Graphical abstract
graphic file with name fx1.jpg
[39]Open in a new tab
Highlights
* •
dSILO is a method to measure proteome-wide protein turnover rates
in an organoid system
* •
Turnover rates are overall higher in PDA metastatic organoids than
in tumor organoids
* •
Mitochondrial proteins have the highest turnover increase
* •
Turnover changes are not generally correlated with changes in
protein abundance
Motivation
We sought to profile protein abundance and turnover rate in organoids
derived from a pancreatic cancer mouse model using dynamic SILAC
labeling. By doing so in paired primary and metastatic tumor organoids
derived from an autochthonous PDA mouse model, we aim to gain deeper,
unbiased insights into potential mechanisms driving PDA metastasis.
__________________________________________________________________
Ross et al. present dynamic Stable-Isotope Labeling of Organoids
(dSILO), a dynamic SILAC approach to measure proteome-wide protein
turnover rates in organoids. They apply the method in a pancreatic
ductal adenocarcinoma mouse model and discover that the metastatic
organoids display accelerated proteome turnover compared to primary
tumor organoids.
Introduction
Pancreatic ductal adenocarcinoma (PDA) is a devastating disease with a
5-year survival rate of only 12%.[40]^1 This poor prognosis is partly
related to the highly metastatic potential of PDA,[41]^2^,[42]^3
compounded by a frequent late-stage diagnosis of the disease.[43]^4
Understanding the cellular processes associated with metastatic
transformation is paramount for developing effective treatment
strategies and addressing the main causes of PDA mortality. Through
genome sequencing efforts,[44]^5^,[45]^6 we now know that KRAS is the
major driver oncogene in PDA, observed in over 90% of PDA cases.[46]^7
Other common mutations include CDKN2A (p14(ARF)/p16(INK4A)),[47]^8
TP53,[48]^9^,[49]^10 and SMAD4/DPC4.[50]^11^,[51]^12 Mutations in TP53
are the most prevalent in human PDA patients and are frequently
associated with the loss of heterozygosity (LOH) of its wild-type
allele.[52]^13^,[53]^14^,[54]^15
Most PDA research has centered on
genomic[55]^7^,[56]^8^,[57]^9^,[58]^10^,[59]^11^,[60]^12 and
transcriptomic[61]^16^,[62]^17^,[63]^18^,[64]^19^,[65]^20 profiling of
primary tumors, which has provided fundamental insights into the
genetic and molecular causes of PDA. However, it remains critically
important to characterize the changes that drive metastatic lesions.
Interestingly, next-generation genome sequencing of patient-matched
primary pancreatic and metastatic lesions revealed striking genetic
similarities between primary tumors and metastases,[66]^21 underscoring
the crucial role of transcriptional and post-transcriptional changes in
promoting PDA metastasis.[67]^22^,[68]^23^,[69]^24^,[70]^25 Indeed,
global proteomic analysis of human PDA liver metastases revealed four
distinct subtypes,[71]^26 laying the groundwork for the potential
application of proteomic classification in developing strategies to
understand and target PDA metastasis.
The maintenance of protein homeostasis involves a constant balance
between protein synthesis and degradation.[72]^27 The disruption of
this balance is linked to various diseases, including
cancer[73]^28^,[74]^29^,[75]^30 and neurodegenerative disorders.[76]^31
A growing body of evidence suggests that further disruption of this
imbalance is a viable therapeutic strategy for
cancer.[77]^32^,[78]^33^,[79]^34^,[80]^35
Recent advancements in proteomics have equipped us with tools to
conduct comprehensive analyses of protein turnover in homeostatic
systems.[81]^27^,[82]^36 One such technique is “dynamic stable-isotope
labeling by amino acids in cell culture”
(dSILAC).[83]^37^,[84]^38^,[85]^39^,[86]^40^,[87]^41^,[88]^42^,[89]^43^
,[90]^44^,[91]^45 In this approach, cells are pulse-labeled with heavy
stable-isotope amino acids, followed by mass spectrometric analysis.
This allows for the measurement of protein half-lives at a
proteome-wide level by comparing the rate of incorporation of heavy
amino acids into new peptides against the reduction of light
(unlabeled) amino acids. Combining dSILAC with isobaric labels such as
tandem mass tags (TMTs) further enables reduced measurement time and
sample-to-sample variability.[92]^43^,[93]^46
In this study, we introduce dynamic stable-isotope labeling of
organoids (dSILO), an application of hyperplexed dSILAC-TMT to ex vivo
organoid models. We apply dSILO to PDA to investigate if protein
turnover is differentially regulated in the metastatic setting. Our
findings suggest that the proteome of PDA metastases generally exhibits
shorter half-lives compared to primary tumors. Specifically, the
respiratory megacomplex I[2]-III[2]-IV[2],[94]^47 also known as the
“respirasome,”[95]^48 is differentially regulated in PDA metastatic
organoids. Given the plethora of recent studies reporting increased
mitochondrial activity in metastatic cancer
cells,[96]^44^,[97]^49^,[98]^50^,[99]^51^,[100]^52^,[101]^53^,[102]^54^
,[103]^55^,[104]^56 our findings suggest that dynamic renewal of the
respiratory chain components in PDA may represent a
post-transcriptional mechanism to support the metabolic needs of
metastases.
