A Lab‐On‐chip Tool for Rapid, Quantitative, and Stage‐selective Diagnosis of Malaria

Abstract Malaria remains the most important mosquito‐borne infectious disease worldwide, with 229 million new cases and 409.000 deaths in 2019. The infection is caused by a protozoan parasite which attacks red blood cells by feeding on hemoglobin and transforming it into hemozoin. Despite the WHO recommendation of prompt malaria diagnosis, the quality of microscopy‐based diagnosis is frequently inadequate while rapid diagnostic tests based on antigens are not quantitative and still affected by non‐negligible false negative/positive results. PCR‐based methods are highly performant but still not widely used in endemic areas. Here, a diagnostic tool (TMek), based on the paramagnetic properties of hemozoin nanocrystals in infected red blood cells (i‐RBCs), is reported on. Exploiting the competition between gravity and magnetic forces, i‐RBCs in a whole blood specimen are sorted and electrically detected in a microchip. The amplitude and time evolution of the electrical signal allow for the quantification of i‐RBCs (in the range 10–105 i‐RBC µL−1) and the distinction of the infection stage. A preliminary validation study on 75 patients with clinical suspect of malaria shows on‐field operability, without false negative and a few false positive results. These findings indicate the potential of TMek as a quantitative, stage‐selective, rapid test for malaria.


RBC,
, is equal to 1.1·10 -13 N. According to the expression of the magnetic force on a superparamagnetic particle, , assuming a difference in susceptibility = 1.8·10 -6 between RBC and plasma, 1 the value of the H 2 gradient needed to balance is of the order of 1·10 15 A 2 ·m -3 .
A similar estimate can be made for a single crystal of hemozoin suspended in plasma. Assuming an average volume V HC = 2.2·10 -14 cm -3 , a density = 1.15 g·cm -3 , and a difference in susceptibility = 4.1·10 -4 compared to plasma, the value of the H 2 gradient needed to balance for hemozoin crystals is of the order of 1.7·10 13 A 2 ·m -3 .

Characterization of the macroscopic field produced by the permanent magnet assembly
The characterization of the field and gradient produced by the two NdFeB magnets sandwiching a 0.2 mm thick -metal foil has been carried out by using a Hall-probe mounted on a linear positioning stage. We measured the field component perpendicular to the surface of the assembly at variable distance from the surface of the magnets, while keeping the probe center in the symmetry plane defined by the -metal foil. As a matter of fact, the stray field emanating from the -metal is essentially parallel to the foil plane (see Fig. 2b in the main text), i.e. perpendicular to the chip surface, at least within ±1 mm from the symmetry plane, corresponding to the area with measurement electrodes on the chip, centered on the foil plane. At 0.5 mm from the magnet assembly surface, roughly corresponding to the position of the concentrators taking into account that during measurements the back of the chip (500 m thick) is placed in close proximity to the magnets, the H field produced is about 6·10 5 A·m -1 while the corresponding H 2 is 7·10 14 A 2 ·m -3 .
These values are definitely enough to saturate the Ni concentrators (see loops below) and produce the macroscopic attracting force towards the chip acting on i-RBC, responsible for the deviation from the vertical of their trajectories (see Fig. 4e in the main text).
The effect of the -metal foil has been evaluated by measuring the field and gradients produced with and without the foil, while keeping the same distance of 0.2 mm between the faces of the magnets. It turns out that the field is very similar (within 5%), while the mechanical stability of the assembly largely improves with the foil because we can better press the magnets using the foil as spacer and avoid the sliding motion due to the magnetic repulsive force. Figure S1 reports the characteristic M(H) curve of the out-of-plane component of the magnetization vs. the out-of-plane magnetic field, measured on a portion of the microchip with 1000 fabricated Ni concentrators arranged in a hexagonal lattice, using a Vibrating Sample magnetometer. The out-ofplane direction has been chosen because the field created by the permanent magnet assembly is applied mainly perpendicularly to the chip surface. The experimental curve shows that, in order to saturate the magnetization (M), an external field of about 300000 A·m -1 is required. As discussed above, the field provided by the permanent magnet assembly is 600000 A·m -1 , thus ensuring that all Ni posts are saturated upon the magnet approach during the assay. Furthermore, the M(H) curve displays a negligible hysteresis and remanent magnetization. This is crucial for TMek operation, because the measurement protocol (see Fig. 2f of the main text) exploits the i-RBC release upon permanent magnet disengagement, which is based on the absence of sizable ferromagnetic remanence. The saturation magnetization is about 4.5·10 5

M(H) behavior of fabricated Ni concentrators
A·m -1 , in nice agreement with the theoretical value of 5·10 5 A·m -1 , taking into account some uncertainty in the determination of the Ni post height due to some non-uniformity in the microfabrication process.

Concentrators layout
Ni concentrators are arranged in a hexagonal lattice, as depicted in Figure S2. The lattice parameter (S) is 160 m, while the diameter and height of each cylinder are 40 m and 20 m, respectively.

Note 2: Chip fabrication
The process flow is described in more details in Figure S3.
Optimized electroplating conditions corresponds to a working temperature of 50 degree and a pH level of 4.0 as recommended by the ATOTECH. The current density ensuring good hole filling is Mechanical polishing is carried out between steps 5 and 6, in order to achieve a roughness lower than 50 nm rms.
A uniform gold coating is finally deposited on the back on the chip to create a reference ground plane and reduce the background.

