Human Subject Characterization and CSF Sample Collection
Postmortem CSF was collected at autopsy from clinically and neuropathologically characterized participants enrolled in the Arizona Study of Aging and Neurodegenerative Disorders and Brain and Body Donation Program41. Briefly, most subjects were volunteers recruited from the surrounding communities of Maricopa County, Arizona, especially the Sun Cities. The demographics of the population consisted largely of Caucasian, middle to high income individuals. For each participant, a subspecialist cognitive-behavioral neurologist performed a comprehensive evaluation. A cognitive diagnosis was assigned at a consensus conference attended by neuropsychologists, neurologists, and cognitive neurology subspecialists. All neuropathological examinations were performed by the same neuropathologist, blinded to clinical findings.
Prior to removing the brain, cerebrospinal fluid (CSF) was drawn from the lateral ventricles, using 30 ml disposable syringes fitted with 8 cm long, 18-gauge needles. The CSF was ejected into 15 ml disposable polyethylene tubes for centrifugation. CSF was centrifuged at 5k rpm for 10 minutes and supernatants from the CSF were aliquoted into 0.5 ml polyethylene microcentrifuge tubes and stored frozen at − 80°C.
Amyloid plaque and neurofibrillary tangle density in brain were graded and staged at standard sites in frontal, temporal, parietal and occipital cortex as well as hippocampus and entorhinal cortex, based on the aggregate impression from the 80 µm sections stained with thioflavine S, Campbell-Switzer and Gallyas methods. The total plaque score, considering all types of plaques (cored, neuritic and diffuse) together, was predominantly derived from the Campbell-Switzer stain while the Gallyas and thioflavine S stains were used for estimating neuritic plaque densities. All three stains show neurofibrillary changes and therefore this score was estimated after viewing slides stained with all three. Both total and neuritic plaque densities were rated as none, sparse, moderate and frequent, using the published CERAD templates42. Conversion of the descriptive terms to numerical values provides scores of 0–3 for each area, with a maximum score of 15 for all five areas combined. Neurofibrillary tangle abundance and distribution was also graded in these thick sections, again using the CERAD templates for this, while the original Braak protocol43 was used for estimating the topographical distribution of neurofibrillary change.
Mini-Mental State Examination.
The Mini-Mental State Examination (MMSE) is a brief, widely-used screening test for measuring cognitive impairment and dementia. The test score ranges from 0–30, with lower scores indicating poorer performance and cutoff scores for cognitive impairment typically range between 24–26 with varying sensitivity and specificity.44
FLOWER microtoroid biosensor.
The FLOWER biosensing system is illustrated in Fig. 1a, including the major components required for measuring the resonant wavelength shift by frequency-locking a tunable laser (TLB-6712, Newport) to the microtoroid resonator. A narrow linewidth laser with a tuning range from 765 nm to 781 nm was chosen, where the absorption by water is minimal compared to higher infrared wavelengths. Depending on the microtoroid geometry, each microtoroid will support an optical resonance at the resonance condition:
$$\begin{array}{c}2\pi R{n}_{eff}=m\lambda \left(1\right)\end{array}$$
where \(R\) is the major radius of the microtoroid resonator, \({n}_{eff}\) is the effective refractive index of the guided mode, \(m\) is an integer, and \(\lambda\) is the free-space wavelength of the laser. Figure 1b shows a scanning electron micrograph of a row of microtoroids on a single chip. The chip is placed in a custom-built fluidic chamber mounted on a 3-axis micrometer and nano positioning piezo stage, which allows for precise coupling between the tapered fiber waveguide and the microtoroid. A 2D axially symmetric COMSOL simulation shows how the optical mode is distributed in a cross section of the microtoroid resonator, along with the evanescent field extending past the surface of the microtoroid and into the surrounding environment (Fig. 1c). Optical resonances appear as sharp dips in the transmission spectrum, and the resonant wavelength shift is measured while the detection antibody is injected into the fluidic chamber (Fig. 1d).
Microtoroid Fabrication
Silicon wafers with a 2 µm thick thermal oxide layer were purchased from UniversityWafer, Inc. In a cleanroom, a maskless, direct-write photolithography tool (Heidelberg Instruments MLA150) was used to create columns of 150 µm photoresist circles on the wafer. The photoresist pattern acts as a mask during the subsequent buffered oxide wet-etch process, which etches the exposed silica and leaves behind 150 µm silica circles on a silicon substrate. Afterwards, the photoresist mask is washed away with acetone and IPA, and the wafer is cut into smaller “chips” before drying in an oven at 130°C for at least 30 minutes. Next, the chips are dry etched using XeF2 (Xactix e2, SPTS) which isotropically etches the exposed silicon substrate to form silica microdisks atop a silicon pillar. A CO2 (λ0 = 10.5 µm) laser is used to reflow the silica microdisks to form the final microtoroid resonator.
Microtoroid Surface Functionalization Using APTES.
Prior to the biosensing experiment, microtoroid chips were incubated in a solution of 1% v/v APTES in chloroform for 15 minutes. Next, the chips were washed with denatured ethanol and dried under a nitrogen stream. Afterwards, the chips were incubated overnight in a solution of 0.1 M succinic anhydride in dimethyl formamide. The next day, the chips were washed with denatured ethanol and dried under a nitrogen stream. Next, the chips were incubated in a solution of 100 mM EDC, 100 mM Sulfo-NHS prepared in MES buffer (pH 6) for 15 minutes at room temperature. Afterwards, the chips were washed with 100 mM PBS before incubation in Aβ42 specific antibody.