Results
Proteomic characterization of paired primary tumor- and metastases-derived
PDA organoids
Using tissues from the Kras^G12D;p53^R172H;PdxCre (KPC) mouse
model,[105]^14 we developed a panel of murine pancreatic ductal
organoids derived from primary pancreatic tumors and paired distant
metastases isolated from the liver or the diaphragm[106]^56
([107]Figures 1A–1C). Malignant PDA is frequently, but not always,
accompanied by a loss of heterozygosity (LOH) of
TP53.[108]^13^,[109]^14^,[110]^15 To separate out p53-driven effects
from the differences between metastatic and primary tumors, we treated
all cells with Nutlin-3a, a mouse double minute 2 (MDM2)
antagonist,[111]^57 for several passages ([112]Figure 1A). As expected,
Nutlin-3a selected for cells with Trp53 LOH ([113]Figure 1D). We
confirmed that Nutlin-3a did not significantly alter the composition of
the metastatic proteome ([114]Figure S1A). One out of five metastatic
organoid lines (M19) retained one wild-type allele prior to Nutlin-3a
treatment ([115]Figure 1D), indicating that TP53 LOH is a frequent, but
non-essential event of PDA metastasis, possibly due to loss of function
mutations in downstream effectors such as CDKN2A.[116]^58 We did not
find strong differences in the proteome of the M19 organoid line
compared to the other metastatic lines prior to Nutlin-3a treatment
([117]Figure S1B).
Figure 1.
[118]Figure 1
[119]Open in a new tab
Biological model and experimental design
(A) Workflow for preparation of organoid lines for dSILO and global
proteomics. Paired organoid cultures were established from the
Kras^G12D; p53^R172H; PdxCre (KPC) mouse model of pancreatic cancer,
comprising primary tumors (n = 4), diaphragm (n = 2), and liver
metastases (n = 3). Organoid cultures were treated with Nutlin-3a for
selection of cells with Trp53 LOH. The resulting panel of paired tumors
(n = 4) and metastases (n = 5) was used for dSILO (protein turnover)
and global proteomics (protein abundance) analyses.
(B) Representative hematoxylin and eosin (H&E) staining of parental
tissues (primary tumors and metastases, “Met”) used for the
establishment of organoid lines. Bright-field (BF) pictures of tumor
and metastasis organoids used for treatment with Nutlin-3a. Scale bars:
100 μm for H&E pictures and 2 mm for BF pictures.
(C) Overview of primary and metastatic organoid pairs from each mouse
in this study.
(D) PCR-based genotyping of primary tumor (“T”) and metastasis (“M”)
organoids, before and after Nutlin-3a treatment. Upon treatment with
Nutlin-3a, tumor organoids with the loss of their wild-type (WT) allele
copy of Trp53 were enriched.
We looked at the total proteome changes among the 6,280 proteins
identified in both primary tumor and metastatic organoid lines
([120]Table S1). 167 proteins exhibit a significantly lower expression
level (p < 0.05 and log[2]FC < −0.58, 1.5-fold decrease) in metastatic
tumors, while 48 proteins exhibit a significantly higher expression
level (p < 0.05 and log[2]FC > 0.58, 1.5-fold increase)
([121]Figure S1C).
dSILO measurements show faster proteome turnover in metastases than primary
tumors
To determine if the proteome is differentially turned over in
metastatic PDA cells, we used dSILO to label five pairs of primary
tumor and metastatic organoids in culture ([122]Figure 2A). Cells were
pulse-labeled over seven time points. Extracted lysates were
subsequently TMT labeled, HPLC fractionated, and measured by liquid
chromatography coupled to tandem mass spectrometry (LC-MS/MS).
Quantifying TMT-labeled samples at the MS2 (tandem mass spectrometry)
level can lead to signal interference from peptides co-isolated and
co-fragmented along with the target peptide.[123]^59 We addressed this
with deep sample fractionation ([124]Figure S1D), which reduces sample
complexity, thereby enhancing accuracy and proteome coverage of
samples.[125]^60 TMT mixes were designed such that each primary tumor
and its corresponding metastatic tumor(s) were in the same plex, along
with a highly heavy-labeled “booster” sample, which was included to
increase the coverage of heavy-labeled peptides picked for MS2 analysis
([126]Figures S1E–S1K and [127]Table S2).[128]^39
Figure 2.
[129]Figure 2
[130]Open in a new tab
Experimental workflow and validation of method to determine protein
half-lives in PDA organoids
(A) Experimental workflow for pulsing and quantification of protein
half-lives in PDA organoids.
(B–F) Heatmaps for protein heavy (H)/light (L) ratios (ln(H/L + 1))
across all biological replicates over the course of 36 h.
(G) Coefficients of determination (R^2) of the fit of the linear
heavy/light label incorporation (in a semi-log plot; see [131]STAR
Methods for details) for each tumor and metastasis organoid line.
Percentages represent the number of proteins with half-lives determined
and R^2 >0.7. Data are shown as mean ± SD.
We identified a total of 49,976 unique peptides, mapping to 6,103
unique protein groups ([132]Table S2). As expected with constant
incorporation rates, heavy/light (H/L) ratios increased with time
([133]Figures 2B–2F). We used a previously described model to calculate
protein half-lives[134]^42 and extracted the R^2 values of the linear
fit of heavy label incorporation as a filter for quality control,
removing any proteins with an R^2 value <0.7. Over 90% of the
identified proteins have an R^2 >0.7, suggesting high quality of the
fit ([135]Figure 2G). Of the 6,103 total proteins identified, 3,924
proteins pass the cutoff criteria of peptide spectrum matches
(PSMs) > 2 and R^2 > 0.7 and were used for subsequent analyses
([136]Figure S2A). We observed a good agreement of half-lives between
biological replicates in primary ([137]Figure S2B) and metastatic
samples ([138]Figure S2C). Consistent with previously published
datasets,[139]^45^,[140]^61 we found that mitochondrial proteins are
the most long-lived proteins in both primary and metastatic tumors,
while cell surface proteins (including cell membrane, cell junction,
and cell projection) are among the most short-lived proteins
([141]Figures S2D and S2E).
Cell doubling rate plays a critical role in determining the degradation
rate of proteins in dividing cells, especially those that turn over
slowly.[142]^27^,[143]^62^,[144]^63 We monitored the growth of
organoids during heavy amino acid labeling and measured the change in
organoid circumference and area over time using microscopy
([145]Figure S3A). Although there is no significant difference in the
proliferation between tumor and metastasis organoids as a group,
metastatic organoids tend to double more slowly, with distinct
differences across each line ([146]Figures S3B–S3D). Thus, to account
for variations across each line and ensure accurate measurement of
protein degradation rates, we calculated the average degradation rates
of the 20 most long-lived proteins in each organoid line and used these
values as cell doubling correction factors ([147]Table S2).
