Analytical Chemistry

This document is confidential and is proprietary to the American Chemical Society and its authors. Do not
copy or disclose without written permission. If you have received this item in error, notify the sender and
delete all copies.

NMR-based mitochondria metabolomic profiling: a new

approach to reveal cancer-associated alterations

Journal: Analytical Chemistry

Manuscript ID Draft

Manuscript Type: Article

Date Submitted by the

n/a
Author:

Complete List of Authors: Domingo Ortí, Inés; CIPF, Polymer therapeutics lab
Ferrer-Torres, Patricia; CIPF, Polymer therapeutics lab
Armiñan, Ana; Centro de Investigacion Principe Felipe, Polymer
Theraprutics Lab
Vicent, Maria; Centro de Investigacion Principe Felipe, Polymer
Therapeutics Lab
Pineda-Lucena, Antonio; University of Navarra,
Palomino-Schätzlein, Martina; ProtoQSAR,

ACS Paragon Plus Environment

Page 1 of 26 Analytical Chemistry

1
2
3 NMR-based mitochondria metabolomic profiling: a new approach to reveal cancer-associated
4
5 alterations
6
7 Inés Domingo-Ortí1,2,3, Patricia Ferrer Torres2, Ana Armiñán1, María J. Vicent1*, Antonio Pineda-
8
Lucena3,4*, Martina Palomino-Schätzlein2,5*
9
10 1Polymer Therapeutics Laboratory and CIBERONC, Centro de Investigación Príncipe Felipe,
11
12 Valencia 46012, Spain
13 2NMR
14 Facility, Centro de Investigación Príncipe Felipe, Valencia 46012, Spain
15 3Drug Discovery Unit, Instituto de Investigación Sanitaria La Fe, Valencia 46026, Spain
16
4Molecular Therapeutics Program, CIMA Universidad de Navarra, Pamplona 31008, Spain
17
18 5ProtoQSAR, CEEI, Parque Tecnológico Valencia, Paterna 46980, Spain
19
20
21
22 *Emails: mjvicent@cipf.es; apinedal@unav.es; mpalomino@protoqsar.com.
23
24 Abstract
25
26 Studying metabolism may assist in understanding the relationship between normal and dysfunctional
27 mitochondrial activity and various diseases such as neurodegenerative, cardiovascular, autoimmune,
28
29 psychiatric, and cancer. Nuclear magnetic resonance (NMR)-based metabolomics represents a
30
powerful method to characterize the chemical content of complex samples and has been successfully
31
32 applied to studying a range of conditions. However, an optimized methodology is lacking for
33
34 analyzing isolated organelles such as mitochondria.
35
36 In this study, we report the development of a protocol to metabolically profile mitochondria from
37 healthy, tumoral and metastatic tissues. Encouragingly, this approach provided quantitative
38
39 information on up to 45 metabolites in one comprehensive and robust analysis. Our results revealed
40
significant differences between whole-cell and mitochondrial metabolites, which supports a more
41
42 refined approach to metabolic analysis.
43
44 We applied our optimized methodology to investigate aggressive and metastatic breast cancer in
45
46 mouse tissues, discovering that lung mitochondria exhibited an altered metabolic fingerprint. Specific
47 amino acids, organic acids, and lipids showed significant increases in levels when compared to
48
49 mitochondria from healthy tissues.
50
51 Our optimized methodology could promote a better understanding of the molecular mechanisms
52
53
underlying breast cancer aggressiveness and mitochondrial-related diseases and support the
54 optimization of new advanced therapies.
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 2 of 26

1
2
3 Keywords
4
5 Nuclear Magnetic Resonance, Mitochondria, Metabolism, Breast Cancer, Metastasis
6
7 Mitochondria play critical roles in various cellular processes, including oxidative phosphorylation
8
9 (OXPHOS), fatty acid oxidation, the urea and Krebs cycles, gluconeogenesis, and ketogenesis1.
10
While ATP generation represents the primary function of mitochondria, they also provide building
11
12 blocks for cell replication, control redox balance, participate in apoptosis, and function as a signaling
13
14 platform1. Normal cellular physiological function requires critical control of mitochondrial-
15 associated activities; furthermore, mitochondrial dysfunction can prompt the onset of diverse
16
17 pathological processes, including neurodegenerative, cardiovascular, autoimmune, and psychiatric
18
diseases, and notably, cancer2.
19
20
Cancer cells obtain energy through anaerobic glycolysis due to low oxygen levels, which impairs
21
22 mitochondrial respiration and increases cell acidity due to increased lactate production (the "Warburg
23
24 effect")3. A higher mitochondrial membrane potential, reflecting the altered functional status of
25 mitochondria, also characterizes tumorigenic development4. Mitochondrial dysfunction induced by
26
27 genetic alterations also triggers alterations to tumor cell energy production mechanisms (from
28 OXPHOS to glycolysis), thus contributing to cancer progression5. In this context, mitochondrial
29
30 metabolism supports cancer cell adaptation, which promotes metastasis4. Furthermore, altered
31
32
mitochondrial content supports additional facets of cancer aggressiveness6.
33
34 A deeper understanding of disease-associated molecular mechanisms may help to develop better
35 therapeutic approaches for aggressive metastatic cancers. In this context, developing specific methods
36
37 to study mitochondrial metabolism may detect disease-associated alterations to inform on these
38 mechanisms. Metabolomic profiling provides a comprehensive quantitative analysis of metabolic
39
40 pathway end products in biological systems7, providing valuable information on the biochemical
41
42
processes at play. Metabolomic analysis generally use biofluids8–10 but also cell and tissue samples11–
43 13. Significantly, data from these latter matrices reflect changes occurring at the origin of the disorder
44
45 and provide relevant information regarding associated mechanisms14.
46
47 Whole tissue metabolomics has been widely applied in biomedical research for the diagnosis of
48 pathological processes15, the discovery of early-disease biomarkers16, optimization of therapies for
49
50 various pathologies17,18, the study of different disease rates between populations19, and identification
51
52
of tumor characteristics and progression markers20. However, we lack protocols that support the
53 metabolomic profiling of organelles such as mitochondria. A more in-depth understanding of
54
55 mitochondrial metabolomics could capture dynamic changes missed by "bulk" whole-cell profiling21.
56 A more targeted approach could also represent a valuable tool in mitochondria-related disease
57
58
59
60 ACS Paragon Plus Environment
Page 3 of 26 Analytical Chemistry

1
2
3 research, helping to evaluate novel drugs (particularly those targeting mitochondria) and supporting
4
5 the early identification of biomarkers in diseases with a known mitochondrial impact (e.g.,
6 Alzheimer's disease, Parkinson's disease, and diabetes2).
7
8
While current mass spectrometry (MS)-based methods22 for quantifying the metabolite content of
9
10 isolated mitochondria are very sensitive and suitable for the study of targeted metabolic changes23,
11
12 the relatively complex sample preparation and analysis currently employed typically involve several
13 steps, which involves a relatively high degree of difficulty22. As an alternative/complementary
14
15 strategy for mitochondrial metabolomics, we sought to develop a more straightforward, robust, non-
16 targeted approach that generates a general mitochondrial metabolic fingerprint. We now describe the
17
18 development of an optimized protocol for the metabolomic analysis of mitochondria isolated from
19
healthy, tumoral and metastatic tissues using nuclear magnetic resonance (NMR) spectroscopy.
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 4 of 26

1
2
3 EXPERIMENTAL SECTION
4
5 Materials and reagents
6
7 Bovine serum albumin (BSA) and Trizma hydrochloride (Tris-HCl) were purchased from Sigma-
8
9 Aldrich (St. Louis, USA). D-mannitol and sucrose were purchased from Merck Millipore (Burlington,
10
USA). Deuterium oxide (D2O) and 3-(trimethylsilyl) propionic acid d4 sodium salt (TSP) were
11
12 obtained from Eurisotop (Tewksbury, USA). Monobasic sodium phosphate (NaH2PO4) was acquired
13
14 from Acros Organics (Geel, Belgium). Milli-Q water was used for all experiments (Millipore;
15 Burlington, USA).
16
17 A 15 mL Dounce tissue homogenizer and Ultra-turrax dispersing machine were purchased from
18
19 Wheaton (Millville, USA) and IKA (Staufen, Germany), respectively. 3K centrifugal filters were
20
acquired from Merck Millipore (Burlington, USA). 5 mm NMR tubes and 5 mm NMR tube caps for
21
22 SampleJet were obtained from Deutero (Kastellaun, Germany) and Bruker (Billerica, USA),
23
24 respectively.
25
26 Biological Material
27
28 Immunodeficient NOD/SCID mice (NOD.CB17-Prkdcscid/NCrHsd) were obtained from Envigo
29 Laboratories Inc. (Spain), and sixteen female mice were used for this study. The animal experiments
30
31 were performed following the European Communities Council Directive (86/609/ECC) guidelines
32
33
and the Spanish Royal Decree 1201/2005, and the procedures were approved by the Institutional
34 Animal Care and Use Committee. The mice were maintained in a specific pathogen-free facility under
35
36 environmental control, and food pellets and water were provided ad-libitum during the experiment.
37 The mice were evaluated twice a week to ensure their well-being.
38
39 Eight mice were induced with MDA-MB-231-Luc tumors by subdermal inoculation of 3x106 MDA-
40
41 MB-231 triple-negative breast cancer cells transduced with a luciferase expression plasmid suspended
42
43
in 100 μl of Matrigel (20%) into the second left mammary fat pad following inhalatory anesthesia
44 (3% sevoflurane in 100% oxygen). Tumor development did not cause any significant animal weight
45
46 loss (Supporting Information 1) or pain-related behavior. The mice were euthanized under a CO2
47 atmosphere when tumors reached 1.0 cm3 (around 42 days after inoculation). The livers, lungs,
48
49 kidneys, brains, and breast tumors were extracted, washed with 1X phosphate-buffered saline (PBS),
50
and snap-frozen in liquid nitrogen for subsequent metabolic analysis.
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
Page 5 of 26 Analytical Chemistry

