Content uploaded by Dorota Skowronska-Krawczyk
Author content
All content in this area was uploaded by Dorota Skowronska-Krawczyk on Nov 17, 2024
Content may be subject to copyright.
Retinal Proteome Profiling of Inherited Retinal
Degeneration Across Three Different Mouse Models
Suggests Common Drug Targets in Retinitis
Pigmentosa
Authors
Ahmed B. Montaser, Fangyuan Gao, Danielle Peters, Katri Vainionpää, Ning Zhibin,
Dorota Skowronska-Krawczyk, Daniel Figeys, Krzysztof Palczewski, and Henri Leinonen
Correspondence Graphical Abstract
ahmed.montaser@uef.fi;henri.
leinonen@uef.fi
In Brief
Inherited retinal degenerations
(IRDs) lack effective treatment
options and they are a major
cause of blindness. In this study,
we analyzed the retinal
proteomes of three
translationally relevant IRD
mouse models (two for retinitis
pigmentosa and one for
congenital amaurosis type 2) to
identify convergent signaling
pathways using global
proteomics. Despite different
disease mechanisms between
the models, we found shared
pathways that could be targeted
for broad therapeutic
applications potentially by using
already existing drugs.
2024, Mol Cell Proteomics 23(11), 100855
© 2024 THE AUTHORS. Published by Elsevier Inc on behalf of American Society for Biochemistry and
Molecular Biology. This is an open access article under the CC BY license (http://creativecommons.org/
licenses/by/4.0/).
https://doi.org/10.1016/j.mcpro.2024.100855
RESEARCH
Highlights
•Proteomic analyses reveal retinal proteome data from three mouse models of degeneration.
•Retinitis pigmentosa and Leber congenital amaurosis type 2 models exhibit distinct proteomic
changes.
•Rd10 and P23H retinitis pigmentosa models display remarkable convergence in proteomic
phenotypes.
•Mutation-agnostic therapies are feasible due to shared pathological phenomena in retinitis
pigmentosa.
RESEARCH
Retinal Proteome Profiling of Inherited Retinal
Degeneration Across Three Different Mouse
Models Suggests Common Drug Targets in
Retinitis Pigmentosa
Ahmed B. Montaser
1,*
, Fangyuan Gao
2,3
, Danielle Peters
4
, Katri Vainionpää
1
,
Ning Zhibin
4
, Dorota Skowronska-Krawczyk
2,3
, Daniel Figeys
5
, Krzysztof Palczewski
2,3,6,7
,
and Henri Leinonen
1,*
Inherited retinal degenerations (IRDs) are a leading cause
of blindness among the population of young people in the
developed world. Approximately half of IRDs initially
manifest as gradual loss of night vision and visual fields,
characteristic of retinitis pigmentosa (RP). Due to chal-
lenges in genetic testing, and the large heterogeneity of
mutations underlying RP, targeted gene therapies are an
impractical largescale solution in the foreseeable future.
For this reason, identifying key pathophysiological path-
ways in IRDs that could be targets for mutation-agnostic
and disease-modifying therapies (DMTs) is warranted. In
this study, we investigated the retinal proteome of three
distinct IRD mouse models, in comparison to sex- and
age-matched wild-type mice. Specifically, we used the
Pde6β
Rd10
(rd10) and Rho
P23H/WT
(P23H) mouse models of
autosomal recessive and autosomal dominant RP,
respectively, as well as the Rpe65
−/−
mouse model of
Leber’s congenital amaurosis type 2 (LCA2). The mice
were housed at two distinct institutions and analyzed us-
ing LC-MS in three separate facilities/instruments
following data-dependent and data-independent acquisi-
tion modes. This cross-institutional and multi-
methodological approach signifies the reliability and
reproducibility of the results. The large-scale profiling of
the retinal proteome, coupled with in vivo electroretinog-
raphy recordings, provided us with a reliable basis for
comparing the disease phenotypes and severity. Despite
evident inflammation, cellular stress, and downscaled
phototransduction observed consistently across all three
models, the underlying pathologies of RP and LCA2 dis-
played many differences, sharing only four general KEGG
pathways. The opposite is true for the two RP models in
which we identify remarkable convergence in proteomic
phenotype even though the mechanism of primary rod
death in rd10 and P23H mice is different. Our data high-
lights the cAMP and cGMP second-messenger signaling
pathways as potential targets for therapeutic intervention.
The proteomic data is curated and made publicly avail-
able, facilitating the discovery of universal therapeutic
targets for RP.
Inherited retinal degenerative diseases (IRDs) are eye dis-
eases that lead to significant disability and eventually to
blindness, as they lack effective treatments. Collectively,
IRDs are relatively common disorders with a global preva-
lence of 1/2000 (1), and they form a leading cause of blind-
ness amongst the youth and working-age population in the
developed world (2). IRDs are excellent targets for gene
therapy as typically a single genetic mutation is the causative
factor (3). Furthermore, local delivery of therapeutic agents in
different compartments of the eye is common practice in eye
clinics. These facts facilitated the clinical approval of the first
gene therapy in 2017/2018 for the treatment of Leber's
congenital amaurosis type 2 (LCA2), a severe type of IRD.
This pioneering gene therapy, voretigene neparvovec (Lux-
turna), is a gene-augmentation therapy that supplies some
normal RPE65 protein, which is crucial for the renewal of
visual pigments in the retinal pigment epithelium (RPE) (4).
Since mutations in RPE65 account for only ~1 to 2% of IRD
cases (5,6) and all affected patients are not eligible for
Luxturna, a great majority of IRD patients are left without
treatment options.
A major challenge with IRDs is that they are very hetero-
geneous genetically. To date, mutations in more than 280
From the
1
School of Pharmacy, Faculty of Health Sciences, University of Eastern Finland, Kuopio, Finland;
2
Center for Translational Vision
Research, Department of Ophthalmology, Gavin Herbert Eye Institute, University of California, Irvine, Irvine, California, USA;
3
Department of
Physiology and Biophysics, University of California, Irvine, California, USA;
4
Ottawa Institute of Systems Biology, and
5
Department of
Biochemistry, Microbiology and Immunology, University of Ottawa, Ottawa, Ontario, Canada;
6
Department of Chemistry, and
7
Department of
Molecular Biology and Biochemistry, University of California, Irvine, California, USA
* For correspondence: Ahmed B. Montaser, ahmed.montaser@uef.fi; Henri Leinonen, henri.leinonen@uef.fi.
RESEARCH
Mol Cell Proteomics (2024) 23(11) 100855 1
© 2024 THE AUTHORS. Published by Elsevier Inc on behalf of American Society for Biochemistry and Molecular Biology.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1016/j.mcpro.2024.100855
genes are causative for IRDs, and the types of mutations in
causative genes can be numerous (7). For instance, at least
150 different disease-causing mutations have been identified
in the rhodopsin gene associated with retinitis pigmentosa
(RP) (8). Although the genetic knowledge of IRDs is expanding
constantly, 30% to 50% of IRDs remain idiopathic (9,10).
Because of these reasons, the applicability of targeted ther-
apies to address all IRDs appears unlikely in the foreseeable
future.
Even though the causative genetic mutations among IRDs
are highly diverse, the pathophysiological signaling events
occurring downstream of the primary insult (typically rod
degeneration) could be shared. Such convergent pathological
mechanisms, potentially driving collateral degeneration (11),
are attractive targets for disease-modifying therapies (DMTs)
that could benefit a broad and diverse patient population (12–
14). However, a comprehensive picture of the pathological
events across distinct IRD etiologies is still missing. Such data
is crucial for enabling the rational design of DMTs. For
instance, no cross-etiology analyses of the global proteome
exist, although technically this would have been possible for at
least a decade (15,16).