We compared the median protein half-lives between primary and
metastatic organoids in a pairwise manner and observed that the median
half-life of the metastatic proteome is significantly lower than the
tumor proteome for every pair measured (p < 0.0001, paired Wilcoxon
test) ([148]Figures 3A–3E). To identify protein-specific changes, we
focused on the median half-life of 2,349 proteins identified in at
least three organoid pairs. In accordance with our global turnover
analysis, the mean protein half-life of this subset of the proteome is
10.9 h lower in the metastatic organoids than their primary
counterparts ([149]Figure S2F). Together, these data suggest that PDA
metastases have a faster proteome turnover rate than primary tumors.
However, the higher turnover is not reflected in protein abundance, as
there is a low correlation between these two parameters
([150]Figures S3E–S3J).
Figure 3.
[151]Figure 3
[152]Open in a new tab
Mitochondrial proteins turn over faster in metastases compared to
tumors
(A–E) Histograms of protein half-life frequency distributions for each
metastasis vs. tumor pair. The “n” value in each graph corresponds to
the number of proteins shared between each tumor and metastases pair.
Differences between half-life distributions were assessed using the
Wilcoxon test.
(F) Volcano plot of differentially turned over proteins in metastases
compared to primary tumors. Blue dots indicate proteins with
significantly decreased half-lives in metastasis compared to tumor (n =
486, p < 0.05, log[2]FC < −0.32). Vertical red lines represent the
bottom (left, log[2]FC = −0.32) and top (right, log[2]FC = 0.32)
log[2]FC cutoffs applied. The horizontal dotted line represents the p
value cutoff (p = 0.05). p value established by paired t test.
(G) KEGG pathway enrichment analysis from DAVID showing pathways with
shortened half-lives (i.e., faster turnover) in metastases compared to
tumors. The x axis represents the −log[10](p value). Only pathways with
p <0.05 are shown. “n” represents the number of proteins associated
with each pathway.
(H) Keyword (KW) cellular compartment analysis from DAVID showing
cellular compartments with shortened half-lives (i.e., faster turnover)
in metastases compared to tumors. The x axis represents the −log[10](p
value). Only cellular compartments with p <0.05 are shown. “n”
represents the number of proteins associated with each cellular
compartment. Mitochondrial and inner mitochondrial membrane
compartments are marked in green.
Mitochondrial proteins show faster turnover in metastases compared to primary
tumors
Of the 2,349 proteins commonly identified in at least 3 pairs of
organoids, 486 proteins exhibit a significant 1.25-fold decrease
(p < 0.05) in half-life in the metastatic setting ([153]Figure 3F). No
protein showed a significant increase in half-life in metastatic
organoids, consistent with the general shift to shorter half-lives in
metastases relative to tumors. The subset of proteins with
significantly shorter half-lives is enriched for participation in
metabolic pathways in the mitochondria ([154]Figures 3G and 3H,
[155]Table S3), such as the tricarboxylic acid cycle and other
metabolic pathways involved in carbon and amino acid metabolism
([156]Figure 3G). Through cellular compartment analysis, we found an
enrichment of mitochondrial and mitochondrial inner membrane proteins
([157]Figure 3H). We tested the ratio of mitochondrial DNA (mtDNA)
content relative to nuclear DNA (nDNA) in tumor and metastasis
organoids and found no difference between the two cell types
([158]Figure S4A). Therefore, the increased turnover of mitochondrial
proteins in the metastatic setting is not due to a difference in
mitochondrial numbers.
The mitochondrial respirasome megacomplex turns over faster in metastases
than in primary tumors
Previous studies suggest that proteins with the capacity to assemble
into complexes may exhibit longer half-lives compared to unassembled
proteins.[159]^45^,[160]^64^,[161]^65^,[162]^66^,[163]^67 Other studies
suggest that the half-lives of complexed proteins show higher coherence
than expected by randomly assigning proteins to groups of the same
size.[164]^45^,[165]^68 We examined our turnover data using the
mammalian protein complexes database (CORUM)[166]^69 and found that
while proteins in complexes do not have significantly different
half-lives than proteins without known protein binding partners
([167]Figure 4A, Kolmogorov-Smirnov test, p > 0.05), proteins involved
in known complexes do exhibit lower variance in half-life compared to
randomized protein pairs ([168]Figure 4B, Kolmogorov-Smirnov test,
p < 0.05). This suggests a coherence in the turnover rates of proteins
involved in complexes. To identify dysregulated complexes that may be
contributing to metastatic dissemination, we investigated the
differences in half-lives between primary tumors and metastases in the
context of their associated complexes. We examined 53 protein complexes
comprising >3 proteins, representing 407 total proteins, and identified
44 protein complexes with significant differences in overall turnover
rates ([169]Table S4). After condensing redundant complex assignments
with minimal unique protein compositions, we noticed that 3 of the 10
complexes with the greatest differences between metastatic and primary
tumors were components of the respirasome, the active form of
mitochondrial complexes I[2]-III[2]-IV[2][170]^48 ([171]Figure 4C;
[172]Table S4). Overall, proteins from the respirasome have shorter
half-lives in metastatic tumors compared to primary tumors
([173]Figure 4D). We identified 20 proteins of the mitochondrial
respiratory complex exhibiting shorter half-lives in metastases
compared to primary tumors (p < 0.05 and |log[2]FC| > 0.32)
([174]Figure 4E). While not statistically significant, there are
similar downward trends in the abundance of these proteins.
Figure 4.