1
2
3 Isolating mitochondria from tissues
4
5 The isolation procedure was performed at 4ºC to preserve mitochondria and avoid significant
6
7 metabolic alterations (see Figure 1A). Organs were removed from storage at -80ºC and kept on ice.
8
After weighing each sample, the tissues were homogenized in Buffer 1 (220 mM mannitol, 70 mM
9
10 sucrose, 10 mM Tris-HCl solution, and 0.5 mg/mL BSA in MilliQ water, pH 7.4) using a Dounce or
11
12 Ultra-turrax homogenizer. The homogenization of soft tissues was performed using two volumes (1:2,
13 tissue:buffer) of Buffer 1 and seventy strokes with a Dounce homogenizer. The homogenate was
14
15 diluted to obtain a 10% (v/v) homogenate solution. Due to its fibrous nature, the breast tumor tissue
16 was homogenized in twenty volumes (1:20, tissue:buffer) of Buffer 1 with the Ultra-turrax
17
18 homogenizer used at 12,000 rpm for precisely 10 seconds. Table 1 describes various parameters,
19
including tissue type, amount of tissue used, the homogenizer used, and the number of scans and
20
21 duration of the NMR experiment.
22
23 Table 1. Parameters for NMR analysis of mitochondrial samples isolated from tissues
24
25 Tissue Homogenizer Organ units Scans 1H-NMR Duration (minutes)
26
Breast tumor Ultra-Turrax 1/2 1280 125
27
28 Lung Dounce 2 1280 125
29 Kidney Dounce 2 640 61
30
31 Liver Dounce 1 640 61
32 Brain Dounce 1 640 61
33
34
35 Each homogenate was centrifuged at 560g for 15 minutes at 4°C. Next, the supernatant containing
36
37 mitochondria, cytosol, and membranes was transferred to a new tube while the nuclear pellet was
38
stored at -80°C for further experiments. Critical step: As the nuclear pellet can easily become
39
40 resuspended in the supernatant, sudden movements should be avoided, and the supernatant should be
41
42 collected carefully. The supernatant was then centrifuged at 7,000g for 15 minutes at 4°C. An aliquot
43 (1 mL) of the supernatant (cytosol and membranes) was stored at -80°C for further experiments. The
44
45 mitochondrial pellet was then washed twice with Buffer 1 (volume approximately equal to the weight
46 of the initial tissue) and centrifuged at 7,000g for 15 minutes at 4°C. Mitochondrial pellets were then
47
48 weighed and stored at -80°C.
49
The isolation of mitochondria was confirmed by Western blotting. Mitochondrial pellets were
50
51 resuspended in RIPA buffer (25 µL/10 mg; 150 mM NaCl, 50 mM Tris Base, 1% SDS, 0.5% sodium
52
53 deoxycholate, and 1% NP40, pH 8) containing a protease inhibitor cocktail (25x). Cytosolic samples
54 were prepared by adding 10 µL of a protease inhibitor cocktail (25x) to 250 µL of the cytosolic
55
56 fraction. After vortexing, the samples were centrifuged at 13,200 rpm for 15 minutes at 4°C, and the
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 6 of 26

1
2
3 supernatant was kept for protein quantification via the Bradford Assay. Twenty micrograms of protein
4
5 extract were used to assess the expression of mitochondrial and cytosolic markers. Table 2 reports
6 the primary and secondary antibodies used.
7
8 Table 2. Primary and secondary antibodies used for the analysis of mitochondrial and cytosolic
9
marker proteins
10
11 Antibody Supplier Dilution S Antibody Supplier Dilution
P
12 R α-tubulin Sigma-Aldrich, T8203 1:10000
E
Anti-mouse Sigma-Aldrich, A9044 1:5000
13 C
I
β-actin Santa Cruz Biotechnology, sc-8432 1:10000 O Anti-mouse Sigma-Aldrich, A9044 1:30000
14 M
N
15 A Cytochrome-C Cell Signalling, 4280 1:1000 Anti-rabbit Sigma-Aldrich, A6154 1:3000
D
R
16 Y
GAPDH Santa Cruz Biotechnology, sc-32233 1:6000 A Anti-mouse Sigma-Aldrich, A9044 1:20000
17 R
VDAC Abcam, ab15895 1:1000 Anti-rabbit Sigma-Aldrich, A6154 1:5000
Y
18
19
20 Metabolite extraction and NMR sample preparation
21
22 The extraction of metabolites from mitochondria employed the following protocol (Figure 1B).
23 Frozen mitochondrial pellets were placed on ice and allowed to thaw for 5 minutes. Then, 700 µL of
24
25 Buffer 2 (20 mM Na2HPO4 and 20 µM 3-(trimethylsilyl)propionic-2,2,3,3-d4 acid sodium salt
26
solution [TSP; used as an internal standard for NMR analysis] in D2O, pH 7.4) was added, and the
27
28 suspension was homogenized with a vortex, resuspended with a pipette, and subjected to ultrasound
29
30 by submerging the tubes in an ultrasonic bath for 5 minutes. Samples were placed in liquid nitrogen
31 for 1 minute to break cell membranes and then allowed to thaw in a temperate water bath. This step
32
33 was repeated two more times (three cycles in total). The suspension was centrifuged at 10,000 g for
34 15 minutes at 4°C, and the supernatant was transferred to 1.5 mL tubes.
35
36
Before NMR analysis, samples were passed through a 3 KDa centrifugal filter. The filter was washed
37
38 three times with 400 µL of Milli-Q water and once with 400 µL of D2O by centrifuging at 14,000 g
39
40 for 15 minutes at room temperature. Samples were then filtered by centrifugation at 14,000 g for 40
41 minutes at 4°C. Critical step: If the filter becomes clogged, transfer the sample to a fresh, washed
42
43 filter. Finally, 550 µL of the samples were transferred to a 5 mm NMR tube.
44
45 Metabolite extraction using whole cells was performed using the same method. 700 µL of Buffer 2
46
was added to the tissue to perform the homogenization, and the metabolite extraction process was
47
48 followed as before.
49
50 NMR analysis
51
52 Samples were introduced into the NMR spectrometer (Bruker AVII-600MHz equipped with a 5 mm
53
54 TCI cryoprobe; see Figure 1C), and the temperature was set to 27ºC. A reference standard sample
55 (Bruker) containing 2 mM sucrose, 0.5 mM sodium trimethylsilylpropanesulfonate (DSS), and 2 mM
56
57
58
59
60 ACS Paragon Plus Environment
Page 7 of 26 Analytical Chemistry

1
2
3 NaN3 in H2O/D2O was first analyzed to ensure adequate spectral resolution and water suppression.
4
1H 1D NOESY (Nuclear Overhauser Effect Spectroscopy) NMR spectra were acquired with
5
6 corresponding scans for each tissue type (see Table 1 and Supporting Information 2). The following
7
8 settings were applied to obtain good quality spectra: 64,000 data points were digitalized over a
9
spectral width of 30 ppm for optimal baseline correction, a 4-s relaxation delay was used between
10
11 FID (free induction decay) readings, and a water presaturation pulse of 25 Hz was applied to minimize
12
13 the water signal. The FID values were multiplied by an exponential function with a 0.5 Hz line
14 broadening factor.
15
16 Spectra processing and data analysis
17
18
The acquired NMR spectra were processed using Mestrenova 14.0.0 before data integration. The
19
20 spectra were referenced to the internal standard (TSP), and the phase was corrected manually to obtain
21
22 the spectra signals in the pure absorption mode. The baseline was corrected using a manual
23 optimization of the Whittaker smoother correction.
24
25 An assignment table for each tissue type was prepared based on information available regarding the
26
27 metabolomic profile of the total tissue (liver24, brain25, lung26, kidney24, breast tissue24), as well as the
28 Human Metabolome and Chenomx databases (see Supporting Information 3). The integration
29
30 templates were copied into a text file (example in Supporting Information 4), and the signals were
31
32
integrated using the "Predefined sum" calculation method in MestreNova software by uploading the
33 text file. Signal integrals were saved as 1D Integral Series (.txt) and opened in Excel.
34
35 For metabolomic analysis, the integral values were normalized to their total area by dividing each
36
37 integral value by the sum of all integrals in the sample. This approach preserves the relative intensities
38 of each sample peak. Signals stemming from the isolation buffer were excluded from data
39
40 normalization. For the quantitative comparison of different procedures, a series of selected
41
42
metabolites were integrated, and the integrals were quantified in relation to the internal standard
43 (TSP) to obtain absolute quantification.
44
45 Statistical analyses followed standard methods for multivariate and univariate analysis of
46
47 metabolomic data27. Unsupervised PCA (principal component analysis) and supervised OPLS-DA
48 (orthogonal partial least squares discriminant analysis) models were obtained using the SIMCA-P 16
49
50 software. Univariate analysis and ANOVA (One-Way Analysis of Variance) were carried out using
51
52
Excel and the MetaboAnalyst 5.0 web-based platform28.
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 8 of 26

