4.2.2. Plant Extraction and Sample Preparation

For each of the three groups (i.e., WT, M, and ECE), 10 biological replicates of rice plants were used for the extraction and subsequent NMR-based metabolomics analysis. Each rice plant was extracted individually using a method modified from previous reports [23,54]. Firstly, the rice sample-containing tube was snap-frozen in liquid nitrogen and the rice tissue was then swiftly ground into fine powder using a pre-cooled pestle. Pre-cooled methanol/water (v/v = 2/1, −20 ◦C) was added into the homogenized sample at a ratio of 600 μL per 100 mg powder. Afterwards, the mixture was further homogenized using a 2010 Geno/Grinder® (SPEX Sample Prep, Metuchen, NJ, USA) at 1300 rpm for 90 s. Then the homogenate mixture was sonicated in an ice bath with 10 cycles of 30 s sonication and 30 s break. Following centrifugation (14,489× *g*, 10 min, 4 ◦C), the supernatant was collected, and the remaining pellets were further treated twice using the same procedure. Three supernatants were combined and lyophilized after removal of methanol *in vacuo* (CentriVap Centrifugal Vacuum Concentrators, Labconco, MO, USA). Each dried extract was reconstituted into 600 μL phosphate buffer prepared as previously described. After a final centrifugation, 500 μL supernatant was transferred into a 5 mm NMR tube for NMR analysis.

#### 4.2.3. NMR Spectroscopy

All NMR spectra were acquired on a Bruker Avance 800 MHz NMR spectrometer (800.15 MHz for proton frequency) at 302 K using a 5 mm cryoprobe (Bruker Biospin, Rheinstetten, Germany) [55]. A one-dimensional (1D) 1H NMR spectrum was acquired for each sample in a random order using the first increment of the gradient selected NOESY pulse sequence (recycle delay−*G*1−90◦−*t*1−90◦−*t*m−*G*2−90◦−acquisition) with water presaturation during both the recycle delay (2 s) and mixing time (tm, 80 ms) [56]. For each spectrum, a total of 64 transients were collected into 32,768 data points over a spectral width of 16,025 Hz with a 90◦ pulse length adjusted to around 11 μs. For selected samples, a variety of two-dimensional (2D) NMR spectra [54] were acquired for the purpose of resonance assignment, including 1H−1H Correlation Spectroscopy (COSY), 1H−1H Total Correlation Spectroscopy (TOCSY), 1H J-Resolved Spectroscopy (JRES), 1H−13C Heteronuclear Single Quantum Correlation Spectroscopy (HSQC), and 1H−13C Heteronuclear Multiple Bond Correlation Spectroscopy (HMBC).

After Fourier transformation with 1 Hz exponential line broadening and 65,536 data points zero-filling, each 1D spectrum was manually corrected for phase and baseline distortions and referenced to TSP (δ 0.00) using Topspin 2.0 (Bruker Biospin, Germany). The spectra regions δ 0.500−9.610 were segmented into discrete bins of 0.003 ppm width using AMIX (V3.9.15, Bruker Biospin, Germany) [57]. Spectra regions with imperfect water suppression (δ 4.500−5.170) or residual methanol signal (δ 3.356−3.370) were discarded. The intersample chemical-shift variations for some metabolites were manually corrected to prevent the possible adverse effect on subsequent data analysis [58].