[175]Figure 4
[176]Open in a new tab
Mitochondrial respiratory chain complexes turn over faster in
metastases than in primary tumors
(A) Half-life cumulative distributions of proteins found to be in
complexes and those without protein binding partners per CORUM.
Proteins found within complexes did not display significant differences
in half-lives compared with randomly grouped (shuffled) proteins within
the dataset (Kolmogorov-Smirnov test, p > 0.05).
(B) Variance cumulative distribution of half-lives of proteins found to
be in complexes per CORUM compared to randomly shuffled groups of
proteins from the dataset. Proteins found within complexes show less
variance in their half-lives than randomly grouped (shuffled) proteins
within the dataset (Kolmogorov-Smirnov test, p < 0.05).
(C) Mean log[2] fold-change differences in half-lives of the top five
protein complexes with significantly different turnover rates in
metastases compared to primary tumors after curating for redundant
complex assignments (Kolmogorov-Smirnov test, p < 0.001).
(D) Half-life comparison of differentially turned over respirasome
proteins in tumors and metastases (two-sided paired t test, p < 0.05).
(E) Log[2] fold changes (FCs) of half-lives and protein abundances of
overlapping proteins from the datasets of respiratory chain complexes
I–V. Significance was assessed using a two-sided paired t test
(p < 0.05). Data are shown as mean ± SD.
Previous reports have shown sub-architectural differences in the
regulation of subunits of the mitochondrial respiratory chain (RC)
complex I[177]^65^,[178]^70^,[179]^71 and other complexes.[180]^45 To
determine whether these differences are present in our model, we mapped
the results from differential turnover analysis of the subunits of
these complexes between metastatic and primary tumors onto the cryo-EM
structure of the assembled respirasome (PDBID: [181]5XTI [182]^47)
([183]Figure S4B). We did not observe any clear trend in turnover
rates associated with differences in the architecture of the
respirasome.
Stability of mitochondrial proteins in light of previously published datasets
In this study, we found that metastatic cells showed faster overall
proteome turnover than primary tumors ([184]Figures 3A–3E and
[185]S2F), even though the metastatic organoids grew at similar or
slightly slower rates ([186]Figures S3A–S3D). Specifically, we found
that mitochondrial proteins were among the most long-lived proteins in
both the primary tumor ([187]Figure S2D) and metastasis
([188]Figure S2E). However, mitochondrial respiratory chain proteins
were found to turn over significantly faster in metastases relative to
primary tumors ([189]Figures 4C–4E). To understand whether this
observation was consistent in other biological systems, we compared our
results to various published datasets.
A previous study by Welle and colleagues compared the turnover rates of
proliferative and quiescent human dermal fibroblasts.[190]^46 Their
data indicate that quiescent cells have a faster turnover rate (higher
k[deg]) than proliferating cells, both globally ([191]Figure S4C) and
specifically for mitochondrial respiratory chain proteins
([192]Figure S4E). This is in line with our observation
([193]Figure S4D) and reinforces the notion that the rate of cell
division and the rate of protein degradation are not necessarily
positively correlated.
Next, we compared our dataset to a recent study by Dong and
colleagues,[194]^72 who applied a pulse of isotopically labeled leucine
to study the turnover of midbrain organoids derived from human induced
pluripotent stem cells. Comparing the 312 proteins common to both
datasets, we observed that the median half-lives of brain organoids are
higher than those of PDA organoids ([195]Figure S4F). Only one protein
from the mitochondrial respiratory chain is common to both datasets,
SDHA, with a half-life of 139.6 h in the midbrain organoids, much
higher than in the tumor (114.8 h) and metastatic (78.2 h) PDA
organoids. In line with our results, the authors also identify
mitochondrial proteins as having the longest half-lives.
Finally, we compared our proteome turnover results to healthy liver
tissue measurements from Rolfs et al.[196]^73 We analyzed 841 proteins
that are present in both datasets and compared their half-lives. Our
analysis showed that the half-lives of liver tissue proteins are
generally longer than that of metastatic organoids but shorter than
that of primary tumor PDA organoids ([197]Figures S4G and S4H). Both
studies concurred that mitochondrial respiratory chain proteins are
among the most stable in the proteome ([198]Figure S4I).
Our findings suggest that the link between turnover and proliferation
as well as the stability of mitochondrial proteins align with previous
research. We hypothesize that a shared mechanism may account for the
slow turnover of mitochondrial proteins in all three reference datasets
and our primary tumors, which could potentially be disrupted in
metastatic cancer.
Discussion
We developed dSILO to perform a proteome-wide characterization of
protein turnover differences between pancreatic primary tumor and
metastasis-derived organoids and show that steady-state proteome
turnover occurs more rapidly in metastatic organoids than in primary
counterparts.
Although some studies[199]^61^,[200]^65 suggested a connection between
protein abundance and rate of turnover, our research has found that
these two parameters are not always coupled in PDA cells. This is also
true in the brain, where protein levels for genes associated with
age-related neurodegenerative diseases remain constant, but their
turnover rates change dynamically.[201]^74 Therefore, it is important
to measure both protein abundance and turnover to provide an accurate
understanding of how gene expression is regulated in different cell
types. The relationship between these two factors appears to be unique
to each context.
Elevated mitochondrial respiration in metastasis has been reported in
various
cancers.[202]^44^,[203]^49^,[204]^50^,[205]^51^,[206]^52^,[207]^53^,[20
8]^54^,[209]^55^,[210]^56 Our findings show that the turnover of
mitochondrial proteins is higher in metastatic PDA cells. This is
consistent with the increased respiratory needs of metastatic cancer.