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32 Figure 1. Schematic representation of the protocol used for (A) mitochondria isolation from
33 tissues, (B) metabolite extraction and sample preparation, and (C) NMR analysis and biological
34
35 interpretation. B1: buffer 1 (220 mM mannitol, 70 mM sucrose, 10 mM Tris-HCl solution, and 0.5
36 mg/mL BSA in MilliQ H2O, pH 7.4); B2 (20 mM Na2HPO4 and 20 μM TSP solution in D2O, pH 7.4),
37
38 C: cytosol, M: mitochondria, and N: nucleus.
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
Page 9 of 26 Analytical Chemistry

1
2
3 RESULTS
4
5 Optimizing the NMR analysis of mitochondrial metabolites
6
7 Creating a mitochondrial metabolic profile involves three steps (Figure 1): (A) isolation of the
8
9 mitochondrial fraction from tissues, (B) extraction of metabolites from the mitochondrial fraction,
10
and (C) quantification of metabolites by NMR. These steps can be carried out sequentially in a single
11
12 day or over several days by freezing the samples in between steps.
13
14 The procedure used for mitochondrial isolation from tissues depends on the final aim of the study and
15
16 the grade of purity and activity of the isolated mitochondria required29,30. We selected tissue
17 homogenization followed by differential centrifugation as the most suitable method for mitochondria
18
19 isolation from tissues to support the robust and reproducible acquisition of metabolomic profiles30.
20
The simplicity and high recovery rate make this approach commonplace in organelle isolation
21
22 protocols, with extensive applications in tissues such as the liver30,31, heart31, musculoskeletal31,
23
24 brain31, kidney32, and testes32. Generally, two standard methods are available for tissue
25 homogenization – manual Dounce or automatic Ultra-turrax approaches. We compared metabolite
26
27 recovery using both homogenization methods for all tissues; only the tumor samples displayed a
28 higher intensity in some NMR signals (e.g., acetate, creatine, and choline derivatives) after
29
30 homogenization with Ultra-turrax compared to Dounce (Supporting Information 5), probably due
31
32
to its more compact nature (given the fibrous, muscular type of tissue). Therefore, we employed Ultra-
33 turrax homogenization for tumor samples and Dounce homogenization for all other samples (Table
34
35 1).
36
37 We confirmed the isolation of mitochondria by Western blotting analysis of the mitochondrial and
38 cytosolic fractions isolated from each tissue (Figure 2A). We used Cytochrome-c and α-tubulin
39
40 antibodies as mitochondrial and cytosolic markers, respectively, and GAPDH, β-actin, or VDAC as
41
42
loading controls (Table 2). Different loading controls were used for the different type of tissues due
43 to the different modulation of the loading controls in each subcellular fraction. For all organs, we
44
45 observed a significantly higher level of mitochondrial marker expression in mitochondrial fractions
46 than in cytosolic fractions, thereby confirming the presence of mitochondria. The cytosolic fractions
47
48 presented comparatively low levels of Cytochrome-c, further confirming the isolation of
49
mitochondria from the cytosol (Figure 2A).
50
51
52
After organelle isolation, we explored the choice of metabolite extraction protocol. Analysis of intact
53 mitochondria by high-resolution magic-angle spinning (HR-MAS) NMR provides lower resolution
54
55 and sensitivity than a 1H NOESY of a mitochondrial extract33, crucial parameters when working with
56 small sample amounts (such as mitochondrial fractions, in our case). To our knowledge, no study has
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 10 of 26

1
2
3 reported the specific optimization of a protocol for extracting mitochondrial metabolites; however,
4
5 some protocols have been established for analyzing limited-quantity samples34. The commonly used
6 Folch extraction method involves adding a chloroform/methanol/water mixture to simultaneously
7
8 extract aqueous and hydrophobic metabolites35; however, alternative methods that avoid organic
9
solvents may represent a more optimal choice for the analysis of small-quantity samples. The use of
10
11 NMR buffer (Buffer 2) in place of organic solvents has been used in combination with sonication36
12
13 or quenching in liquid nitrogen37 to avoid metabolite loss during solvent evaporation. Moreover,
14 sonication alone or combined with final sample ultrafiltration provides better quality spectra than the
15
16 Folch method37. Given these data, we chose to resuspend mitochondrial extracts in NMR buffer
17
(Buffer 2), quench them in liquid nitrogen, and then carry out centrifugal ultrafiltration to extract the
18
19 metabolites for the present study. We also introduced 3 KDa centrifugal filters to eliminate broad
20
21 protein signals during centrifugal ultrafiltration.
22
23 We first compared metabolite extraction carried out using two methods, namely the optimized
24 ("NOESY with filtration") and the traditional extraction ("NOESY Folch extraction"), in liver and
25
26 breast tumor mitochondria samples, which represent normal and diseased tissues, respectively
27 (Figure 2B and Supporting Information 6). A comparison of the obtained NMR spectra
28
29 demonstrated a significantly higher metabolite recovery when using the "NOESY with filtration"
30
31
method. This was further confirmed via the absolute quantification of a series of selected metabolites
32 (valine, lactate, glutamate, glucose, fumarate, tyrosine, histidine, and phenylalanine) over the entire
33
34 spectral range at different concentration levels (Supporting Information 7).
35
36 Next, we evaluated the efficacy of our method to remove protein signals by comparing the obtained
37 spectra with the NMR spectra resulting from a specific NMR experiment that filters out
38
39 macromolecule signals (Carr–Purcell–Meiboom–Gil, CPMG method) from an unfiltered sample
40
41
(Figure 2C). The CPMG method consists of a 1H 1D NMR experiment using a relaxation filter that
42 minimizes the intensity of broad signals38,39. Therefore, we compared the CPMG analysis of unfiltered
43
44 liver mitochondria samples ("CPMG without filtration") with NOESY analysis ("NOESY without
45 filtration" and "NOESY with filtration") to choose the optimal method. The NOESY experiment
46
47 without filtration showed broad protein signals, altering the baseline of the whole spectrum, while the
48
CPMG analysis showed the partial deletion of these protein signals in specific spectra regions, leaving
49
50 cleaner metabolite signals (Figure 2C). However, the NOESY experiment with filtration gave rise to
51
52 better-quality spectra with further reduced protein signals, enabling the correct quantification of the
53 metabolite signals. Therefore, metabolite quantification relied on acquiring a 1D 1H spectrum with
54
55 water suppression by presaturation40 and a short NOESY mixing time to improve phase and
56
57
58
59
60 ACS Paragon Plus Environment
Page 11 of 26 Analytical Chemistry

1
2
3 baseline41. Spectra from 1H NOESY experiments with the previous settings usually result in
4
5 reproducible and quantitatively accurate data, which remains crucial for metabolomics analysis42.
6
7 The optimized methodology for NMR analysis of mitochondria isolated from tissues consisted of
8
tissue homogenization using Dounce or Ultra-turrax for soft and hard tissues, respectively, followed
9
10 by differential centrifugation to obtain the mitochondrial fraction. Metabolite extraction from
11
12 mitochondria involved dissolving the extract in the NMR buffer, quenching the samples, and filtering
13 them to eliminate broad protein signals. Finally, NMR analysis of mitochondria was carried out using
14
15 a 1H NOESY experiment. The NMR spectra obtained from this procedure resulted in good-quality
16 spectra, making this methodology suitable for studying metabolic alterations occurring in
17
18 mitochondrial diseases.
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
Figure 2. (A) Representative Western blots and protein quantitation confirming mitochondrial
47 isolation from tissues. Protein quantification was performed by densitometry analysis (average ±
48
49 SEM (n=3)). Cytochrome-c: mitochondrial marker; α-tubulin: cytosolic marker; GAPDH, β-actin,
50 and VDAC: loading controls. M: mitochondria; C: cytosol. (B) Aliphatic region of the 1H-NMR
51
52 spectra of liver mitochondria obtained by the "NOESY with filtration" and the "NOESY Folch
53
extraction" metabolite extraction protocols. (C) Comparison of the NOESY spectra aliphatic
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 12 of 26