In this study, we aimed to find convergent downstream
retinal signaling events using global proteomics analyses from
the retinas of three distinct IRD model mice. Specifically, we
analyzed retinal samples from Pde6β
Rd10
and Rho
P23H/WT
(in
short, rd10 and P23H) mouse models of autosomal recessive
and autosomal dominant retinitis pigmentosa (RP), respec-
tively; and samples from the Rpe65
−/−
mouse model of LCA2.
As controls, we analyzed retinal samples from age- and sex-
matched wild-type (WT) mice corresponding to the respec-
tive IRD models. The mouse models in this work are highly
utilized disease models in translational IRD research and they
recapitulate the human diseases remarkably well (17). We
found that the great majority of pathological proteomic
changes were the same in rd10 and P23H RP-model mice,
whereas in Rpe65
−/−
LCA2-model mice numerous proteomic
network changes were distinct. We focused our attention on
the RP models and analyzed datasets not only by following
the traditional LC-MS/MS Data-Dependent Acquisition (DDA)
mode but also by using the Data-Independent Acquisition
(DIA) mode for improved sensitivity. This study is the first one
to analyze global proteomic data from several IRD models
simultaneously. Our analysis identifies numerous convergent
signaling pathways within the common RP models, facilitating
the rational design of DMT strategies.
EXPERIMENTAL PROCEDURES
Animal Models and Study Design
We used three different mouse models of inherited retinal degen-
erative diseases (IRD) to discover retinal proteomic changes that are
common to retinal degeneration (RD). The models used in this study
were B6.CXB1-Pde6b
rd10
/J (RRID: IMSR_JAX:004,297, referred to as
rd10), B6.129S6(Cg)-Rho
tm1.1Kpal
/J (RRID: IMSR_JAX:017,628,
referred to as P23H), mouse models of recessive and autosomal
dominant retinitis pigmentosa (RP), respectively (18,19), and B6.129-
Rpe65
tm1Tmr
/J (RRID:IMSR_JAX:035,329, referred to as Rpe65
−/−
)
model of Leber congenital amaurosis type 2 (LCA2) which was a kind
gift from Dr Michael Redmond (National Institutes of Health) (20). The
rd10 colony was kept as a homozygote. Age- and sex-matched
C57BL/6J mice (RRID: IMSR_JAX:000,664) were used as controls.
P23H heterozygote mice were bred with C57BL/6J mice yielding
P23H heterozygote and wild-type (WT) littermates. To get Rpe65
−/−
and their WT littermate mice, we bred heterozygote Rpe65
+/−
mice
together. Only WT and homozygote offsprings were used in this study.
Several cohorts of mice were raised, and their retinal samples were
collected, at two different institutions: the University of California Irvine
(UCI) and the University of Eastern Finland (UEF) over the years 2019
to 2021 and 2023, respectively (Table 1). Mice were given water and
standard feed ad libitum at both institutions. Retinal proteome analysis
was conducted using liquid chromatography-tandem mass spec-
trometry (LC-MS/MS), employing data-dependent acquisition (DDA) or
data-independent acquisition (DIA) modes.
The rd10 mice carry a naturally occurring point mutation in the
phosphodiesterase 6b (Pde6β) gene. This mutation leads to instability
and dysfunction of PDE6 protein and phototransduction, increased
free cGMP, subsequent opening of cGMP-gated channels, and
TABLE 1
Age and housing conditions of mouse cohorts used in the study
Condition DDA_rd10_DR DDA_rd10_CLR DDA_P23H DDA_Rpe65
−/−
DIA_rd10_DR DIA_P23H
Disease model RP RP RP LCA RP RP
Age P37 P38 P90 P45 P23–25 P60
Housing place UCI UCI UCI UCI UEF UEF
Rearing
condition
DR DR-to-CLR
a
Normal Normal DR Normal
LC-MS/MS UCI UO UCI UCI UEF UEF
MS instrument Orbitrap
Lumos
Orbitrap
QExactive
Orbitrap
Lumos
Orbitrap
Lumos
Orbitrap
QExactive
Orbitrap
QExactive
DDA, Data-dependent acquisition; DIA, Data-independent acquisition; DR, Dark rearing; CLR, Cyclic light rearing; RP, Retinitis pigmentosa;
LCA, Leber congenital amaurosis; UCI, University of California, Irvine; UEF, University of Eastern Finland; MS, Mass spectrometry; UO, Uni-
versity of Ottawa.
a
Mice reared P0-P29 in DR and in CLR (vivarium) P29-P38.
Retinal Proteome Profiling in Inherited Retinal Degeneration
Mol Cell Proteomics (2024) 23(11) 100855 2
increased Ca
2+
influx into rods, resulting in robust rod photoreceptor
degeneration (21). The rd10 mice are highly susceptible to the
damaging effects of light and most of their rods die by post-natal day
24 (P24) if mice are reared in cyclic light rearing (CLR), or vivarium
conditions (18). The disease progression is substantially slower if rd10
mice are reared in a dim-light environment, or in a dark room. Several
cohorts of rd10 mice and C57BL6/6J wild-type (WT) control mice were
housed in different light environments, and used for this study:
UCI_DDA_rd10_CLR_cohort 1: rd10 (N = 6; n = 3 females, n = 3
males) and WT (N = 6; n = 3 females, n = 3 males) were housed in
darkroom between P0 and P28. At P29, mice were transferred to
standard vivarium housing conditions (CLR = lights on 6:30 AM, and
off 6:30 PM) at the UCI laboratory animal center (LAC). Mice were
euthanized by cervical dislocation at P38 (9 days in vivarium), their
eyes were enucleated, and their retinas were harvested.
UCI_DDA_rd10_DR_cohort 2: rd10 (N = 6; n = 4 females, n = 2
males) and WT (N = 4; n = 2 females, n = 2 males) were housed in a
darkroom at UCI-LAC throughout their lifespan. The only light expo-
sure they experienced was the dim red light required for daily hus-
bandry. Mice were euthanized by cervical dislocation at P37 in a dark
room, their eyes were enucleated, and their retinas were harvested.
UEF_DIA_rd10_DR_cohort 3: rd10 (N = 4; n = 2 females, n = 2
males) and WT (N = 4; n = 2 females, n = 2 males) were housed in a
dim light environment at UEF-LAC throughout their lifespan. The mice
were kept in a Scantainer, in which glass doors were covered with a
darkening film, limiting light exposure to <0.01 lux inside the mouse
cages. The rd10 and WT mice were euthanized in the same session
with a 3 min exposure to CO
2
and subsequent cervical dislocation at
P23 and P25, respectively.
P23H mice harbor a proline to histidine mutation in codon 23 in the
rhodopsin gene, which leads to early rhodopsin misfolding, mis-
localization, and subsequent ER stress in their retinas (22). The mu-
tation leads to gain-of-function pathology, and Rho
P23H/WT
mice
undergo an intermediately progressing rod degeneration whereby
roughly half of their rods die by 3 months of age (23). Unlike with the
rd10 mice, the housing light conditions do not distinctly affect the
disease progression in Rho
P23H/WT
(P23H) mice, and all experiments in
P23H mice were performed in standard vivarium conditions. Two
cohorts of P23H mice and WT control mice were used:
UCI_DDA_P23H_cohort 1: P23H mice (N = 10; n = 5 females, n = 5
males) and WT littermates (N = 8; n = 4 females, n = 4 males) were
housed at UCI-LAC. All mice were euthanized by cervical dislocation
at ~ P90, their eyes were enucleated, and their retinas were harvested.
UEF_DIA_P23H_cohort 2: P23H mice (N = 10; n = 5 females, n = 5
males) and C57BL/6J WT control mice (N = 5; n = 2 females, n = 3
males) were housed at UEF-LAC. The P23H and WT mice were
euthanized by cervical dislocation at P60 and P57, respectively. Their
eyes were enucleated, and their retinas were harvested.