The increased turnover of mitochondrial proteins could be a result of
higher levels of reactive oxygen species production in the mitochondria
of metastatic PDA cells, as has been previously reported.[211]^56
Indeed, the genetic depletion of the mitochondrial quality control
protein HSP60 (also known as HSPD1) impairs PDA cell proliferation and
migration.[212]^75 Although HSP60 is not identified in our turnover
analysis, other members of the quality control machinery, such as
LONP1, HSPA9, PITRM1, and PARK7, exhibit significantly faster turnover
in metastasis ([213]Table S3). Lastly, our findings suggest that the
regulation of the respirasome as a whole may vary in the context of
metastatic PDA, contrary to previous studies suggesting that there are
functional differences between subunits of RC complexes at the
sub-architectural level.[214]^65
Despite organoid models being more costly and time-consuming compared
to monolayer cultures, organoid cultures are a valuable resource as
they maintain key features of the tissue of origin, including cell
polarity and intra-tumor heterogeneity.[215]^2 We anticipate that our
method of measuring proteome turnover using dSILO will be useful in
numerous comparative organoid studies, to determine how changes in
proteome homeostasis contribute to developmental and pathological
processes. Our research also opens up the possibility of using dSILO to
investigate proteome turnover differences in personalized medicine
development using patient-derived organoids.
Limitations of the study
To determine which protein complexes have the greatest differences in
turnover between conditions, we sorted proteins into complexes using
the CORUM database as a reference. However, it is important to note
that this analysis may not be completely exhaustive. These databases
are biased toward protein complexes that have been investigated and
published more frequently. This means that there could be complexes
that would invalidate the trends we observed. Nonetheless, these
databases are among the few tools we have to assess protein complex
dynamics. Furthermore, although we did not detect any clear trend in
turnover rates linked to differences in the architecture of the
respirasome, this conclusion is based on limited observations since
only a few membrane-bound RC proteins were identified in our dataset.
STAR★Methods
Key resources table
REAGENT or RESOURCE SOURCE IDENTIFIER
Chemicals, peptides, and recombinant proteins
__________________________________________________________________
L-Lysine-2HCL Thermo Scientific Cat# 89987
L-Arginine-HCL Thermo Scientific Cat# 89989
L-Lysine-2HCL, 13C6, 15N2 Thermo Scientific Cat# 88209
L-Arginine-HCL, 13C6, 15N4 Thermo Scientific Cat# 89989
Nutlin-3a Sigma-Aldrich Cat# SML0580-5MG
Dithiothreitol Fisher Cat# BP172-5
Iodoacetamide Sigma-Aldrich Cat# I6125-25G
Sequencing Grade Modified Trypsin Promega Cat# V511X
Formic acid Fisher Cat# A117-50
TMT16plex mass-tag labeling reagent Thermo Scientific Cat# [216]A44520
SYBR green Thermo Scientific Cat# 4367659
Advanced DMEM/F12 Gibco Cat# 12634010
Penicillin/Streptomycin Gibco Cat# 15140163
GlutaMAX Gibco Cat# 35050061
HEPES Gibco Cat# 15630080
Gem21 neuroplex Gemini Bio Cat# 400-160
N-Acetylcysteine Sigma-Aldrich Cat# A9165
Gastrin Sigma-Aldrich Cat# G9145
EGF PeproTech Cat# 315-09
FGF10 PeproTec Cat# 100-26
Nicotinamide Sigma-Aldrich Cat# N0636
Growth Factor Reduced (GFR)-Matrigel Corning Cat# 356231
SILAC media Thermo Fisher Scientific Cat# A2494301
__________________________________________________________________
Critical commercial assays
__________________________________________________________________
Quick-DNA™ Miniprep Plus Kit Zymo Research Cat# D4068
BCA kit Pierce Cat# PI23227
__________________________________________________________________
Deposited data
__________________________________________________________________
Protein turnover data of proliferative and quiescent human dermal
fibroblasts Welle et al.[217]^46 ProteomeXchange Consortium identifier:
[218]PXD004725
Protein turnover data of human induced pluripotent stem cells
(iPSC)-derived midbrain organoids Dong et al.[219]^72 ProteomeXchange
Consortium identifier: [220]PXD032169
Protein turnover data of healthy mouse liver tissue Rolfs
et al.[221]^73 MassIVE identifier: MSV000086426
Proteome expression and turnover datasets This study MassIVE ID number:
MSV000092498
__________________________________________________________________
Experimental models: Cell lines
__________________________________________________________________
Mouse: KPC derived PDAC organoids He et al.[222]^56 N/A
__________________________________________________________________
Oligonucleotides
__________________________________________________________________
Trp53 loxP Fw: AGCCTGCCTAGCTTCCTCAGG He et al.[223]^56 N/A
Trp53 loxP Rv: CTTGGAGACATAGCCACACTG He et al.[224]^56 N/A
CO1 Fw: TGCTAGCCGCAGGCATTAC Jovanovic et al.[225]^80 N/A
CO1 Rv: GGGTGCCCAAAGAATCAGAAC Jovanovic et al.[226]^80 N/A
NDUFV1 Fw: CTTCCCCACTGGCCTCAAG Jovanovic et al.[227]^80 N/A
NDUFV1 Rv: CCAAAACCCAGTGATCCAGC Jovanovic et al.[228]^80 N/A
__________________________________________________________________
Software and algorithms
__________________________________________________________________
Original code for data analysis This study Figshare DOI:
[229]https://doi.org/10.6084/m9.figshare.25245997.v1
ImageJ National Institutes of Health [230]https://imagej.nih.gov/ij/;
RRID: [231]SCR_003070
GraphPad Prism GraphPad [232]http://www.graphpad.com/; RRID:
[233]SCR_002798
R studio Posit [234]https://posit.co/download/rstudio-desktop/
Spectronaut Biognosys [235]https://biognosys.com/shop/spectronaut
Xcalibur Thermo Fisher Scientific
[236]https://chemistry.unt.edu/∼verbeck/LIMS/Manuals/XCAL_Quant.pd;
RRID: [237]SCR_014593
MaxQuant Max Planck Institute of Biochemistry
[238]https://maxquant.org/; RRID: [239]SCR_014485
__________________________________________________________________
Other
__________________________________________________________________
25 cm sub-1.6-μm Aurora C18 column IonOpticks Cat# AUR2-25075C18A
25 cm sub-2-μm Aurora C18 column IonOpticks Cat# AUR2-25075C18A
Axygen Gel Documentation System Corning Product number GDBL-1000
[240]Open in a new tab
Resource availability
Lead contact
Further information and requests for resources and reagents should be
directed to and will be fulfilled by the lead contact, Marina Ayres
Pereira (marinaayrespereira@gmail.com).