1
2
3 region obtained with/without sample filtration compared to CPMG without filtration. Regions
4
5 of the spectra experiencing the most significant changes are highlighted in grey.
6
7
8
9
10 A specific mitochondrial metabolomic profile characterizes distinct tissue types
11
12 Using mitochondria isolation through differential centrifugation, buffer plus filtration metabolite
13
14 extraction, and NOESY NMR on healthy (liver, lung, kidney, and brain) and breast tumor tissues
15 from a mouse model allowed us to obtain distinct mitochondrial metabolic profiles. Figure 3A depicts
16
17 the assigned 1H-NOESY spectra of mitochondrial extracts isolated from the various tissue types under
18
investigation. We adjusted the number of required scans for each experiment based on the weight of
19
20 the mitochondrial extract obtained from each tissue (Table 1). After carefully analyzing all the
21
22 spectra, we identified 45 metabolites, including amino acids, organic acids, sugars, lipids, and
23 nucleotides (Supporting Information 3). We observed high-quality spectra for breast tumors,
24
25 kidney, and brain, and an excellent signal-to-noise ratio for the liver, perhaps due to the relatively
26 high mitochondrial content of liver cells (1,000-4,000 per cell)43. The signal-to-noise ratio was a bit
27
28 lower for the spectra of lung mitochondria, which may be due to the relatively low mass of this
29
organ44.
30
31
32 Further analysis of the spectra (Figure 3A) provided evidence for the specificity of certain
33 mitochondrial metabolites in specific tissues. For example, we only detected ornithine and maltose in
34
35 the liver, betaine in the kidney, acetone in the lung, and nicotinurate, N-acetyl aspartate, and 4-
36 aminobutyrate in the brain (Figure 3A). We next performed a PCA on mice from three independent
37
38 litters to evaluate potential differences in the composition of tissue-specific mitochondrial
39
metabolomic profiles (Figure 3B). The resulting score plot demonstrated that samples from the same
40
41 tissues tended to cluster together, confirming that the mitochondrial metabolomic profiles
42
43 systematically differ between organs (Figure 3B). Interestingly, the lung exhibited some
44 compositional similarity to breast tumors, while brain and liver samples differed from the remaining
45
46 tissues.
47
48 The loading plot obtained from this analysis (Figure 3C) facilitated the identification of the
49
mitochondrial metabolites that supported the discrimination between tissues. Amino acids
50
51 significantly contributed to the differentiation of liver and kidney samples, whereas nucleotides
52
53 played a crucial role in discriminating brain tissue samples. Lipids and organic acids (such as lactate,
54 citrate, and ascorbate) contributed to separating lung and breast tumor samples. We observed
55
56 exceptionally high amounts of the organic acid succinate in lung samples, while breast tumor samples
57
58
59
60 ACS Paragon Plus Environment
Page 13 of 26 Analytical Chemistry

1
2
3 contained high amounts of creatine and acetate. Branched amino acids (such as valine, leucine, and
4
5 isoleucine) and alanine characterized the kidney and breast tumors, while brain samples displayed
6 high levels of aspartate. The non-essential amino acids glutamine and glutamate also contributed to
7
8 separating breast tumor samples.
9
10 Overall, the application of our optimized methodology to breast tumors, liver, lung, kidney, and brain
11
12 tissues resulted in a specific metabolic profile of mitochondria due to the differential content of amino
13 acids, organic acids, lipids, sugars, and nucleotides.
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35 Figure 3. (A) Representative 1H-NOESY spectra from isolated mitochondria from healthy and
36
37 tumorigenic tissues. The NMR region between 3.3 ppm and 6 ppm corresponding to water and
38
sucrose/mannitol signals present in Buffer 1 were removed to facilitate visual inspection. (B) PCA
39
40 score plot of mitochondrial metabolomic profiles obtained from healthy and tumorigenic
41
42 tissues. (C) PCA loading plot colored according to metabolite chemical nature. Ace: acetate, Ala:
43 alanine, AMP: adenosine monophosphate, Asc: ascorbate, Cre: creatine, Cho: choline, Cit: citrate,
44
45 For: formate, Gln: glutamine, Glu: glutamate, GPC: glycerophosphocholine, Ile: isoleucine, IMP:
46
inosine monophosphate, Lac: lactate, Leu: leucine, Lys: lysine, Met: methionine, Myo: myo-inositol,
47
48 Naa: N-acetyl aspartate, Nia: niacinamide, Nic: Nicotinurate, Phe: phenylalanine, PCre:
49
50 phosphocreatine, Suc: succinate, Tau: taurine, Thr: threonine, UDP-der: UDP-derivative, Ura: uracil,
51 Val: valine.
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 14 of 26

1
2
3 Mitochondrial metabolomic profiles differ from whole-cell metabolomic profiles in distinct
4
5 tissue types
6
7 We next performed a similar analysis using total tissue samples to evaluate the differences between
8
the mitochondrial profile and the profile of the whole cell (Supporting Information 7). PCA of
9
10 whole cell metabolomic profiles showed good separation between tissues (Supporting Information
11
12 8A), with breast tumor and kidney samples closer than compared to the PCA for mitochondrial
13 metabolomic profiles. The loading plot of whole cell samples showed similarities to the mitochondrial
14
15 samples (Supporting Information 8B). Nucleotides again contributed to the separation of brain
16 samples, while amino acids differentiated kidney and breast tumor samples from the remaining
17
18 samples, and the lung displayed high choline content. In contrast to the mitochondrial analysis,
19
several sugar metabolites, such as glucose, maltose, and ribose, separated liver samples from other
20
21 tissues.
22
23 We performed a pairwise comparison (mitochondria vs. whole tissue; Figure 4A) for each different
24
25 organ by first generating a validated discriminating OPLS-DA model and selecting metabolites with
26 a variable importance plot (VIP) values > 1, which were then subjected to univariate statistical
27
28 analysis (Supporting Information 9). Figure 4A reports those metabolites displaying the most
29
significant differences in intensity between isolated mitochondrial samples (bottom-dark colored bar)
30
31 and whole-tissue samples (upper-light colored bar). In general, isolated mitochondria contained
32
33 elevated levels of glycerophosphocholine, valine, leucine, isoleucine, lysine, and UDP-derivatives
34 and decreased levels of phosphocreatine, glutamate, glutamine, glutathione, lactate, myo-inositol, and
35
36 taurine compared to the whole cell samples (Figure 4A). These data agree with the distinct metabolic
37 pathways that take place in the mitochondria and the cytosol (Figure 4B). For example, amino acids
38
39 can supply tricarboxylic acid (TCA) cycle intermediates in the mitochondria45, while lactate,
40
41
phosphocreatine, and myo-inositol mainly locate in the cytosol46–48. Metabolites such as glutathione
42 and taurine are synthesized in the cytosol and move into the mitochondria using specific carrier
43
44 molecules49,50, therefore their levels will change depending on the state of the cell.
45
46 Our analysis also supported the discovery that specific metabolites exhibited organ-specific changes.
47 Tumor and brain samples possessed a higher proportion of most metabolites in whole cells than
48
49 mitochondria, whereas lung and liver exhibited higher relative quantities of most metabolites in
50
51
mitochondria (Figure 4A). Kidney samples displayed different tendencies in metabolite levels; for
52 example, we observed higher amounts of choline, P-choline, taurine, glutamate, myo-inositol, and
53
54 lactate in whole cells but higher amounts of the amino acids alanine, isoleucine, leucine, lysine,
55 methionine, phenylalanine, threonine, and valine and a UDP-derivative in mitochondria (Figure 4A).
56
57
58
59
60 ACS Paragon Plus Environment
Page 15 of 26 Analytical Chemistry

1
2
3 Whole-cell samples from breast tumors showed significantly increased levels of lactate and P-choline
4
5 compared to other organs (Figure 4A), which agrees with the metabolic reprogramming known to
6 occur in tumor cells. Lactate levels tend to increase in tumor cells due to the Warburg effect, in which
7
8 the final product of anaerobic glycolysis, pyruvate, becomes converted into lactate51. Additionally,
9
choline metabolism also undergoes alterations in tumor cells, including breast cancer cells52.
10
11
12 We also performed an ANOVA analysis on mitochondrial and whole cell samples from all organs to
13 obtain a more detailed description of any differences. The results demonstrated the alteration of 42
14
15 metabolites when comparing mitochondria isolated from distinct tissues but only 27 when performing
16 the same analysis in whole cell samples (Supporting Information 10).
17
18
Overall, these data confirm a significant contribution from organelle-specific metabolites and the
19
20 relevance of the organelle-specific analysis.
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 16 of 26

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41 Figure 4. (A) Bar graphs corresponding to significantly altered metabolites from whole cells (top)
42
43 and mitochondria (bottom) in healthy and tumorigenic tissues. Concentration values normalized
44
to total intensity. Data represented as mean ± SEM.* p < 0.05, ** p < 0.01, *** p < 0.001. T: breast
45
46 tumor; L: lung; V: liver; B: brain; K: kidney; P-choline: phosphocholine; GPC: glycerophosphocholine;
47
48 P-creatine: phosphocreatine; UDP-der: UDP-derivative. (B) Summary of metabolic pathways in the
49
mitochondria and cytosol and their interconnections. Mitochondrial metabolites in green are
50
51 increased, and in red are decreased. Glucose-6-P: glucose-6-phosphate, myo-inositol 3-P: myo-
52
53 inositol 3-phosphate, LDH: lactate dehydrogenase, Cys: cysteine, Tau: taurine, GSH: glutathione, α-
54
KG: α-ketoglutarate, Leu: leucine, Ile: Isoleucine, Lys: Lysine, ADP: adenosine monophosphate,
55
56 ATP: adenosine triphosphate, Cre: creatine, PCre: phosphocreatine.
57
58
59
60 ACS Paragon Plus Environment
Page 17 of 26 Analytical Chemistry