The Rpe65
−/−
mice are fully deficient of RPE65 which is an enzyme
that is necessary for the functioning of the classical visual cycle in the
retinal pigment epithelium (RPE) (24). In Rpe65
−/−
mice, the cone
photoreceptors do not respond to light at all (25). They also die quickly
after the opening of the eyes due to cone opsin mislocalization in the
absence of 11-cis-retinal production and supply by the RPE (26). In
contrast, rods die slowly in Rpe65
−/−
mice and they retain some re-
sidual light responsivity despite the severe chromophore insufficiency.
In our experiment, the Rpe65
−/−
mice were housed in standard vi-
varium conditions and one cohort was used:
UCI_DDA_Rpe65
−/−
cohort 1: Rpe65
−/−
mice (N = 8; n = 4 females,
n = 4 males) and C57BL/6J WT control mice (N = 9; n = 3 females, n =
6 males) were housed at UCI-LAC. Both retinas from two Rpe65
−/−
mice were pooled during harvesting leading to replicates: N = 4; n = 2
females, n = 2 males. Similarly, retinas from three WT mice were
pooled during harvesting leading to replicates: N = 3; n = 1 female, n =
2 males. All mice were euthanized by cervical dislocation at P45, their
eyes were enucleated, and their retinas were harvested.
At both institutions, mice were housed in a temperature-controlled
animal facility with a 12-h light/dark cycle and fed a standard rodent
diet ad libitum. All procedures were conducted in accordance with the
ARVO Statement for the Use of Animals in Ophthalmic and Vision
Research. Animal subjects at UCI were treated in accordance with the
NIH guidelines for the care and use of laboratory animals, and all
experimental procedures have been approved by the Institutional
Animal Care and Use Committee (IACUC, protocol #AUP-21-096).
Animal experiments at UEF were conducted in accordance with the
Directive 86/609/EEC for animal experiments, and FELASA Guidelines
and Recommendations, and were approved by the Finnish Project
Authorization Board, with protocol number ESAVI/26320/2021.
Retina Dissection
Retina dissection was performed as previously described (23).
Briefly, following euthanasia, the eyes were quickly enucleated, and
the retinas were excised by performing three incisions starting from
the optic nerve head and cutting toward the ora serrata, which allowed
easy and quick separation of the retina from the rest of the eye cup.
Anterior parts of the eye were discarded, whereas the retinas were
transferred to 1.5 ml Eppendorf tubes and snap-frozen using liquid
nitrogen, and then stored for later processing.
In Vivo Phenotyping by Electroretinography
We tested retinal function by electroretinogram (ERG) recordings
under anesthesia with ketamine (100 mg/kg) and xylazine (10 mg/kg).
ERG recordings were performed with a Diagnosys Celeris ERG device
(Lowell), as described previously (27,28).
Label-Free Protein Quantifications Following Data-dependent
Acquisition Mode
UCI_DDA_rd10_DR_cohort 2, UCI_DDA_P23H_cohort 1, and
UCI_DDA_Rpe65
−/−
cohort one were analyzed following the method A
steps, while UCI_DDA_rd10_CLR_cohort one was analyzed following
method B steps (see below).
Sample Preparation and Protein Digestion
Method A–The retinas were suspended in Urea buffer containing
8 M Urea, 0.1 M Tris-HCl (pH 8.5), 1% protease inhibitor cocktail, and
ultrasonicated on ice for 4 min, followed by centrifugation at 12,000g
at 4 ◦C for 10 min. The resulting supernatant (i.e., extracted retinal
proteins) was collected for digestion using the filter-aided sample
preparation (FASP) method (29). In brief, the extracted retinal proteins
were loaded into an Amicon centrifugal filter (Millipore) with a 30-kDa
cutoff. The proteins were then reduced with 10 mM dithiothreitol (DTT)
at 56 ◦C for 1 h and alkylated with 20 mM iodoacetamide (IAA) at room
temperature (RT) in the dark for another 1 h. Subsequently, the buffers
were replaced with 50 mM ammonium bicarbonate (AmBi) through
three washes of the filter membrane. Modified trypsin (Promega) was
added to the protein solution in a 1:50 (w/w) ratio and incubated at 37
◦C overnight. The digested peptides were collected via centrifugation,
along with an additional water rinse. This solution was then vacuum-
dried and reconstituted in 200 μl of 0.5% acetic acid. The peptide
mixture was desalted by C18 solid-phase extraction (provided by The
Nest Group, Inc), and vacuum-dried. For each analysis, 0.5 μgof
digested proteins were loaded for LC-MS/MS analysis in randomized
order.
Method B –Frozen specimens were ground to a powdered state
over liquid nitrogen. Fresh chilled sodium deoxycholate (SDC) lysis
buffer (4 % wt/vol SDC and 100 mM Tris-HCl, pH 8.5) was then added,
and the samples were incubated at 95 ◦C for 5 min with shaking at
Retinal Proteome Profiling in Inherited Retinal Degeneration
3Mol Cell Proteomics (2024) 23(11) 100855
1500 rpm. This was followed by sonication (Q125 Sonicator, Qsonica,
LLC) at 4 ◦C at maximum output power for two 10-min cycles. The
samples were reduced and alkylated using 10 mM TCEP and 40 mM
2-chloroacetamide (pH 7) (Sigma) while incubating for 5 min with
shaking at 1500 rpm at 45 ◦C, and then cooled. Lys-C (Thermo Sci-
entific, 90,051) and TPCK-treated trypsin (Worthington-biochem,
LS02124) were added at an enzyme-to-substrate ratio of 1:100 (w/w)
and digested for 16 h at 37˚C with shaking at 1500 rpm. The SDC was
removed from the samples by acidification with formic acid (FA)
to ~ pH 2, followed by centrifugation at 16,000 g for 5 min. Next, 5 mg
of a slurry of 10-μm C18 column beads (Dr Maisch GmbH) in aceto-
nitrile (Sigma-Aldrich, 1000294000) was loaded into a 200 μlfilter tip
(Vertex, 4237NSF) and the tryptic peptides were washed three times
with 0.1% FA (v/v). The peptides were then eluted with 80% aceto-
nitrile (v/v) (Sigma-Aldrich, 1000294000) and 0.1% FA (v/v), freeze-
dried, and reconstituted to 1 μg/μl with 0.5% formic acid (v/v)
(Sigma-Aldrich, 5330020050). An aliquot (1 μl) of each sample was
loaded for LC-MS/MS analysis in randomized order.
Peptide Separation and Data-dependent Acquisition (DDA)
Method A–Proteomics analysis was performed using an UltiMate
3000 UHPLC system (Thermo Fisher Scientific) connected directly to
an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scien-
tific) equipped with an ESI nanospray source. The mobile phase A
consisted of 0.1% FA in water, and mobile phase B consisted of 0.1%
FA in ACN. The peptides were eluted at a constant flow rate of 300 nl/
min and separated over an active 57-min gradient from 4% to 25%
buffer B, with a total sample runtime of 90 min on an Acclaim PepMap
RSLC column (50 cm ×75 μm). Survey MS scans were conducted in
the Orbitrap (FT) with an automated gain control (AGC) target of 8E5, a
maximum injection time of 50 ms, and a dynamic exclusion period of
30 s covering a scan range of 375 to 1800 m/z. MS/MS spectra were
gathered in data-dependent acquisition (DDA) mode at the highest
speed setting for 3-s cycles; the AGC target was set at 1E4 with a
maximum injection time of 35 ms. The ions underwent stepped-
energy, higher-energy collisional dissociation (seHCD) with a
normalized collision energy (NCE) of 20 ± 5%.