Materials availability
All organoid lines can be obtained via a CUIMC materials transfer
agreement (free of charge for non-commercial purposes).
Data and code availability
* •
The proteome expression and turnover datasets that support the
findings on this study was deposited to MassIVE under the ID number
MSV000092498. Data are publicly available as of the date of
publication. Accession numbers are listed in the [241]key resources
table. This paper analyzes existing, publicly available data. The
accession numbers for these datasets are listed in the [242]key
resources table.
* •
All original code for data analysis was deposited on Figshare (DOI:
[243]https://doi.org/10.6084/m9.figshare.25245997.v1) and to the
Jovanovic Lab github account:
[244]https://github.com/mjlab-Columbia/2023_dSILO. Code is publicly
available as of the date of publication. DOIs are listed in the
[245]key resources table.
* •
Any additional information required to reanalyze the data reported
in this work is available from the [246]lead contact upon request.
Experimental model and study participant details
Organoid isolation and culture
Detailed procedures to isolate and propagate murine primary and
metastatic pancreatic organoids have been described
previously.[247]^56^,[248]^76 Briefly, organoids were maintained in
complete organoid media: Advanced DMEM/F12 (Gibco, Cat# 12634010)
supplemented with 1% Penicillin/Streptomycin (PS) (Gibco, Cat#
15140163), 1x GlutaMAX (Gibco, Cat# 35050061), 1x HEPES (Gibco, Cat#
15630080), Gem21 neuroplex (Gemini Bio, Cat# 400-160), 1.25 mM
N-Acetylcysteine (NAC) (Sigma-Aldrich, Cat# A9165), 10 nM gastrin
(Sigma-Aldrich, Cat# G9145), 50 ng mL-1 EGF (PeproTech, Cat# 315-09),
10% RSPO1-conditioned media, 20% Noggin-FC-conditioned media (the
Noggin-Fc-expressing cell line was a kind gift from Dr. Gijs R. van den
Brink, University of Amsterdam), 100 ng mL-1 FGF10 (PeproTech, Cat#
100-26), and 10 mM Nicotinamide (Sigma-Aldrich, Cat# N0636). To
passage, organoids were washed out from the Growth Factor Reduced
(GFR)-Matrigel (Corning, Cat# 356231) using ice-cold PBS, mechanically
dissociated into small fragments using fire-polished glass pipettes,
and then seeded into fresh GFR-Matrigel. Passaging was performed at a
1:4 split ratio roughly twice per week. All experiments described were
done without EGF and NAC.[249]^24 To isolate p53 LOH (loss of
heterozygosity) organoids, early-passage tumor (Pdx1-Cre;
Kras^+/LSL−G12D; Trp53^+/LSL−R172H) organoids (as previously
described[250]^56) were cultured in complete organoid media with 10 mM
Nutlin-3a (Sigma-Aldrich, Cat# SML0580-5MG) and propagated for at least
three passages (or until confirmation of p53 LOH by PCR). After
verification of p53 LOH, organoid lines were allowed to recover for
three passages before being seeded for experiments. All cells were
cultured at 37°C with 5% CO[2].
Method details
Genotyping
Organoids were harvested from three wells of a 24-well plate and
centrifuged at 1500 xg for 5 min at 4°C. Organoids were washed three
times with ice-cold PBS, after which genomic DNA was extracted using
the Quick-DNA Miniprep Plus Kit (Zymo Research, Cat# D4068). Each PCR
reaction for Trp53 1loxP genotyping was performed in a 20 μL mixture
containing 1x GoTaq G2 Hot Start master mix, 0.5 μM each primer, and
100ng template DNA. p53 LOH was confirmed by PCR[251]^14 using the
following primers.
* •
Trp53 loxP Fw: 5′-AGCCTGCCTAGCTTCCTCAGG-3’
* •
Trp53 loxP Rv: 5′-CTTGGAGACATAGCCACACTG-3′
The PCR cycling conditions were 95°C for 2 min, followed by 34 cycles
at 95°C for 30 s, 62°C for 30 s, and 72°C for 15 s, with a final
extension step at 72°C for 5 min. PCR products were separated on a 2%
agarose gel in 1X TAE buffer, and gel imaging was performed using an
Axygen Gel Documentation System.
Dynamic stable isotope labeling in organoids (dSILO)
Murine PDA primary and metastatic organoids were grown in EGF and
NAC-free media to unmask mitogenic and redox-dependent
mechanisms.[252]^24 When organoids reached the desired confluency, the
media was changed to SILAC media (Thermo Fisher Scientific, Cat#
A2494301) containing either light (0.50 mM L-Lysine-2HCL, Thermo
Scientific, Cat# 89987 and 0.70 mM L-Arginine-HCL, Thermo Scientific,
Cat# 89989) or heavy-labeled (0.50 mM L-Lysine-2HCL, 13C6, 15N2, Thermo
Scientific, Cat# 88209 and 0.70 mM L-Arginine-HCL, 13C6, 15N4, Thermo
Scientific, Cat# 89989) amino acids. The remaining growth factor
composition of the SILAC media is identical to the organoid culture
media described above. Cells were treated with heavy-labeled media for
0, 3, 6, 12, 18, 24, and 36 h. All wells experience the same number of
media changes at each time point and were harvested simultaneously.
Multiple ice-cold PBS washes were performed to ensure Matrigel removal.
Snap-frozen cell pellets were stored at −80°C until they were
subsequently processed for protein extraction and mass spectrometric
analysis.