1
2
3 Optimized analysis of mitochondrial metabolic profiles in breast cancer models distinguishes
4
5 metastatic sites
6
7 We validated our mitochondrial metabolic profiling methodology by analyzing the impact of
8
metastasis on mitochondrial metabolism in target and non-target tissues using a mouse model of
9
10 metastatic breast cancer. We induced MDA-MB-231-Luc tumors in five NOD/SCID mice, which we
11
12 extracted after euthanizing them when tumors reached 1.0 cm³, approximately 42 days after
13 inoculation. Previous work by our group revealed the complete invasion of the lungs at day 35, while
14
15 liver did not present signs of metastasis53. We compared the metabolic profiles of mitochondria
16 isolated from a breast cancer-related metastatic site (lung) and non-metastatic sites (liver and kidney)
17
18 from healthy mice and breast cancer model mice (Figure 5, Supporting Information 11).
19
20 The PCA of mitochondria samples isolated from tissues from healthy and breast cancer model mice
21
22 revealed robust discrimination between healthy and breast cancer model mice for the lungs
23 (metastatic site), a slight discrimination in the liver (non-metastatic site), and a lack of separation in
24
25 the kidney (non-metastatic site) (Figure 5A). The OPLS-DA model analysis of this data provided a
26 similar outcome (Table 3), providing evidence for the significant impact of breast cancer metastasis
27
28 on mitochondria in the lung. Notably, we failed to obtain an OPLS-DA model for liver and kidney
29
mitochondria samples where metastasis does not tend to occur in this model. This result agrees with
30
31 a previous study reporting that MDA-MB-231-Luc tumors produce multiple organ metastases,
32
33 especially in the lung and axillary lymph node53. Overall, these data demonstrate the validity of the
34 methodology for metabolic studies of mitochondria.
35
36
37 Table 3. Statistical values for the OPLS-DA models obtained from the comparison between the
38
39 healthy and breast cancer mouse models
40 Mitochondria Total tissue
41
42 Lung Liver Kidney Lung Liver Kidney
43 R2Y(cum) 0.756 - - 0.934 0.853 0.973
44
Q2(cum) 0.614 - - 0.866 0.676 0.735
45
46 CV-Anova p-value 0.0355 - - 0.0014 0.444 0.370
47 Intercept R2 0.229 - - 0.425 0.781 0.849
48
49 Intercept Q2 -0.652 - - -0.695 -0.148 -0.037
50
51
52 R2Y (cum): goodness of the fit; Q2(cum): goodness of the prediction. R2Y(cum) and Q2(cum) values closer to a value of 1
53 indicate a reliable model. Intercept R2 and Q2: Cumulative R2 and Q2 values obtained from the permutation analysis of 200
54 permutation tests. Intercept R2 below 0.3-0.4 and intercept Q2 below zero demonstrates model validity.
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 18 of 26

1
2
3 We conducted a univariate statistical analysis on mitochondrial metabolites displaying VIP values of
4
5 >1 in the OPLS-DA model when comparing healthy and metastatic tissues (see Supporting
6 Information 11). Figure 5B shows the boxplot of those mitochondrial metabolites exhibiting
7
8 significant changes (p<0.05) when comparing healthy and metastatic tissues and the corresponding
9
NMR signal. Mitochondria from metastatic lung tissues had higher levels of eight metabolites when
10
11 compared to healthy tissues (noted by black arrows in Figure 5C). Choline and phosphocholine, both
12
13 exhibiting increased levels (known tumor-associated metabolic alterations52), were found in
14 metastatic lung mitochondria. The RAS (rat sarcoma) oncogenic signaling pathway and the oncogenic
15
16 transcription factor HIF1 (hypoxia-inducible factor 1) regulate the expression of the CHK1
17
(checkpoint kinase 1) and CTL1 (choline transporter-like protein 1) choline metabolism-associated
18
19 enzymes, and their upregulation increases the total choline intracellular concentration54. We also
20
21 observed an increase in glucose and lactate in metastatic lung mitochondria, which correlates with
22 the impact of the Warburg effect on cancer cells51. Many cancers, including breast cancer, are
23
24 characterized by elevated lactate concentrations that can be explained by altered glucose metabolism
25 by different tumor suppressor genes and oncoproteins55. HIF1 upregulates the expression of the
26
27 GLUT1 (glucose transporter 1) and GLUT2 (glucose transporter 2), and the HK2 (hexokinase 2) that
28
initiates glycolysis55. Despite the relevance of glucose and lactate in tumor cell metabolism, metastatic
29
30 liver mitochondria did not display alterations to these metabolites, which may reflect this tissue's
31
32 lower degree of metastasis. Creatine, whose level increases in metastatic lung mitochondria, plays an
33 essential role in cancer cells by maintaining energy homeostasis. Furthermore, increased creatine
34
35 kinase levels occur in distinct cancers56. Metastatic lung mitochondria also had increased levels of
36 glutamate, alanine, and methionine, supporting the theory of dysregulated amino acid metabolism in
37
38 cancer cells57. The increased levels of glutamate could also relate to tumor-associated increases in
39
40
glutaminolysis58, which occurs when the MYC oncogene prompts the overexpression of the SLC1A5
41 (solute carrier family 1 member 5) glutamine transporter and upregulates the glutaminase enzyme
42
43 (Figure 5C).
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
Page 19 of 26 Analytical Chemistry

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22 Figure 5. (A) PCA score plots of mitochondrial metabolite profiles of samples isolated from
23
24 healthy and metastatic mouse models. (B) Mitochondrial metabolite signals and boxplots of
25
26
relevant metabolites for group discrimination. Data represented as mean ± SEM. * p < 0.05, ** p
27 < 0.01, *** p < 0.001. (C) Choline and glutamine metabolic pathways and the Warburg effect
28
29 in cancer. GLS: glutaminase, SLC1A5: Solute Carrier Family 1 Member 5, GLUT1/2: glucose
30 transporter 1/2, HIF1-α: hypoxia-inducible factor 1- α, HK: hexokinase, LDHA: lactate
31
32 dehydrogenase A, CHK: choline kinase, GPC: glycerophosphocholine.
33
34
35
36 To further test the relevance of our method, we performed a similar comparison, assessing the whole
37
38 cell rather than mitochondria (Table 3, Supporting Information 12). Interestingly, in this analysis
39 we failed to detect alterations in alanine, lactate, phosphocholine, and methionine in the metastatic
40
41 lung. Choline and glutamate showed the same trend in both whole-cell and mitochondria of metastatic
42
lung samples, while other metabolites (glycerophosphocholine, myo-inositol, leucine, lysine, and
43
44 valine) displayed significant changes only in whole-cell analysis.
45
46 These findings support the hypothesis that mitochondrial and whole-cell metabolite analysis provides
47
48 differing but complementary information, further evidence for the utility of organelle-specific
49 analysis in exploring the mechanisms controlling metastasis.
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 20 of 26

1
2
3 Conclusions
4
5 In this work, we describe an optimized methodology to study metabolomic profiles of mitochondria
6
7 isolated from mouse tissues through NMR spectroscopy. The procedure involves: i) isolation of
8
mitochondria through differential centrifugation, ii) extraction of metabolites using an optimized
9
10 protocol based on quenching, and iii) NMR analysis. After confirming the presence of mitochondria
11
12 in the isolated fractions by Western blotting, we compared the metabolomic content of whole cells
13 and mitochondria in healthy and tumorigenic mouse tissues. This comparison provided evidence of
14
15 significant changes in the levels of many metabolites; furthermore, we observed specific changes
16 occurring in the mitochondria that were not observed in the whole-cell analysis. Our proposed
17
18 methodology is relatively fast and straightforward, avoids solvent evaporation to prevent metabolite
19
loss/degradation, and does not destroy samples, which allows further analysis with complementary
20
21 analytical techniques (e.g., mass spectrometry). Moreover, our methodology allows direct
22
23 quantification of metabolite levels without a calibration curve. Overall, robust and reproducible
24 metabolic data can be obtained in only a few hours.
25
26 Our methodology can also be applied to tissue samples obtained from animal models or human
27
28 biopsies and could provide helpful information for clinical and preclinical research. Additionally, our
29
methodology may find use in research projects focused on identifying biomarkers in diseases that
30
31 impact mitochondrial metabolism, such as Alzheimer's disease, Parkinson's disease, or diabetes, using
32
33 small muscle or fat biopsies before the manifestation of the clinical symptoms.
34
35 Given the findings observed when comparing mitochondria isolated from tissues of healthy and
36 metastatic breast cancer mouse models, we also hope our methodology will play an essential role in
37
38 cancer research. The increase in certain metabolite levels in lung mitochondria associated with
39
tumorigeneses/metastasis suggests that components of the associated metabolic pathways could serve
40
41 as therapeutic targets. Therefore, mitochondria constitute promising targets for developing novel
42
43 anticancer agents, and this method could also provide critical information regarding the impact of
44 novel anticancer drugs on mitochondrial metabolism.
45
46
47
48
Supporting Information
49
50
51
Supporting Information 1. Graph representing the percentage change in NOD/SCID mice body
52 weight (%) following the induction of tumorigenesis via the inoculation of MDA-MB-231-Luc cells
53
54 into the second left mammary fat pad. (PPTX)
55
56
57
58
59
60 ACS Paragon Plus Environment
Page 21 of 26 Analytical Chemistry