Method B –Proteomics analysis was conducted using an EASY-
nLC 1200 System (Thermo Fisher Scientific) linked to a Q-Exactive
mass spectrometer (Thermo Electron) via a nano-electrospray inter-
face operating in positive ion mode. Mobile phase A consisted of
0.1% FA in water, while mobile phase B was comprised of 0.1% FA in
80% acetonitrile. Peptides were loaded into a 75 μm I.D. ×150 mm
fused-silica analytical column packed in-house with 3 μm ReproSil-
Pur C18 resin (Dr Maisch GmbH). The flow rate was set at 250 nl/
min, and peptides were separated over a 105-min active gradient from
5% to 35% buffer B (total method duration 120 min). Survey MS scans
were captured in the Orbitrap with a resolution of 70k at m/z 400, and
the spray voltage was held at 2.0 kV. The capillary temperature was
maintained at 300 ◦C. Data-dependent MS/MS scans were performed
targeting the 12 most intense precursor ions with a dynamic exclusion
of 30 s. MS/MS resolution was set at 17.5k, and real-time internal
calibration was used to enhance mass accuracy with a lock mass of
background ion at 445.120025. Charge states that were unknown or
singly charged were excluded from MS/MS analysis. All data were
collected using Xcalibur software (ThermoFisher Scientific).
DDA Spectrum Match and Identification of Proteins
The raw LC-MS/MS data files were processed using MaxQuant
(version 2.1.0.0) (15), with the spectra matched against the Uniprot
mouse database (UP000000589, updated in April 2022, which con-
tains 21,957 protein entries plus additional Uniprot mouse database
which contains 41,543 protein isoform entries) and cRAP contaminant
database (updated in March 2019, which contains 100 protein entries).
For peptide identification, the mass tolerances applied were 20 ppm
for initial precursor ions and 0.5 Da for fragment ions. Specific Trypsin/
P digestion and up to two missed cleavages were permitted in the
tryptic digests and spectral search. Cysteine residues were treated as
static modifications, and oxidation of methionine was considered a
variable modification. Peptide identification filtering was conducted at
a 1% false discovery rate.
Label-Free Quantification of Proteins Following Data-Independent
Acquisition Mode
UEF_DIA_rd10_DR_cohort 3, and UEF_DIA_P23H_cohort two were
analyzed following method C steps.
Sample Preparation and Protein Digestion
Method C–The mouse retinas were homogenized in 100 μlof
protein extraction buffer (ab193970) (Abcam) using a handheld ho-
mogenizer (Pellet Pestle Cordless Motor, FisherScientific) for 30 s on
ice. The retinal homogenates were then solubilized by sonication in
two 30-s bursts, interspersed with 30-s intervals on ice. The lysates
were then centrifuged at 18,000gfor 20 min at 4 ◦C. The supernatants
(extracted solubilized proteins) were collected in separate Eppendorf
low-protein binding tubes (Thermo Fisher Scientific). Total protein
content was measured from retinal lysates using the BCA protein
assay (Thermo Fisher Scientific), and the total protein concentration
was adjusted to 1 mg/ml for all samples using the same lysis buffer.
The extracted and solubilized retinal proteins were processed using
FASP, as previously described (29,30). Briefly, a total of 50 μgof
protein fraction was loaded on a centrifugal filter 30-kDa cutoff (Merk
Millipore), and the buffer was exchanged with 0.1 M DTT (Merk) in UA
buffer (8 M Urea in 0.1 M Tris/HCl, pH 8.5). The mixture was incubated
at RT on a thermomixer at 500 rpm for 60 min and then washed twice
with UA buffer. Alkylation of the protein samples was performed using
0.05 M iodoacetamide (Merck) in UA buffer on the thermomixer at
300 rpm for 20 min in the dark. The reduced and alkylated proteins on
the filter were washed twice with UA buffer, then incubated with 49 μl
of 50 mM AmBi digestion buffer, 1:100 (w/w) endoproteinase LysC,
and 0.05% ProteaseMax (Promega) on the thermomixer at 600 rpm
and 30 ◦C for 3 h. Subsequently, 1 μl of TPCK-treated trypsin
(Promega) was added to the filter at a 1:50 (w/w) ratio and incubated
for 16 h at 37 ◦C. The digested peptides were recovered by centri-
fugation followed by two elution steps using 50 μl of 50% acetonitrile
in AmBi buffer. The solvent was evaporated via a SpeedVac vacuum
concentrator (Thermo Fisher Scientific) at RT. The dried samples were
reconstituted in a 2% acetonitrile/5% FA solution, and mixed on a
thermomixer at 300 rpm and RT for 20 min. Finally, 10 μl of each
sample, containing 10 μg of protein, was loaded for LC-MS/MS
analysis in randomized order.
Peptide Separation and Data-independent Acquisition (DIA)
Proteomics analysis was conducted using a UPLC (Vanquish Flex,
Thermo Scientific) coupled to a high-resolution Orbitrap Q Exactive
Classic mass spectrometer (Thermo Scientific) operating in positive
ion mode. Mobile phase A consisted of 0.1% FA in water, while mobile
phase B comprised 0.1% FA in ACN. Peptides were loaded into an
Agilent AdvanceBio Peptide Map column (2.1 mm ×250 mm, 2.7 μm,
Agilent Technologies). The flow rate was set at 0.3 ml/min, and pep-
tides were separated over an 80-min active gradient from 2% to 45%
buffer B (total method duration 90 min). Mass spectrometric detection
encompassed Full MS–SIM (Resolution: 35k; AGC target: 3e6; max
injection time: 60 ms; scan range: 385–1015 m/z) and DIA (Resolution:
17.5k; AGC target: 2e6; max injection time: 60 ms; loop count: 25;
isolation window: 24 m/z).
Retinal Proteome Profiling in Inherited Retinal Degeneration
Mol Cell Proteomics (2024) 23(11) 100855 4
DIA Spectrum Match and Identification of Proteins
The raw data were processed by DIA-NN software (version 1.8)
using the library-free DIA analysis mode (31) with default settings.
The MS/MS spectra library and retention times of peptides were
predicted using the UniProt reference proteome database for
mouse (UP000000589, updated in April 2022, which contains
21,957 protein entries plus additional Uniprot mouse database
which contains 41,543 protein isoform entries). Cysteine residues
were set as static modifications. Oxidation of methionine and N-
terminal acetylation were set as variable modifications while a
maximum number of variable modifications per peptide was set to
2. The predicted MS library was used to search the raw data,
applying 1% thresholds for both precursor and protein group false
detection rates, and requiring the presence of at least one proteo-
typic peptide ranging from 7 to 30 amino acids in length. For data
evaluation, the resulting MaxLFQ normalized intensities (15)were
used.
Experimental Design and Statistical Rationale
The output protein groups and their normalized intensities,
calculated by the MaxLFQ algorithm (15) integrated into Maxquant
and DIA-NN software, were used for further downstream analysis. The
number of replicates used in each mouse cohort is described in
Table 2, while detailed sample characteristics are described in the
supplementary file (Supp4-Sample_characteristics).
The MaxLFQ intensities were filtered and processed further using
Perseus software (32). First, potential contaminants, only identified
by site, or reversed hits were filtered out. The intensities of the
protein groups were log-transformed, and the study groups were
defined using the category-annotation tab. The groups were then
divided based on data completion and missing values into two
groups; complete dataset (100% valid values in total) and missing
values dataset (at least three valid values in at least one group). The
categorical groups in the complete datasets were compared and
analyzed using a linear model for microarray data test (LIMMA) (33)
followed by multiple testing false discovery rate (FDR) corrections;
and the statistical significance was set at q-value < 0.05. The cat-
egorical groups in the missing values datasets were analyzed using
the Student’sttest and the statistical significance was considered if
there were at least three valid values in each of the compared
groups and a p-value < 0.01. The KEGG (Kyoto Encyclopedia of
Genes and Genomes) pathway analysis from statistically significant
differentially expressed proteins (DEPs) was performed using
SRplot (34). Data visualization was performed using several R
packages such as dplyr (35), ggplot2 (36), KEGGREST (37), LIMMA
(33) pheatmap (38), and VennDiagram (39). Figure 1 and
Supplemental Figs. S1 and S2 were created using Jvenn (40), while
violin plots were prepared by SRplot (34). Inkspace 1.2 software
was used to combine the figures.