Organoids growth assessment
In order to assess the sizes of organoids in both tumor and metastasis
settings, we used UCB Vision Science’s "Hough Circle Transform" plugin
for Fiji (ImageJ). We identified all the organoids in the microscopy
images at multiple focus levels and measured their relative radius in
pixels. This is then used to calculate total area and circumference.
Sample processing and mass spectrometry measurements
Cells were lysed for 30 min in urea buffer (8 M urea; 75 mM NaCl, 50 mM
Tris HCl pH 8.0, 1 mM EDTA) with 1X Protease Inhibitor Cocktail (Sigma,
Cat# P8340). Lysates were centrifuged at 20,000 g for 10 min, and
protein concentrations of the clarified lysates were measured via BCA
assay (Pierce, Cat# PI23227). Protein disulfide bonds of the lysates
were reduced for 45 min with 5 mM dithiothreitol (Fisher, Cat# BP172-5)
and alkylated for 45 min with 10 mM iodoacetamide (Sigma, Cat#
I6125-25G). Samples were then diluted 1:6 with 50 mM Tris HCl, pH 8.0,
to reduce the urea concentration to <2 M. Lysates were digested
overnight at room temperature with trypsin in a 1:50
enzyme-to-substrate ratio (Promega, Cat# V511X) on a shaker. Peptide
mixtures were acidified to a final volumetric concentration of 1%
formic acid (Fisher, Cat# A117-50). Tryptic peptides were desalted on
C18 StageTips as previously described,[253]^77 and evaporated to
dryness in a vacuum concentrator.
For global expression data, approximately 1 μg of total peptides were
analyzed on a Waters M-Class UPLC using a 25 cm sub-1.6-μm Aurora C18
column (IonOpticks Cat# AUR2-25075C18A) coupled to a benchtop Thermo
Fisher Scientific Orbitrap Q Exactive HF mass spectrometer. Peptides
were separated at a flow rate of 400 nL/min with a 160 min gradient,
including sample loading and column equilibration times. Data was
acquired in data-independent mode (DIA) using Xcalibur 4.5 software.
MS1 Spectra were measured with a resolution of 120,000, an AGC target
of 5e6 and a mass range from 350 to 1650 m/z. Per MS1, 38 equally
distanced, sequential segments were triggered at a resolution of
30,000, an AGC target of 3e6, a segment width of 36 m/z, and a fixed
first mass of 200 m/z. The stepped collision energies were set to 22.5,
25, and 27.
For proteome turnover (dSILO) analysis, desalted peptides were labeled
with the TMT16plex mass tag labeling reagent according to the
manufacturer’s instructions (Thermo Scientific, Cat# [254]A44520) with
small modifications. Briefly, 0.2 units of the TMT16plex reagent was
used per 20 μg of sample. Peptides were dissolved in 30 μL of 50 mM
HEPES pH 8.5 solution, and the TMT16 plex reagent was added in 12.3 μL
of MeCN. After 1h incubation, the reaction was quenched with 2.4 μL 5%
Hydroxylamine for 15 min at 25°C. Differentially labeled peptides were
mixed for each replicate. To reduce peptide complexity and achieve
deeper proteome coverage, samples were then separated by basic
reversed-phase chromatography as described previously,[255]^60 with 15
final concatenated fractions ([256]Figure S2A). Approximately 1 μg of
total peptides per sample were then analyzed on a Thermo Scientific
Orbitrap Q Exactive HF mass spectrometer coupled via a 25 cm sub-2-μm
Aurora C18 column (IonOpticks, Cat# AUR2-25075C18A) to an Acquity M
Class UPLC system (Waters). Peptides were separated at a flow rate of
400 nL/min with a linear 95 min gradient from 2% to 22% solvent B (100%
acetonitrile, 0.1% formic acid), followed by a linear 30 min gradient
from 22 to 90% solvent B. Each sample was run for 160 min, including
sample loading and column equilibration times. Data was acquired in
data-dependent (DDA) mode using Xcalibur 4.1 software. MS1 Spectra were
measured with a resolution of 120,000, an AGC target of 3e6, and a mass
range from 300 to 1800 m/z. Up to 12 MS2 spectra per duty cycle were
triggered at a resolution of 60,000, an AGC target of 1e5, an isolation
window of 0.8 m/z, and a normalized collision energy of 28.
TMT16 plexes were designed so that every primary tumor and its
corresponding metastatic tumor were run in the same mix
([257]Figure S2B, [258]Table S1). A strongly labeled “booster” spike-in
channel was added to the first and last channels of each mix in order
to increase the chances of heavy peptides being picked for
quantification,[259]^39 and to increase the likelihood of the same
peptides picked for quantification across mixes ([260]Figures S2C–S2H).
One mouse included in the study had metastatic tumors from both the
peritoneum and liver and as such organoid samples from its primary
tumor were run in two separate TMT16 plexes ([261]Figure S2B,
[262]Table S2).
Identification and quantification of proteins
Global expression data were searched using Spectronaut version 17.2
using the default BSG settings and median normalization. dSILO data
were searched with MaxQuant software version 1.7.0 using a modified
mouse UniProt database (March 2013) with added entries for mutant
KRASG12D and p53R172H. MS/MS searches for the proteome datasets were
performed with the following parameters: TMT16 labels were added as
isobaric labels and quantified on MS2. Oxidation of methionine, protein
N-terminal acetylation, and heavy SILAC amino acids (Lys8 and Arg10)
were added as variable modifications. Carbamidomethylation was added as
a fixed modification. Trypsin/P was selected as the digestion enzyme. A
maximum of 3 labeled amino acids and 2 missed cleavages per peptide
were allowed. The mass tolerance for precursor ions was set to 20 ppm
for the first search (used for nonlinear mass recalibration) and 6 ppm
for the main search. Fragment ion mass tolerance was set to 20 p.p.m.