1
2
3 Supporting Information 2. Parameters for the NMR analysis of whole cell samples from tissues.
4
5 (PDF)
6
7 Supporting Information 3. NMR signal assignation after analysis of mitochondrial samples isolated
8
from tissues. (XLSX)
9
10
Supporting Information 4. Integration template for the integration of NMR signals in MestreNova.
11
12 (TXT)
13
14 Supporting Information 5. Aliphatic region of the 1H-NMR spectra of tumor mitochondria samples
15
16 obtained following Dounce or Ultra-turrax homogenization. Both methods employed the same tumor
17 sample. The spectral regions experiencing the most significant changes are highlighted in grey.
18
19 (PPTX)
20
21 Supporting Information 6. Aliphatic region of the 1H-NMR spectra of breast tumor mitochondria
22
23 samples obtained following the optimized "NOESY with filtration" method and the traditional
24 "NOESY Folch extraction" method. Signals could be easily assigned after buffer extraction with
25
26 filtration, while low signals were detected using Folch extraction. (PPTX)
27
28 Supporting Information 7. Quantification of 1H-NMR spectra of breast tumor mitochondria samples
29 obtained following the optimized "NOESY with filtration" method and the traditional "Folch
30
31 extraction" method. (XLSX)
32
33 Supporting Information 8. PCA of whole cell metabolomic profiles of distinct tissues. (A) 3D score
34
35 plot. (B) 3D loading plot colored by metabolite type. 4ab: 4-aminobutyrate, Ala: alanine, AMP:
36 adenosine monophosphate, Asc: ascorbate, Asn: asparagine, Cre: creatine, Cho: choline, Cit: citrate,
37
38 DMA: dimethylamine, Glu: glutamate, Gluc: glucose GPC: glycerophosphocholine, His: histidine,
39 Ile: isoleucine, Ino: Inosine, Lac: lactate, Leu: leucine, Mal: maltose, Met: methionine, Myo: myo-
40
41 inositol, Naa: N-acetyl aspartate, Nia: niacinamide, Nic: Nicotinurate, Orn: ornithine, Phe:
42
43
phenylalanine, PCre: phosphocreatine, Rib: ribose, Suc: succinate, Try: tryptophan, UDP-NAG: UDP
44 N-acetylglucosamine, Uri: uridine, Val: valine. (PPTX)
45
46 Supporting Information 9. Mean, standard deviation, VIP values, and p-values for the comparative
47
48 analysis of mitochondria and whole cell metabolomic profiles in each tissue type. Light blue: Mean
49 and standard deviation of mitochondrial samples; dark blue: mean and standard deviation of whole-
50
51 cell samples; green: VIP values above 1; and red: p-values under 0.05. (XLSX)
52
53 Supporting Information 10. Significantly altered metabolites obtained from ANOVA analysis of
54
55 mitochondrial and whole-cell samples from distinct tissues. Red: breast tumor samples; blue: lung
56 samples; green: liver samples; purple: brain samples; and yellow: kidney samples. (XLSX)
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 22 of 26

1
2
3 Supporting Information 11: Mean, standard error of the mean, VIP, and p-values for the
4
5 comparative analysis of mitochondrial metabolomic profiles from healthy vs. metastatic tissues.
6 (XLSX)
7
8
Supporting Information 12: Mean, standard error of the mean, VIP, and p-values for the
9
10 comparative analysis of whole cell metabolomic profiles from healthy vs. metastatic tissues. (XLSX)
11
12
13
14 References
15
16 (1) Duchen, M. R. Roles of Mitochondria in Health and Disease. Diabetes 2004, 53 (SUPPL. 1).
17 https://doi.org/10.2337/diabetes.53.2007.s96.
18
(2) Murphy, E.; Ardehali, H.; Balaban, R. S.; DiLisa, F.; Dorn, G. W.; Kitsis, R. N.; Otsu, K.; Ping, P.; Rizzuto,
19
R.; Sack, M. N.; Wallace, D.; Youle, R. J. Mitochondrial Function, Biology, and Role in Disease: A
20
Scientific Statement from the American Heart Association. Circ. Res. 2016, 118 (12), 1960–1991.
21
https://doi.org/10.1161/RES.0000000000000104.
22
(3) Modica-Napolitano, J. S.; Singh, K. K. Mitochondrial Dysfunction in Cancer. Mitochondrion 2004, 4 (5-
23
6 SPEC. ISS.), 755–762. https://doi.org/10.1016/j.mito.2004.07.027.
24
(4) Scheid, A. D.; Beadnell, T. C.; Welch, D. R. Roles of Mitochondria in the Hallmarks of Metastasis. Br. J.
25
Cancer 2021, 124 (1), 124–135. https://doi.org/10.1038/s41416-020-01125-8.
26
(5) Zong, W.-X.; Rabinowitz, J. D.; White, E. Mitochondria and Cancer. Mol Cell 2016, 61 (5), 667–676.
27
28 https://doi.org/10.1016/j.molcel.2016.02.011.Mitochondria.
29 (6) Simonnet, H.; Alazard, N.; Pfeiffer, K.; Gallou, C.; Béroud, C.; Demont, J.; Bouvier, R.; Schägger, H.;
30 Godinot, C. Low Mitochondrial Respiratory Chain Content Correlates with Tumor Aggressiveness in
31 Renal Cell Carcinoma. Carcinogenesis 2002, 23 (5), 759–768. https://doi.org/10.1093/carcin/23.5.759.
32 (7) Nicholson, J. K.; Lindon, J. C.; Holmes, E. “Metabonomics”: Understanding the Metabolic Responses of
33 Living Systems to Pathophysiological Stimuli via Multivariate Statistical Analysis of Biological NMR
34 Spectroscopic Data. Xenobiotica 1999, 29 (11), 1181–1189.
35 https://doi.org/10.1080/004982599238047.
36 (8) Miller, I. J.; Peters, S. R.; Overmyer, K. A.; Paulson, B. R.; Westphall, M. S.; Coon, J. J. Real-Time Health
37 Monitoring through Urine Metabolomics. npj Digit. Med. 2019, 2 (1). https://doi.org/10.1038/s41746-
38 019-0185-y.
39 (9) Kok, M. G. M.; Nix, C.; Nys, G.; Fillet, M. Targeted Metabolomics of Whole Blood Using Volumetric
40 Absorptive Microsampling. Talanta 2019, 197 (January), 49–58.
41 https://doi.org/10.1016/j.talanta.2019.01.014.
42 (10) Gowda, G. A. N.; Shanaiah, N.; Cooper, A.; Maluccio, M.; Raftery, D. Bile Acids Conjugation in Human
43 Bile Is Not Random: New Insights from 1H-NMR Spectroscopy at 800 MHz. Lipids 2009, 44 (6), 527–
44 535. https://doi.org/10.1007/s11745-009-3296-4.
45 (11) Lima, A. R.; Pinto, J.; Bastos, M. de L.; Carvalho, M.; Guedes de Pinho, P. NMR-Based Metabolomics
46 Studies of Human Prostate Cancer Tissue. Metabolomics 2018, 14 (7), 0.
47 https://doi.org/10.1007/s11306-018-1384-2.
48 (12) Mirnezami, R.; Jiménez, B.; Li, J. V.; Kinross, J. M.; Veselkov, K.; Goldin, R. D.; Holmes, E.; Nicholson, J.
49 K.; Darzi, A. Rapid Diagnosis and Staging of Colorectal Cancer via High-Resolution Magic Angle Spinning
50 Nuclear Magnetic Resonance (HR-MAS NMR) Spectroscopy of Intact Tissue Biopsies. Ann. Surg. 2014,
51 259 (6), 1138–1149. https://doi.org/10.1097/SLA.0b013e31829d5c45.
52 (13) Wang, H.; Zhang, H.; Deng, P.; Liu, C.; Li, D.; Jie, H.; Zhang, H.; Zhou, Z.; Zhao, Y. L. Tissue Metabolic
53 Profiling of Human Gastric Cancer Assessed by 1H NMR. BMC Cancer 2016, 16 (1), 1–12.
54 https://doi.org/10.1186/s12885-016-2356-4.
55 (14) Naz, S.; Moreira Dos Santos, D. C.; García, A.; Barbas, C. Analytical Protocols Based on LC-MS, GC-MS
56 and CE-MS for Nontargeted Metabolomics of Biological Tissues. Bioanalysis 2014, 6 (12), 1657–1677.
57
58
59
60 ACS Paragon Plus Environment
Page 23 of 26 Analytical Chemistry