Amyloid-beta 1–42 Standard Curve and CSF FLOWER Assay.
To construct a standard curve for Aβ42, the APTES functionalized microtoroids were conjugated with 10 µg/mL anti- Aβ42 (12F4 clone) capture antibody in PBS, which binds to epitopes 36–42 (C-terminal) of the Aβ42 peptide45. Dried HFIP Aβ42 (AS-64129-05, Anaspec) peptide film was dissolved in 10 mM NaOH to a stock concentration of 1 mg/mL, and then serially diluted in 100 mM PBS. Next, the chips were incubated in the HFIP treated Aβ42 peptide for 2 hours at room temperature and then washed with HEPES sample buffer (HEPES 25 mM, NaCl 125 mM, BSA 0.1% w/v, EDTA 1 mM, pH 7.5). The chips were kept in HEPES sample buffer on ice until data measurement. The resonant wavelength shift was measured while the anti- Aβ42 detection antibody46 (6E10 clone) (1 µg/mL, diluted in HEPES sample buffer) was perfused into the fluidic chamber containing the microtoroid chip. To measure Aβ42 in CSF, the microtoroid chips were functionalized with 10 µg/mL anti- Aβ42 capture antibody (12F4) in PBS. CSF samples were thawed on ice from − 80 degrees C and then centrifuged for 10 minutes at 3000 x g. The chips were incubated in the CSF samples for 2 hours at room temperature before washing with sample buffer. Samples were kept on ice in HEPES sample buffer until the resonant wavelength shift was measured using 1 µg/mL anti- Aβ42 detection antibody (6E10).
CSF Ultrasensitive Human Aβ42 ELISA.
To compare the diagnostic performance of FLOWER vs. ELISA, we measured CSF Aβ42 using an Invitrogen ultrasensitive human Aβ42 ELISA kit. CSF samples were diluted 1:4 using the included standard diluent buffer before being added to the wells on a 96-well plate and incubated for 3 hours with detector antibody at room temperature. After, the wells were washed 4 times with 1X wash buffer and the HRP conjugated secondary antibody was added and incubated for 30 minutes at room temperature. Next, the wells were washed and Tetramethylbenzidine (TMB) solution was added and incubated for 30 minutes in the dark. Lastly, stop solution was added to each well before reading the plate. The absorbance at 450 nm was measured using a Biotek Synergy HT Microplate Reader.
FLOWER Experimental Setup and Data Acquisition.
A single-mode optical fiber (SM600, Thorlabs) was tapered using a custom built, motorized pulling stage and a hydrogen torch. After tapering, the fiber remains in the pulling stage and is moved over to the experimental setup and table (See Fig. S1). The functionalized microtoroid chip is affixed with double sided tape in a custom 3D-printed fluidic chamber (internal volume ~ 120 µl) attached to a rod. The rod is mounted onto a 3-axis micrometer and nanopositioning stage (P-611.3 Nanocube, PI) to allow for precise coupling between the microtoroid and the tapered fiber. A glass coverslip is cut to size and placed on top of the fluidic chamber to contain the fluid, chip, and tapered fiber. A 100 µm diameter perfusion pencil tip (AutoMate Scientific) is inserted into the fluidic chamber to allow for delivery of the samples to the toroid via an 8-channel pressurized perfusion system (AutoMate Scientific) and electric rotary valve system (ASP-ERV-O1.2-08, Aurora Pro Scientific). Each chip contains ~ 6–8 toroids, which are checked for high-Q resonances (Q > 105) by evanescently coupling the tapered fiber and measuring the transmission spectrum over the laser’s tuning range.
The shift in the microtoroid resonance frequency is measured using the top-of-fringe locking function on a Toptica Digilock 110. The photodetector’s (Nirvana 2007, Newport) signal output was connected to the Digilock input, and the Digilock analog voltage output was connected to the tunable laser’s frequency modulation input and an analog voltage data acquisition card (DAQ) (PCI-4461, National Instruments). The Digilock modulates the laser’s frequency with a 2 kHz sine wave to generate an error signal. Any shift in the microtoroid resonance frequency is compensated for by the Digilock sending a voltage signal to both the laser and the DAQ card (PCI-4461, National Instruments).
Wavelength Shift Curve Fitting.
The resonant wavelength shift data was fitted using a one-phase association function describing the binding kinetics between a receptor and its ligand47
$$\begin{array}{c}y={Y}_{max}\left(1-{exp}^{-kx}\right)\left(2\right)\end{array}$$
where \({Y}_{max}\) is the projected maximum shift at infinite time with units of femtometer and \(k\) is a rate constant with units of s− 1.
The maximum wavelength shifts obtained from Eq. 2 were used to construct calibration curves for Aβ42 using a 5-parameter logistic fit48:
$$\begin{array}{c}y={A}_{1}+\frac{{A}_{2}-{A}_{1}}{{\left(1+{10}^{\left(Log{x}_{0}-x\right)p}\right)}^{s}} \left(3\right)\end{array}$$
where \({A}_{1}\) is the bottom asymptote, \({A}_{2}\) is the top asymptote, \({Logx}_{0}\) is the center, \(p\) is the hill slope, and \(s\) is the symmetry factor.