For identification, we applied a maximum FDR of 1% separately on the
protein and peptide levels. We required 2 or more unique/razor peptides
for protein identification and a ratio count of 2 or more for protein
quantification per replicate measurement.
Protein half-life determination
Calculation of half-lives was performed as previously
described.[263]^42 Briefly, starting with the evidence.txt file of the
MaxQuant output, PSM entries were adjusted to reflect their relative
fraction of corresponding MS1 values. Heavy-labeled PSMs were filtered
to only contain fully labeled peptides. Subsequently, heavy and light
peptides of PSMs with the same protein group IDs and charge states were
matched so that H/L ratios could be calculated. PSM-level H/L ratios
were then median-summarized into protein-level H/L ratios.
The rate constant of protein degradation (k[deg]) was obtained by
linear regression using the equation:
[MATH: kdeg=∑i=1mloge(rti<
/msub>+1)ti∑i=1mti2−<
mfrac>loge2tcd :MATH]
where m is the number of pulsing timepoints (t[i]) and r[ti] the ratio
of H/L amino acids for a specific protein at each timepoint, and t[cd]
is the estimation of cell doubling. Protein half-life (T[1/2]) then was
calculated as per Schwanhausser et al.[264]^42
[MATH: T12=loge2kdeg :MATH]
To estimate cell doubling times (t[cd]), proteins for each sample were
ranked according to their kdeg. The 20 proteins with the lowest kdeg
for each sample were selected, and the mean value was used to estimate
cell doubling.
Only proteins with at least 3 PSMs measured by MS were considered for
half-life quantification. As a quality test, the coefficient of
determination R^2 was calculated for each fit. Only proteins with
half-lives with a R^2 > 0.7 were considered. As the resolution
associated with very long half-life measurements is not compatible with
the short-time courses used in the present study, proteins with very
long half-lives (>200 h) were excluded from the subsequent comparative
analysis between tumors and metastasis. The matrigel contaminants Col4
and Laminin were filtered out.
KEGG pathway, UP_KW_Cellular compartment term enrichment, and MitoPathway
analysis
KEGG pathway analysis and UP_KW_Cellular compartment analysis were
performed using the DAVID Bioinformatics Database (DAVID Bioinformatics
Resources, [265]https://david.ncifcrf.gov/).[266]^78 Calculation of
over-represented KEGG pathways and UP_KW_Cellular compartment was done
using the entire list of proteins with half-lives calculated as
background (threshold count = 2, and EASE score = 1). Pathways and GO
terms with p < 0.05 were selected.
MitoCarta3.0 was used for annotation of mitochondrial proteins.[267]^79
Assessment of mitochondrial pathways was performed using information
retrieved from MitoCarta3.0_MitoPathways.
Analysis of half-life distribution within protein complexes
Proteins identified in our dSILO dataset were cross-referenced with
complexes annotated in the CORUM database.[268]^69 Consequently, a
single protein could potentially be associated with multiple distinct
protein complexes. To examine differences in half-life distributions
between complexed and uncomplexed proteins, we compared those proteins
found to be involved in one or more complexes per CORUM (n = 841) to
those proteins not found to have any binding partners in CORUM (n =
1508), and applied Kolmogorov-Smirnov tests to assess significant
differences in the half-life distributions between the two groups. To
assess the overall coherence of half-lives of proteins involved in
complexes compared to those proteins not associated in complexes, we
randomly shuffled any CORUM-annotated proteins found in the dataset
proteins into groups and compared their variance distribution to those
of proteins found in true complexes, then performed Kolmogorov-Smirnov
tests to assess significant differences between the two groups. To
identify those complexes with the greatest aggregate differences in
turnover rates between primary and metastatic tumors, we performed
Kolmogorov-Smirnov tests to assess significant differences in the
half-life distributions between each organoid line of the primary
tumors and metastases. Analysis was performed in Python 3.10.
Mitochondrial:Nuclear DNA Rrtio Qqantification
We quantified the mitochondrial DNA (mtDNA)/nuclear DNA (nDNA) ratio as
described previously[269]^80 with slight modifications. Total cellular
DNA was extracted using the Quick-DNA Miniprep Plus Kit (Zymo Research,
Cat# D4068). Primers against mitochondrial-encoded NDUFV1 and
genomically-encoded CO1 were used to quantify mtDNA and nDNA,
respectively. The following primer sequences were used.
* •
CO1 Fw: 5′-TGCTAGCCGCAGGCATTAC-3’
* •
CO1 Rv: 5′-GGGTGCCCAAAGAATCAGAAC-3’
* •
NDUFV1 Fw: 5′-CTTCCCCACTGGCCTCAAG-3’
* •
NDUFV1 Rv: 5′-CCAAAACCCAGTGATCCAGC-3′
Quantitative PCR was performed using an SYBR green-based detection
reagent (Thermo Scientific Cat# 4367659) and an Applied Biosciences
StepOne Plus Real-Time PCR system.
Quantification and statistical analysis
Statistical analysis of global expression and turnover data
Statistical data analysis was performed in R Studio 2022.07.2 or Prism
9.0.
For protein expression data, comparisons between primary and metastatic
tumor organoids were performed on protein intensity levels, normalized
with Spectronaut’s default median normalization settings, using a
two-tailed paired t-test. Results are shown as mean ± SD. Proteins that
met our threshold parameters of |log2FC| ≥ 1.5 and p < 0.05 were
considered differentially expressed between conditions.
Half-live distributions between primary and metastatic tumors were
compared using Wilcoxon-rank testing to compare half-live distributions
generated as described in the section above. Correlations were computed
using Prism 9.0, and a linear regression model was fitted using Prism
statistics.
For complex analysis and for the identification of complexes with
differences in turnover rates between primary tumors and metastases, we
performed Wilcoxon-rank tests. A p < 0.05 was considered statistically
significant throughout all analyses.
Acknowledgments