1
2
3 https://doi.org/10.4155/bio.14.119.
4 (15) Domingo-Ortí, I.; Lamas-Domingo, R.; Ciudin, A.; Hernández, C.; Herance, J. R.; Palomino-Schätzlein,
5 M.; Pineda-Lucena, A. Metabolic Footprint of Aging and Obesity in Red Blood Cells. Aging (Albany. NY).
6 2021, 13 (4), 4850–4880. https://doi.org/10.18632/aging.202693.
7 (16) Mayers, J. R.; Wu, C.; Clish, C. B.; Kraft, P.; Torrence, M. E.; Fiske, B. P.; Yuan, C.; Bao, Y.; Townsend,
8 M. K.; Tworoger, S. S.; Davidson, S. M.; Papagiannakopoulos, T.; Yang, A.; Dayton, T. L.; Ogino, S.;
9 Stampfer, M. J.; Giovannucci, E. L.; Qian, Z. R.; Rubinson, D. A.; Ma, J.; Sesso, H. D.; Gaziano, J. M.;
10 Cochrane, B. B.; Liu, S.; Wactawski-Wende, J.; Manson, J. E.; Pollak, M. N.; Kimmelman, A. C.; Souza,
11 A.; Pierce, K.; Wang, T. J.; Gerszten, R. E.; Fuchs, C. S.; Vander Heiden, M. G.; Wolpin, B. M. Elevation
12 of Circulating Branched-Chain Amino Acids Is an Early Event in Human Pancreatic Adenocarcinoma
13 Development. Nat. Med. 2014, 20 (10), 1193–1198. https://doi.org/10.1038/nm.3686.
14 (17) Kaddurah-Daouk, R.; Kristal, B. S.; Weinshilboum, R. M. Metabolomics: A Global Biochemical Approach
15 to Drug Response and Disease. Annu. Rev. Pharmacol. Toxicol. 2008, 48, 653–683.
16 https://doi.org/10.1146/annurev.pharmtox.48.113006.094715.
17 (18) Armiñán, A.; Palomino-Schätzlein, M.; Deladriere, C.; Arroyo-Crespo, J. J.; Vicente-Ruiz, S.; Vicent, M.
18 J.; Pineda-Lucena, A. Metabolomics Facilitates the Discrimination of the Specific Anti-Cancer Effects
19 of Free- and Polymer-Conjugated Doxorubicin in Breast Cancer Models. Biomaterials 2018, 162, 144–
20
153. https://doi.org/10.1016/j.biomaterials.2018.02.015.
21
(19) O’Keefe, S. J. D.; Li, J. V.; Lahti, L.; Ou, J.; Carbonero, F.; Mohammed, K.; Posma, J. M.; Kinross, J.; Wahl,
22
E.; Ruder, E.; Vipperla, K.; Naidoo, V.; Mtshali, L.; Tims, S.; Puylaert, P. G. B.; Delany, J.; Krasinskas, A.;
23
Benefiel, A. C.; Kaseb, H. O.; Newton, K.; Nicholson, J. K.; De Vos, W. M.; Gaskins, H. R.; Zoetendal, E.
24
G. Fat, Fibre and Cancer Risk in African Americans and Rural Africans. Nat. Commun. 2015, 6.
25
https://doi.org/10.1038/ncomms7342.
26
(20) Chan, C. Y. E.; Koh, P. K.; Mal, M.; Cheah, P. Y.; Eu, K. W.; Backshall, A.; Cavill, R.; K, N. J.; C, K. H.
27
Metabolic Profiling of Human Colorectal Cancer Using High-Resolution Magic Angle Spinning Nuclear
28
Magnetic Resonance (HR-MAS NMR) Spectroscopy and Gas Chromatography Mass Spectrometry
29
30 (GC/MS). J. Proteome Res. 2009, 8, 352–361.
31 (21) Pan, D.; Lindau, C.; Lagies, S.; Wiedemann, N.; Kammerer, B. Metabolic Profiling of Isolated
32 Mitochondria and Cytoplasm Reveals Compartment-Specific Metabolic Responses. Metabolomics
33 2018, 14 (5), 1–12. https://doi.org/10.1007/s11306-018-1352-x.
34 (22) Liu, X.; Xu, G. Recent Advances in Using Mass Spectrometry for Mitochondrial Metabolomics and
35 Lipidomics - A Review. Anal. Chim. Acta 2018, 1037, 3–12. https://doi.org/10.1016/j.aca.2017.11.080.
36 (23) Gravel, S.-P.; Andrzejewski, S.; Avizonis, D.; St-Pierre, J. Stable Isotope Tracer Analysis in Isolated
37 Mitochondria from Mammalian Systems. Metabolites 2014, 4 (2), 166–183.
38 https://doi.org/10.3390/metabo4020166.
39 (24) Carneiro, T. J.; Araújo, R.; Vojtek, M.; Gonçalves-Monteiro, S.; Diniz, C.; de Carvalho, A. L. M. B.;
40 Marques, M. P. M.; Gil, A. M. Multi-Organ NMR Metabolomics to Assess in Vivo Overall Metabolic
41 Impact of Cisplatin in Mice. Metabolites 2019, 9 (11), 1–17. https://doi.org/10.3390/metabo9110279.
42 (25) Glinskikh, A.; Snytnikova, O.; Zelentsova, E.; Borisova, M.; Tsentalovich, Y.; Akulov, A. The Effect of
43 Blood Contained in the Samples on the Metabolomic Profile of Mouse Brain Tissue: A Study by NMR
44 Spectroscopy. Molecules 2021, 26 (11). https://doi.org/10.3390/molecules26113096.
45 (26) Ling, Y. S.; Liang, H. J.; Chung, M. H.; Lin, M. H.; Lin, C. Y. NMR- and MS-Based Metabolomics: Various
46 Organ Responses Following Naphthalene Intervention. Mol. Biosyst. 2014, 10 (7), 1918–1931.
47 https://doi.org/10.1039/c4mb00090k.
48 (27) Puchades-Carrasco, L.; Palomino-Schätzlein, M.; Pérez-Rambla, C.; Pineda-Lucena, A. Bioinformatics
49 Tools for the Analysis of NMR Metabolomics Studies Focused on the Identification of Clinically
50 Relevant Biomarkers. Brief. Bioinform. 2016, 17 (3), 541–552. https://doi.org/10.1093/bib/bbv077.
51 (28) Pang, Z.; Chong, J.; Zhou, G.; De Lima Morais, D. A.; Chang, L.; Barrette, M.; Gauthier, C.; Jacques, P.
52 É.; Li, S.; Xia, J. MetaboAnalyst 5.0: Narrowing the Gap between Raw Spectra and Functional Insights.
53 Nucleic Acids Res. 2021, 49 (W1), W388–W396. https://doi.org/10.1093/nar/gkab382.
54 (29) Barbieri, L.; Luchinat, E.; Banci, L. Structural Insights of Proteins in Sub-Cellular Compartments: In-
55 Mitochondria NMR. Biochim. Biophys. Acta - Mol. Cell Res. 2014, 1843 (11), 2492–2496.
56 https://doi.org/10.1016/j.bbamcr.2014.06.009.
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 24 of 26

