, G.M., Westervelt, R.M., Ingber, D.E., 2006. Biomed. Microdev. 8, 299308.
Highlights
3D adaptive particle enrichment extends the dynamic range in magnetic flow cytometry Mechanical and magnetophoretic focusing minimizes coincidences in undiluted samples Functionalized semiconductor package enables sheath-less sensor integration Injection-molded cartridge with an integrated workflow for Point-of-Care testing
23