RESULTS
Profiling of the Mouse Retinal Proteome
In the present study, we identified and quantified over
7000 mouse retinal proteins from 6 mouse cohorts raised at
two different institutions, UCI and UEF. Frozen retinal samples
were processed in three different LC-MS laboratories (at UCI,
UO, and UEF) using two different label-free protein quantifi-
cation methods, the DDA and DIA modes. Consistent and
reliable identifications are presented in (Fig. 1).
The standard data-dependent acquisition (DDA) LC-MS
mode was employed to analyze the retinal proteome of
three different IRD mouse models (rd10, P23H, and Rpe65
−/−
).
Roughly 53% of identified retinal proteins were quantified in all
three models (Fig. 1A). To investigate the RP-associated
retinal proteome more comprehensively, we employed the
DIA analysis mode through the spectral library prediction tool,
as previously described (31). Importantly, by using the DIA
mode the overall proteome coverage was improved more than
four-fold (Fig. 1Band Supplemental Figs. S1 and S2). In the
DIA analyses, 77% of identified proteins were quantified in
both RP models.
Nevertheless, the consistent identification of proteins across
various research sites and analytical methods reinforces the
reliability and potential for comparative (semi-quantitative)
analysis between the datasets (Supplemental Fig. S4). Indeed,
selected marker proteins showed a similar pattern of expres-
sion change regardless of mouse housing institution (UCI &
UEF), mass spectrometer used, or analysis mode (DDA & DIA)
used in dark-reared rd10 mice (Supplemental Fig. S4A)or
vivarium-housed P23H mice (Supplemental Fig. S4B). Instead,
laboratory light conditions affected the retinal phenotype of the
rd10 mouse significantly (Supplemental Fig. S5), which has also
been described before (41).
Phenotypes of the Three IRD Models Based on
Electroretinography Recording and Marker Protein
Expression
Although the phenotypes of rd10, P23H, and Rpe65
−/−
mice
are well-characterized in the literature (see e.g.,(18–20), we
recorded scotopic electroretinograms (ERG) to highlight some
major differences between the models (Fig. 2A). At the time of
ERG recording, the mice were housed in similar conditions
TABLE 2
Number of replicates used for retinal proteome analysis
Condition DDA_rd10_DR DDA_rd10_CLR DDA_P23H DDA_Rpe65
−/−
DIA_rd10_DR DIA_P23H
RD model 6 (4f + 2m) 6 (3f + 3m) 7 (2f + 5m) 4 (2f + 2m)
a
4 (2f + 2m) 10 (5f + 5m)
WT 4 (2f + 2m) 6 (3f + 3m) 8 (4f + 4m) 3 (1f + 2m)
b
4 (2f + 2m) 4 (1f + 3m)
RD: retinal degeneration, WT: wildtype, f: female, m: male.
a
both retinas of two mice were pooled per replicate.
b
single retina of three mice were pooled per replicate.
Retinal Proteome Profiling in Inherited Retinal Degeneration
5Mol Cell Proteomics (2024) 23(11) 100855
and were at similar ages as those used for the DDA mode-
based proteomics (DDA proteomics data presented in
Figs. 3–7). The ERG responses in rd10 mice at this stage (P39)
were very small (Fig. 2A), and likely dominated by the activity
of surviving cone photoreceptors (42). In contrast, the P23H
mice (P90) displayed intermediately well-preserved and sen-
sitive ERGs, particularly with respect to the b-wave (Fig. 2A).
These responses are driven by the remaining rods together
with the well-preserved cone population (23). The Rpe65
−/−
mice (P42) displayed a minuscule a-wave but surprisingly
strong b-wave amplitudes (Fig. 2A). It is notable, however, that
the Rpe65
−/−
mice only started to respond at the highest
stimulus intensities, which indicates low sensitivity. These
unusual light responses of the Rpe65
−/−
mice are known to
arise from residual rod activity (43). In fact, retinal light re-
sponses are severely attenuated from birth in Rpe65
−/−
mice
(25), and they may be practically blind at low light levels such
as in standard vivarium conditions.
For proteomic data processing, before moving into analysis
of DEPs and KEGG pathways, we selected several rod- and
cone-enriched markers (44)(Supplemental Fig. S3), as well as
common inflammation markers, to provide an estimate of how
the different models compare with respect to photoreceptor
degeneration and inflammatory status at the time of sample
collection. Analysis of rod-enriched markers indicated a sig-
nificant loss of rod photoreceptors, with rd10 being the most
affected model, as evidenced by the decreased expression of
rod-enriched proteins (Fig. 2B). The pronounced down-
regulation of RHO and GNAT1 was comparable in rd10 and
P23H models. Rod degeneration is relatively slow in Rpe65
−/−
mice (26), which is evident in our data (Fig. 2A).
Analysis of cone-enriched markers OPN1MW, OPN1SW,
PDE6C, GNAT2, and GNGT2 showed downregulation collec-
tively only in the rd10 mice, but not in P23H mice (Fig. 2B).
None of the cone-enriched markers were detected in Rpe65
−/−
mice (Fig. 2B), highlighting the severe anatomic cone-
degeneration in these mice (45). Interestingly, while the
short-wavelength sensitive S-opsin (OPN1SW) was down-
regulated in P23H mice also, the middle-wavelength sensitive
M-opsin (OPN1MW) showed upregulation. This finding sug-
gests that the S-cones may be more susceptible to cell death
in the P23H mice, or that retinal degeneration overall is more
pronounced in the inferior retina where these cells are pre-
dominantly expressed (46). The trend towards increased
OPN1MW expression in P23H retinas could be due to ho-
meostatic regulation to counterbalance rod-pathway
dysfunction. Indeed, we previously observed oversensitive
cone-mediated ERG b-wave responses in P23H mice up to
3 months of age (28). In the same study, S-cone responses
appeared to start declining earlier than M-cone responses.
Inflammation markers GFAP, VIM, APOE, CTSB, GSTO1,
and A2M indicated significant inflammatory response in the
80
AB
196
111
57
42 353
952
D
D
A
_
r
d
1
0
_
C
L
R
D
D
A
_
P
2
3
H
D
D
A
_
R
p
R
e
6
5
-
/
-
DDA mode : size of each list
0
779
1558
113 1
rd10_CLR
1558
P23H
1458
Rpe65-/-
DIA mode: size of each list
0
779
1558
5377
rd10_CLR
6331
P23H
293 1247
5084
D
I
A
_
r
d
1
0
_
D
R
D
I
A
_
P
2
3
H
FIG.1. Comparison of the total number of reliably identified and quantified proteins between the DDA and DIA analysis modes. The
criterion for reliable identification: at least three valid values per group were required. A, DDA analysis mode; DDA_rd10_CLR cohort (rd10, n = 6;
WT, n = 6), DDA_P23H (P23H, n = 10; WT, n = 8), DDA_Rpe65
−/−
cohort (Rpe65
−/−
, n = 4; WT, n = 3). B, DIA analysis mode; DIA_rd10_CLR
cohort (rd10, n = 4; WT, n = 4) and DIA_P23H cohort (P23H, n = 10; WT, n = 5). The DDA spectral data were deconvoluted with the Andromeda
search engine integrated into MaxQuant software, while the DIA spectral data were deconvoluted using DIANN software. Both datasets were
filtered at a 1% FDR identification rate. CLR, Cyclic light rearing; DR, Dark rearing; WT, wildtype.
Retinal Proteome Profiling in Inherited Retinal Degeneration
Mol Cell Proteomics (2024) 23(11) 100855 6
retinas of all three IRD models (Fig. 2C). Overall, the order of
inflammatory status in the models at this disease stage
appeared to be from highest to lowest: rd10, Rpe65
−/−
, and
P23H. Table 3 provides a qualitative summary of retinal
degeneration severity in the three models.