1
2
3 (30) Go, Y. M.; Uppal, K.; Walker, D. I.; Tran, V.; Dury, L.; Strobel, F. H.; Baubichon-Cortay, H.; Pennell, K.
4 D.; Roede, J. R.; Jones, D. P. Mitochondrial Metabolomics Using High-Resolution Fourier-Transform
5 Mass Spectrometry. Methods Mol. Biol. 2014, 1198 (1), 43–73. https://doi.org/10.1007/978-1-4939-
6 1258-2_4.
7 (31) Franko, A.; Baris, O. R.; Bergschneider, E.; Von Toerne, C.; Hauck, S. M.; Aichler, M.; Walch, A. K.;
8 Wurst, W.; Wiesner, R. J.; Johnston, I. C. D.; De Angelis, M. H. Efficient Isolation of Pure and Functional
9 Mitochondria from Mouse Tissues Using Automated Tissue Disruption and Enrichment with Anti-
10 TOM22 Magnetic Beads. PLoS One 2013, 8 (12). https://doi.org/10.1371/journal.pone.0082392.
11 (32) Pallotti, F.; Lenaz, G. Isolation and Subfractionation of Mitochondria from Animal Cells and Tissue
12 Culture Lines. Methods Cell Biol. 2007, 80 (06), 3–44. https://doi.org/10.1016/S0091-679X(06)80001-
13 4.
14 (33) Gogiashvili, M.; Nowacki, J.; Hergenröder, R.; Hengstler, J. G.; Lambert, J.; Edlund, K. HR-MAS NMR
15 Based Quantitative Metabolomics in Breast Cancer. Metabolites 2019, 9 (2).
16 https://doi.org/10.3390/metabo9020019.
17 (34) Mason, S.; Terburgh, K.; Louw, R. Miniaturized 1 H-NMR Method for Analyzing Limited-Quantity
18 Samples Applied to a Mouse Model of Leigh Disease. Metabolomics 2018, 14 (6), 74.
19 (35) Folch, J.; Lees, M.; Sloane Stanley, G. . A Simple Method for the Isolation and Purification of Total
20
Lipides from Animal Tissues. 1957, 226 (1), 497–509.
21
(36) Matheus, N.; Hansen, S.; Rozet, E.; Peixoto, P.; Maquoi, E.; Lambert, V.; Noël, A.; Frédérich, M.; Mottet,
22
D.; De Tullio, P. An Easy, Convenient Cell and Tissue Extraction Protocol for Nuclear Magnetic
23
Resonance Metabolomics. Phytochem. Anal. 2014, 25 (4), 342–349.
24
https://doi.org/10.1002/pca.2498.
25
(37) Gómez-Archila, L. G.; Palomino-Schätzlein, M.; Zapata-Builes, W.; Galeano, E. Development of an
26
Optimized Method for Processing Peripheral Blood Mononuclear Cells for 1H-Nuclear Magnetic
27
Resonancebased Metabolomic Profiling. PLoS One 2021, 16 (2 February), 1–20.
28
https://doi.org/10.1371/journal.pone.0247668.
29
30 (38) Meiboom, S.; Gill, D. Modified Spin-Echo Method for Measuring Nuclear Relaxation Times. Rev. Sci.
31 Instrum. 1958, 29 (8), 688–691. https://doi.org/10.1063/1.1716296.
32 (39) Carr, H. Y.; Purcell, E. M. Effects of Diffusion on Free Precession in Nuclear Magnetic Resonance
33 Experiments. Phys. Rev. 1954, 94 (3), 630–638. https://doi.org/10.1103/PhysRev.94.630.
34 (40) Price, W. S. Water Signal Suppression in NMR Spectroscopy; 1999; Vol. 38.
35 https://doi.org/10.1016/S0066-4103(08)60040-X.
36 (41) Emwas, A. H.; Saccenti, E.; Gao, X.; McKay, R. T.; dos Santos, V. A. P. M.; Roy, R.; Wishart, D. S.
37 Recommended Strategies for Spectral Processing and Post-Processing of 1D 1 H-NMR Data of Biofluids
38 with a Particular Focus on Urine. Metabolomics 2018, 14 (3), 1–23. https://doi.org/10.1007/s11306-
39 018-1321-4.
40 (42) Emwas, A. H.; Roy, R.; McKay, R. T.; Tenori, L.; Saccenti, E.; Nagana Gowda, G. A.; Raftery, D.; Alahmari,
41 F.; Jaremko, L.; Jaremko, M.; Wishart, D. S. Nmr Spectroscopy for Metabolomics Research. Metabolites
42 2019, 9 (7), 123. https://doi.org/10.3390/metabo9070123.
43 (43) Degli Esposti, D.; Hamelin, J.; Bosselut, N.; Saffroy, R.; Sebagh, M.; Pommier, A.; Martel, C.; Lemoine,
44 A. Mitochondrial Roles and Cytoprotection in Chronic Liver Injury. Biochem. Res. Int. 2012, 2012.
45 https://doi.org/10.1155/2012/387626.
46 (44) Bitar, A.; Lisbona, A.; Thedrez, P.; Sai Maurel, C.; Le Forestier, D.; Barbet, J.; Bardies, M. A Voxel-Based
47 Mouse for Internal Dose Calculations Using Monte Carlo Simulations (MCNP). Phys. Med. Biol. 2007,
48 52 (4), 1013–1025. https://doi.org/10.1088/0031-9155/52/4/010.
49 (45) Martínez-Reyes, I.; Chandel, N. S. Mitochondrial TCA Cycle Metabolites Control Physiology and
50 Disease. Nat. Commun. 2020, 11 (1), 1–11. https://doi.org/10.1038/s41467-019-13668-3.
51 (46) Schlattner, U.; Tokarska-Schlattner, M.; Wallimann, T. Mitochondrial Creatine Kinase in Human Health
52 and Disease. Biochim. Biophys. Acta - Mol. Basis Dis. 2006, 1762 (2), 164–180.
53 https://doi.org/10.1016/j.bbadis.2005.09.004.
54 (47) Croze, M. L.; Soulage, C. O. Potential Role and Therapeutic Interests of Myo-Inositol in Metabolic
55 Diseases. Biochimie 2013, 95 (10), 1811–1827. https://doi.org/10.1016/j.biochi.2013.05.011.
56 (48) Adeva-Andany, M.; López-Ojén, M.; Funcasta-Calderón, R.; Ameneiros-Rodríguez, E.; Donapetry-
57
58
59
60 ACS Paragon Plus Environment
Page 25 of 26 Analytical Chemistry

1
2
3 García, C.; Vila-Altesor, M.; Rodríguez-Seijas, J. Comprehensive Review on Lactate Metabolism in
4 Human Health. Mitochondrion 2014, 17, 76–100. https://doi.org/10.1016/j.mito.2014.05.007.
5 (49) Marí, M.; Morales, A.; Colell, A.; García-Ruiz, C.; Fernández-Checa, J. C. Mitochondrial Glutathione, a
6 Key Survival Antioxidant. Antioxidants Redox Signal. 2009, 11 (11), 2685–2700.
7 https://doi.org/10.1089/ars.2009.2695.
8 (50) Hansen, S. H.; Andersen, M. L.; Cornett, C.; Gradinaru, R.; Grunnet, N. A Role for Taurine in
9 Mitochondrial Function. J. Biomed. Sci. 2010, 17 (SUPPL. 1), 1–8. https://doi.org/10.1186/1423-0127-
10 17-S1-S23.
11 (51) Warburg, O.; Wind, F.; Negelein, E. The Metabolism of Tumours in the Body. Br. Med. J. 1927, 8 (6),
12 519–530. https://doi.org/10.1136/bmj.1.3653.74-a.
13 (52) Glunde, K.; Penet, M. F.; Jiang, L.; Jacobs, M. A.; Bhujwalla, Z. M. Choline Metabolism-Based Molecular
14 Diagnosis of Cancer: An Update. Expert Rev. Mol. Diagn. 2015, 15 (6), 735–747.
15 https://doi.org/10.1586/14737159.2015.1039515.
16 (53) Arroyo-Crespo, J. J.; Armiñán, A.; Charbonnier, D.; Deladriere, C.; Palomino-Schätzlein, M.; Lamas-
17 Domingo, R.; Forteza, J.; Pineda-Lucena, A.; Vicent, M. J. Characterization of Triple-Negative Breast
18 Cancer Preclinical Models Provides Functional Evidence of Metastatic Progression. Int. J. Cancer 2019,
19 145 (8), 2267–2281. https://doi.org/10.1002/ijc.32270.
20
(54) Kanchan Sonkar, Vinay Ayyappan, Caitlin M. Tressler, Oluwatobi Adelaja, Ruoquing Cai, Menglin
21
Cheng, K. G. Focus on the Glycerophosphocholine Pathway in Choline Phospholipid Metabolsim of
22
Cancer. NMR biomed. 2019, 32 (10), e4112. https://doi.org/10.1002/nbm.4112.Focus.
23
(55) Hay, N. Reprogramming Glucose Metabolism in Cancer: Can It Be Exploited for Cancer Therapy? Nat.
24
Rev. Cancer 2016, 16 (10), 635–649. https://doi.org/10.1038/nrc.2016.77.
25
(56) Kazak, L.; Cohen, P. Creatine Metabolism: Energy Homeostasis, Immunity and Cancer Biology. Nat.
26
Rev. Endocrinol. 2020, 16 (8), 421–436. https://doi.org/10.1038/s41574-020-0365-5.
27
(57) Lieu, E. L.; Nguyen, T.; Rhyne, S.; Kim, J. Amino Acids in Cancer. Exp. Mol. Med. 2020, 52 (1), 15–30.
28
https://doi.org/10.1038/s12276-020-0375-3.
29
30 (58) Altman, B. J.; Stine, Z. E.; Dang, C. V. From Krebs to Clinic: Glutamine Metabolism to Cancer Therapy.
31 Nat. Rev. Cancer 2016, 16 (10), 619–634. https://doi.org/10.1038/nrc.2016.71.
32 (59) Jang, W. J.; Choi, B.; Song, S. H.; Lee, N.; Kim, D. J.; Lee, S.; Jeong, C. H. Multi-Omics Analysis Reveals
33 That Ornithine Decarboxylase Contributes to Erlotinib Resistance in Pancreatic Cancer Cells.
34 Oncotarget 2017, 8 (54), 92727–92742. https://doi.org/10.18632/oncotarget.21572.
35
36
37
38 Acknowledgments
39
40 We thank D. Charbonnier for support regarding animal tissues, Dr. Stuart P. Atkinson for his
41
42 collaboration in manuscript preparation and English revision, and funding from AECC Valencia, the
43 Spanish Ministry of Science and Innovation (PID2019-108806RB-I00, SAF2017-89229-R), and the
44
45 FPU grant to IDO Ref: FPU19/03729. Part of the equipment employed in this work was funded by
46
Generalitat Valenciana and co-financed with FEDER funds (PO FEDER of Comunitat Valenciana
47
48 2014–2020).
49
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment
 Analytical Chemistry Page 26 of 26

1
2
3 For Table of Contents Only
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 ACS Paragon Plus Environment