Note 3: Impedimetric detection and multiphysical simulations
According to the Maxwell mixture theory 2 the conductivity of the medium is modified by the presence of insulating particles according to the following expression: (1) where  represents the volume fraction of the particles in the medium. For small values of the volumetric fraction the equation above can be rewritten in terms of resistivity: (2) The total resistance percentage variation due to the capture of a number N p of particles, each of them with volume V p , within the electrode probing volume, is thus given by: where V e is the volume sensed by the electrodes, approximately given by area of the electrodes multiplied by their separation.
Each sensor, as described in the main text (Methods -Electronic platform for impedimetric detection and data acquisition), is composed by two areas: a measurement area, in which the magnetic concentrators are located, and a reference area. The inner elements of the electrodes from both areas are connected together (Fig. S4) to the virtual ground of a transimpedance amplifier (TIA) in order to implement an analog differential measurement enabling both the partial compensation of the common fluctuations and the removal of the offset. Figure S4: Working principle of the differential measurement. The measurement electrodes (top) and the reference electrodes (bottom) are stimulated with counter-phase signals. Without i-RBCs captured by the magnetic concentrators (a) the measured current is ideally zero. The capture of i-RBC (b) produces a net differential signal.
When two counter-phase sinusoidal signals with amplitude V a = 100 mV are applied to the outer elements of the electrodes in the measurement and reference areas, the TIA measures the resulting differential current proportional to the impedance mismatch between the measurement and reference electrodes, originated by the capture of i-RBCs.
The differential current variation is thus given by: where R 0 = 170  is the initial resistance of the electrodes and R the impedance mismatch due RBC capture. This is the expression which has been used to evaluate the impedance variation from the COMSOL simulation of the t-RBC capture and detaching, by monitoring over time the number of particles N p which contribute to the impedance variation according to equation (3).
In figure S5 we report the simulated current variations for the three values of parasitemia considered in the text: 0.01%, 0.1% and 1%, from which the signal amplitudes corresponding to black squares in Figure 4a of the main text have been obtained.
The discretized behavior of the signals is due to the limited portion of the chip used in the numerical simulations.

Note 4: Sensitivity to free hemozoin crystals
We used synthetic hemozoin crystals (-hematin) from Invivogen suspended in PBS at calibrated concentrations to estimate the test sensitivity to free hemozoin crystals, in analogy to what has been done for t-RBCs. In Figure S6 we show a typical differential signal taken using a protocol similar to that used for i-RBCs and described in the main text, after subtraction of a linear drift. Upon signal stabilization, at 60 s we approach the magnet and the current decreases due to HC accumulation.
When we re-engage the magnet, at 200 s, HC are released and the initial base-line is recovered. The HC concentration here was 15 g/ml, and the corresponding signal is about 300 nm. By decreasing the concentration, the lowest detectable limit is mainly determined by the tendency of hemozoin to stick on the concentrators and the difficulty to evaluate net signals over the signal drift. We found distinguishable signals, on the order of 50 nA, down to 3 g/ml, which corresponds to the amount of hemozoin contained in about 5000 i-RBC/l. To determine the percent of infected RBCs in patients from Sacco University Hospital we used the thin smear method.  Note also that a second population of results (empty red dots in Fig. S7) deviates from linearity, but these anomalous data correspond to samples where microscopy analysis revealed the presence of either a lot of free pigment or gametocytes. The presence of these corpuscles with higher magnetic susceptibility and capture efficiency with respect to i-RBCs in the asexual stage (see Table 1), gives rise to an anomalous increase of the TMek signal, in particular for gametocytes which have a larger volume with respect to HC. However, as shown in Figure 5f of the main text, the presence of a sizable amount of gametocytes can be detected looking at the signal waveform during capture, without microscopy investigation. Similar considerations apply also to free-pigment, 7 thus enabling in perspective a self-consistent use of TMek.

Note 7: Data analysis algorithm used for the prevalidation campaign
In order to ensure a robust data analysis compatible with automatic classification we developed an algorithm based on three steps: (1)  the error  w of the blank signal. For the latter we assume that the error on the blank amplitude is on the same order of the blank amplitude itself, due to some irreproducibility in the blank measurement which depends on spurious coupling induced by the magnet motion:  w = A 1w . The total error on the net signal A is thus  = ( 1 2 +  w 2 ) 1/2 .

Note 8: Data analysis for laboratory tests on t-RBCs
For tests performed on blood samples with treated RBCs mimicking i-RBCs, we used a data analysis protocol very similar to that presented above. The amplitude A 1 is evaluated by considering a linear background between 300 s and 600 s and measuring the signal at 420 s with respect to the background, as shown in Fig. 2f. The net signal A is then obtained by subtracting the amplitude A 1HB of the blank signal of the day, measured with the same chip and setup on a reference sample just made of the healthy donor whole blood diluted in PBS and heparin: A = (A 1 -A 1HB ). Due to high repeatability of blank measurements in good laboratory conditions, in this case the error () associated to A is just given by the square root of the sum of the squares of the errors evaluated on A 1 and A 1HB as explained above. The lowest detectable concentration is estimated by assuming that a meaningful signal must fulfil the condition A > .

Additional information
Movie M1 (separate file). A movie showing in detail the sample load and test operation is available online.
Dataset D1 (separate file). Dataset with anonymous patient IDs, clinical data, three independent evaluation of the parasitemia by independent microscopists, the malaria classification carried out by an infectious disease specialist on the basis of the above data, the results of a RDT (SD Bioline) and TMek carried out in parallel.