Commonly Regulated Retinal Proteome in the RP and LCA
Mouse Models
The retinas from the three IRD models shared only a few
tens of commonly regulated DEPs (Fig. 3A), leading to a
handful of KEGG pathways that were commonly enriched
among the models (Fig. 3B). The common DEPs largely con-
sisted of cell stress-related proteins such as VIM and GFAP
(Fig. 3C). The four commonly enriched KEGG pathways were:
“Tight junction”,“GABAergic synapse”,“Phototransduction”,
and “Bacterial invasion of epithelial cells”(Fig. 3B). Down-
wards regulation in the “Phototransduction”pathway
(Supplemental Fig. S6) is clearly evident, as a majority of its
components are expressed in the photoreceptor cilia that
stereotypically degenerate in IRDs, including in the disease
models used here. The “Tight junction”pathway tends to be
upregulated in all three models, and commonly regulated
proteins include several actin- and myosin-related proteins
such as EZR, MYH9, MSN, ACTN1, and SLC9A3R1. These
changes may act to improve cell mobility, contractility, po-
larity, and survival. The regulation mediated by the KEGG
pathway “Bacterial invasion of epithelial cells”is also largely
linked to changes in cytoskeleton-related components such
as actin, catenin, and filament-forming proteins. The overall
upregulation of “GABAergic pathway”proteins (Supplemental
Fig. S7) indicates modified GABA activity.
Selectively Regulated Retinal Proteome in the LCA Model
Versus the RP Models
A total of 373 DEPs were dysregulated in Rpe65
−/−
mice
compared to WT (Fig. 4 and Supp2-proteins list). Among these
proteins, 83 showed expression changes in the opposite di-
rection (either upregulation or downregulation) compared to
those seen in rd10 or P23H mice, as illustrated in (Fig. 5). Many
crystallin proteins were highly downregulated in Rpe65
−/−
-4
-2
0
2
4
Rod markers
logFC (Disease / WT)
-4
-3
-2
-1
0
1
2
3
4
Cone markers
logFC (Disease / WT)
-4
-2
0
2
4
Inflammatory markers
logFC (Disease / WT)
RHO
PDE6b
GNAT1
GNB1
PDC
CNGA1
OPN1MW
OPN1SW
PDE6C
GNAT2
ARR3
GNGT2
GFAP
VIM
APOE
CTSB
GSTO1
A2M
rd10
P23H
Rpe65-/-
-4 -2 0 2
0
200
400
600
Stimulus intensity (log cd·s/m2)
Amplitude (μV)
b-wave
a-wave
-4 -2 0 2
0
200
400
600
Stimulus intensity (log cd·s/m2)
Amplitude (μV)
-4 -2 0 2
0
200
400
600
Stimulus intensity (log cd·s/m2)
Amplitude (μV)
100 200 300
-400
-200
200
400 C57BL/6J
Time (ms)
Amplitude (μV)
b-wave
a-wave
100 200 300
-400
-200
200
400 rd10
Time (ms)
Amplitude (μV)
-4 -2 0 2
0
200
400
600
Stimulus intensity (log cd·s/m2)
Amplitude (μV)
100 200 300
-400
-200
200
400
Rpe65-/-
Time (ms)
Amplitude (μV)
100 200 300
-400
-200
200
400 P23H
Time (ms)
Amplitude (μV)
A
B
FIG.2. Phenotypic characterization of the three IRD models used in DDA-mode proteomics. Phenotyping is based on scotopic ERG
recordings and the expression of cell-enriched marker proteins. A,upper panels show scotopic ERG waveforms (thick line represents group-
mean responses, while thin lines present individual mice) in response to an amplitude-saturating stimulus (log 1.7 cd s/m
2
). Lower panels
present ERG a- and b-wave amplitudes across a stimulus-intensity series. Gray, WT, n = 5; green, rd10, n = 7; blue, P23H, n = 5; red,Rpe65
−/−
,
n=7.B, protein expression changes expressed as log2FC values on the y-axis across three IRD mouse models compared to their respective
WT littermates, showing selected retinal rod, cone, and inflammatory markers. The data are presented as mean ± SEM.
Retinal Proteome Profiling in Inherited Retinal Degeneration
7Mol Cell Proteomics (2024) 23(11) 100855
samples while being stable or slightly upregulated in the RP
mouse retinas (Fig. 4). Similarly, some proteins related to
nucleotide metabolism (ARFGAP1, ARL6IP5, & CDC42EP4), or
cytoskeleton (EIF4G2, HDHD2, & TUBB4A), were down-
regulated in Rpe65
−/−
, but trending towards upregulation in RP
mice. Fatty acid-binding protein 5 (FABP5) was distinctly
downregulated in Rpe65
−/−
mouse retinas (>−2log2FC),while
being upregulated in rd10 and P23H mice (>0.5 log2FC).
Neurochondrin (NCDN) also followed the same pattern. In
contrast, some proteins such as PDC, SAMD11, STX3, SFXN5,
YBX3, AND HMGB2 were upregulated in Rpe65
−/−
mice, but
downregulated in both rd10 and P23H mice.
There was more discrepancy in Rpe65
−/−
versus
rd10 mouse proteomic regulation, as compared to Rpe65
−/−
versus P23H (Fig. 5). For instance, proteins related to nucle-
osome and chromosome phasing and histone binding to DNA
(e.g., H1F0, H2AFV; H2AFZ, HIST1H1A, HIST1H1B,
HIST1H1C, HIST1H1E, HMGA1, LMNB1, NOP58, RALY,
SMC1A, TOP1, and UPF1) were upregulated in Rpe65
−/−
but
downregulated in rd10 mouse retinas, while this pattern was
absent in P23H retinas.
The proteomic phenotype of Rpe65
−/−
mouse retina had
several distinct features (Fig. 6). These included dysregulation
in oxidative phosphorylation, the citrate cycle (TCA), and the
spliceosome pathway. Several components of the mitochon-
drial respiratory chain complex I (NDUFV2, NDUFAB1,
NDUFS3), also known as the NADH: ubiquinone oxidoreduc-
tase, were significantly downregulated in Rpe65
−/−
mouse
retinas (Fig. 7A). Instead, a major component of respiratory
chain complex II, SDHA, was significantly upregulated.
Several components of the downstream complexes III and IV
showed bidirectional regulations (e.g., opposite regulation of
UQCRH and UQCRB, or COX5A and COX5B); whereas the
subunits of complex V, or ATP synthase (such as ATP5D and
ATP5C1), collectively were upregulated in the Rpe65
−/−
sam-
ples. The protein expression in the components mentioned
above remained stable in both RP models. Tens of proteins
included in the TCA cycle pathway also showed bidirectional
regulation in Rpe65
−/−
retinas, but their regulation remained
collectively more stable in rd10 and P23H retinas
(Supplemental Fig. S8). Interestingly, while aconitase 1 (ACO1)
and mitochondrial isocitrate dehydrogenase (IDH2) were
clearly downregulated in Rpe65
−/−
retinas, the same proteins
were distinctly upregulated in rd10 retinas.
Changes in the proteome of the Rpe65
−/−
retina also sug-
gested an altered spliceosome pathway through overall
19
4
11
17
4
8
8
14
6
0 5 10 15 20
Number of DEPs
16
4
8
Ph
o
t
o
tr
a
n
sduc
ti
o
n
Bac
t
e
r
ial
i
nv
asion
of epithelia
l
cells
T
i
g
h
t
j
unct
i
o
n
GABAergic
s
y
napse
19 104
28
118 22
373
314
DEPs
37
723 425 22
Pathways
Common pathways
rd10
P23H
Rpe65-/-
GFAP
VCL
LRP1
CLU
APOE
MSN
VIM
LMNA
TF
GAP43
CORO1C
CTNNB1
NRCAM
GM20390;NME2
ATP6V1G2
NEBL
LASP1
TWF1
GSTT1
CAPN2
ARL3
IMPDH1
GNB1
HMGB2
PDC
ARHGAP1
DCLK1
TPPP3
-2
02
logFC
Common DEPs
AB C
D
FIG.3. Retinas from rd10, P23H and Rpe65
−/−
mice display only a handful of common DEPs. Green, Rd10; blue, P23H; red, Rpe65
−/−
.A,
The Venn diagram shows the number of common and specific DEPs. B, the Venn diagram shows the number of common and specific enriched
KEGG pathways across the three IRD models. C, the bar graph shows the common DEPs across the three models and their expression, as
depicted by their log2FC compared to their respective WT mice. D, the bar graph shows all the common enriched KEGG pathways across the
three models.
Retinal Proteome Profiling in Inherited Retinal Degeneration
Mol Cell Proteomics (2024) 23(11) 100855 8
rd10
P23H
Rpe65
-/-
rd10
P23H
Rpe65
-/-
rd10
P23H
Rpe65
-/-
rd10
P23H
Rpe65
-/-
P23H
Rpe65
-/-
rd10
P23H
Rpe65
-/-
L
o
g
2
f
o
f
f
l
d
c
h
a
n
g
e
FIG.4. The altered retinal proteome of LCA2-Rpe65
−/−
mice is largely different from those of RP mice. Heatmap focusing on the log2FC
of the DEPs in retinas of Rpe65
−/−
mice (Rpe65
−/−
, n = 4; WT, n = 3). Expression of the same proteins is shown for the rd10 (rd10, n = 4; WT,
n = 4) and P23H models (P23H, n = 10; WT, n = 5). Statistical analysis was performed by the LIMMA test followed by FDR correction. The + signs
indicate statistically significant changes at q-value < 0.05. DEPs that show opposing directions of regulation between the Rpe65
−/−
and RP
models are highlighted with boxes.
Retinal Proteome Profiling in Inherited Retinal Degeneration
9Mol Cell Proteomics (2024) 23(11) 100855
upregulation of spliceosome components, such as ribonu-
cleoproteins (HNRNPK, HNRNPA3, SF3B2, & SNRPD2), RNA
helicases (DDX46 & EIF4A3), and other regulatory or RNA-
binding proteins (FUS, SF3B1, SNRPA, RBM8A) (Fig. 7B).
This phenotype was not evident in the RP mouse retinas.
Analysis of the Retinal Proteome of the RP Mouse Models
with Increased Coverage
The DIA-MS-based analysis of the retinal proteome for rd10
and P23H mice showed higher proteome coverage compared
to the earlier DDA-MS-based method (~four-fold increase)
(Fig. 3 and Supplemental Fig. S2). This increased coverage
enabled a more detailed analysis of the proteome of the two
RP models. As a distinct general feature, significantly
upregulated DEPs (q-value < 0.05 and log2FC > 1) greatly
outnumbered downregulated DEPs (Fig. 8). A total of 151 and
197 proteins were upregulated compared to only 41 and 25
downregulated proteins in the rd10 and P23H retinas,
respectively (Fig. 8,Aand B). As expected, due to photore-
ceptor degeneration, most distinctly downregulated proteins
included photoreceptor cilia-specific components such as
RHO, ROM1 and subunits of PDE6. The most strongly upre-
gulated proteins in both RP models included immune system/
inflammation/stress response-associated proteins such as
GFAP, FGF2, H2-D1, S100 A.
Both rd10 and P23H retinas share 600 DEPs (Fig. 9A) which
form 90 commonly enriched KEGG pathways (Fig. 9B).
Selected pathways are shown in (Fig. 9C), while the rest of the
Opposite protein expression in P23H mouse retinas versus Rpe65-/-
+
+
++
++
+
++
+
+
+
+
+
+
+
+
+
++
+
+
+
++
++
++
++
+
+++
+
+
++
++
+
++
-2
0
2
AHCYL1
CADM1
CROCC
EIF2S3X;EIF2S3Y
EPB4.1;EPB41
EPB41L2;EPB4.1L2
FAM169A
H1F0
H2AFV;H2AFZ
HIST1H1A
HIST1H1B
HIST1H1C
HIST1H1E
HMGA1
HMGB2
LMNB1
NOP58
PDC
RALY
RCVRN
SLC1A2
SMC1A
STX3
TOP1
UPF1
logFC
++
+
++
+
+
+
+
-2
0
2
CNGB1
HMGB2
MFGE8
PDC
RCVRN
SLC24A1
logFC
+
+
+
+
+
+
+
+
+
++
+++
+
+
+
+
++
++
+
+
+
+
++
++
+
++
+
++
+
+
+
+
++
+
+
+
+
++
++
+
++
+
++++
++
++
+
+++
++
-2
0
2
ACO1
ACTR1A
ALDH3A2
ANXA1
ARHGAP1
ARPC5L
ATP1B3
CD81
CDH2
CKB
CNRIP1
CPNE6
CRMP1
CTSD
DBI
DCLK1
EEF1B;EEF1B2
ERP29
EZR
GUCY1A3
HINT1
IDH2
INA
ITPA
MIF
NAP1L1
OAT
PCP2
PDIA6
PEA15;PEA15A
PEBP1
PRDX3
PRDX5
PSMA2
PVALB
PYGM
RPLP2
RPN2
RPS5
RPSA
SYNCRIP
TOM1L2
TPPP3
TPT1
TUBB2A
UBQLN2
logFC
r
d10
R
pe65
-/-
+
+
++
++
+
+
++
+
+
+
+
++
+
+
+
-2
0
2
A
CAT2;ACAT3
ACSL3
ADSSL1
ARHGAP1
ARVCF
CORO2B
DCLK1
GUCY1A3
PTBP1
PYGM
SEPTIN2
SPTB
TPPP3
VAT1
logFC
P
23
H
R
pe65
-/-
Opposite protein expression in rd10 mouse retinas versus Rpe65-/-
A
B
FIG.5. Many proteins show differential direction of expression change in Rpe65
−/−
versus rd10, or Rpe65
−/−
versus P23H mouse
retinas. DEPs (based on LIMMA test followed by FDR correction with parameters: q-value < 0.05 and log2FC > 0.5) in the retinas of rd10 mice
(graph A; rd10, n = 6; WT, n = 6), or P23H mice (graph B; P23H, n = 10; WT, n = 8) as compared to those of Rpe65
−/−
mice (Rpe65
−/−
, n = 4; WT,
n = 3) are plotted on the x-axis, and the amplitude of log2FC (disease model versus WT) is plotted on the y-axis. Proteins selected for this
comparison needed to be statistically significant DEPs in the retinas of at least the rd10 or P23H mice; but statistical significance (q-value < 0.05)
is also marked with + signs for the retinas of Rpe65
−/−
mice.
Retinal Proteome Profiling in Inherited Retinal Degeneration
Mol Cell Proteomics (2024) 23(11) 100855 10
enriched pathways are depicted in (Supplemental Fig. S9). The
findings included, e.g., pathways related to the metabolism of
nucleotides, glutathione, and lipids (Supp3-List of Pathways).
The “Environmental Information Processing”KEGG category
included pathway regulation in, e.g., “cell communication,”
“adhesion,”and “intracellular regulation.”Changes in the
regulation of synaptic/neuronal remodeling and plasticity were
suggested by the enrichment of several pathways such as
regulation of actin cytoskeleton, axon guidance, and focal
adhesion. The KEGG “organismal systems”category showed
involvement of various neurotransmitter and hormone
signaling systems in the RP pathophysiology, such as estro-
gen, insulin, thyroid hormone, prolactin, oxytocin, growth
hormone, neurotrophins, cholinergic, glutamate, GABA,
adrenergic, and dopamine pathways. In the KEGG “human
diseases”category, several cancer-related pathways were
also enriched, indicating common signaling phenomena with
retinal neurodegeneration.
From these numerous common KEGG pathways “Purine
metabolism,”“cAMP signaling pathway,”and “cGMP-PKG
signaling pathway”are particularly relevant targets to thera-
peutic interventions (Y. (11,14,47,48); therefore, they are
considered here in more detail. Except for rod-enriched
components that are downregulated as a direct result of rod
degeneration (e.g., PDE6A, PDE6B, PDE6G, GUCY2E,
GUCY2F, NT5E, IMPDH1), the purine metabolism pathway
showed an overall trend towards upregulation in both rd10
and P23H retinas (Fig. 10A). IMPDH1, which catalyzes the
synthesis of xanthine monophosphate from inosine-5′-
monophosphate is downregulated, whereas ADSS1, an
enzyme that plays a role in the conversion of inosine mono-
phosphate to adenosine monophosphate is upregulated in RP
retinas. However, IMPDH1 is enriched in photoreceptors so it
could be decreased simply because of degeneration (44).
Several subtypes of adenylyl cyclases (ADCY2, ADCY5, and
ADCY8), which catalyze the formation of cAMP from ATP,
were significantly upregulated in P23H retinas. In addition,
several cAMP- and cGMP-degrading phosphodiesterases
(PDE1C, PDE3A, PDE4B, PDE4D) were upregulated in P23H
retinas. Adcy2 and Adcy8 mRNA primarily localizes to mouse
cone bipolar cells (CBCs), whereas Adcy5 mRNA is expressed
most abundantly in amacrine cells (AC) and retinal ganglion
cells (RGCs) (44)(Supplemental Fig. S10). Pde4b is mostly
expressed in CBCs while Pde4d expression is rather low but
may be primarily expressed in astrocytes (44). ADCY2,
PDE1C, and PDE3A proteins are overexpressed in both rd10
and P23H mouse retinas. Adcy2 mRNA is detected most
highly in CBCs, horizontal cells (HCs), and RGCs, Pde1c ap-
pears to localize primarily to CBCs, and Pde3a to HCs and
Acs in the mouse retinas (44).
0 5 10 15 20 25 30 35 40 45
Number of differentially expressed proteins
Oxidative phosphorylation
Citrate cycle (TCA cycle)
Cysteine and methionine metabolism
2-Oxocarboxylic acid metabolism
Spliceosome
Aminoacyl-tRNA biosynthesis
Phagosome
Cell cycle
Gap junction
Thermogenesis
Cardiac muscle contraction
Insulin secretion
Vasopressin-regulated water reabsorption
Alzheimer disease
Huntington disease
Prion disease
Chemical carcinogenesis - reactive oxygen species
Diabetic cardiomyopathy
Non-alcoholic fatty liver disease
Hepatitis C
Spinocerebellar ataxia
17
7
5
4
20
7
12
9
8
22
10
7
5
38
37
35
21
17
15
11
10
Human Diseases
Organismal Systems
Cellular Processes
Genetic Information Processing
Metabolism
FIG.6. Several KEGG pathways are uniquely altered in Rpe65
−/−
mouse retinas. KEGG pathways enriched exclusively in the retinas of
Rpe65
−/−
showing the number of DEPs involved in each pathway. Statistical analysis was performed by LIMMA test followed by FDR cor-
rections. DEPs here were identified with parameters: q-value < 0.05 and log2FC > 0.5, compared to protein expression of their respective WT
retinas. The arrows point to pathways of specific interest.
Retinal Proteome Profiling in Inherited Retinal Degeneration
11 Mol Cell Proteomics (2024) 23(11) 100855
A similar trend toward upregulation as in the “purine
metabolism”pathway is seen in the cAMP- and cGMP-
signaling pathways; in contrast, the rod-enriched cyclic
nucleotide-gated channels alpha and beta 1 (CNGA1, CNGB1)
are highly downregulated (Fig. 10,Band C). Both homologs of
rho kinases (ROCK1 & ROCK2) are upregulated in RP retinas.
The increased activity of the rho kinases has been implicated
frequently in neurodegeneration (49). Calcium/calmodulin-
dependent protein kinases CAMK2G and CAMK4 are also
upregulated in rd10 and P23H retinas. Camk2g and Camk4
mRNAs are mostly localized to ACs and RGCs, or CBCs,
respectively (44). Overall, it appears that the inner retina re-
sponds to photoreceptor degeneration by upregulating pro-
teins that belong to the purine metabolism and cAMP/cGMP
pathways, suggesting increased activity of the cAMP and/or
cGMP second messengers in retinal compartments that are
not affected primarily.
The retinal proteomes of both rd10 and P23H mice generally
exhibited changes in the same direction, either upregulated or
downregulated, compared to their respective WT controls;
NDUFV2
UQCRH
NDUFAB1
NDUFS3
UQCRFS1
ATP6V1G2
UQCRC2
ATP6V1G1
ATP6V1E1
ATP5D
CYC1
SDHA
COX6C
ATP5C1
UQCRB
COX5B
COX5A
NDUFV2
UQCRH
NDUFAB1
NDUFS3
UQCRFS1
ATP6V1G2
UQCRC2
ATP6V1G1
ATP6V1E1
ATP5D
CYC1
SDHA
COX6C
ATP5C1
UQCRB
COX5A
NDUFV2
UQCRH
NDUFAB1
NDUFS3
UQCRFS1
ATP6V1G2
UQCRC2
ATP6V1G1
ATP6V1E1
ATP5D
CYC1
SDHA
COX6C
ATP5C1
UQCRB
COX5B
COX5B
−3
−2
−1
0
1
2
rd10 P23H Rpe65 -/-
LogFC (Disease - Wild type)
Oxidative phosphorylation
Spliceosome
U2af2
Hnrnpa3
Srsf1
Rbm8a
Snrpa
Sf3b1
Hnrnpk
Srsf4
Srsf7
Fus
Snrpd2
Ddx46
Sf3b2
Eif4a3
EIF4A3
U2AF2
HNRNPA3
SRSF1
RBM8A
SNRPA
SF3B1
HNRNPK
SRSF4
SRSF7
FUS
SNRPD2
DDX46
SF3B2
EIF4A3
HNRNPA3
SRSF1
RBM8A
SNRPA
SF3B1
HNRNPK
SRSF4
SRSF7
FUS
SNRPD2
DDX46
SF3B2
−3
−2
−1
0
1
2
LogFC (Disease - Wild type)
Rpe65 -/-
P23H
rd10
B
A
FIG.7. Many key components of the oxidative phosphorylation and spliceosome system are uniquely altered in Rpe65
−/−
mouse
retinas. Many key components of the oxidative phosphorylation (A) or spliceosome (B) system are uniquely altered in Rpe65
−/−
mouse retinas.
The proteins highlighted in this figure are DEPs in the Rpe65
−/−
mouse retinas (Rpe65
−/−
, n = 4; WT, n = 3), and their expression change is
contrasted to the expression change of rd10 (rd10, n = 6; WT, n = 6) and P23H (P23H, n = 10; WT, n = 8) retinas. Statistical analysis was
performed by LIMMA test followed by FDR multiple testing corrections. DEPs were identified as differentially expressed proteins with the
following parameters versus WT: q-value < 0.05 and log2FC > 0.5.
TABLE 3
Summary of disease model phenotype
Model rd10 P23H Rpe65
−/−
Primary associated disease RP RP LCA
Rod degeneration progression +++ ++ +
Cone degeneration progression ++ + +++
Combined rod-cone dysfunction
(ERG)
+++ ++ +++
Inflammation markers +++ ++ +++
Analysis derived from literature data, as well as ERG and marker
protein expression data from this work.
Retinal Proteome Profiling in Inherited Retinal Degeneration
Mol Cell Proteomics (2024) 23(11) 100855 12
however, the extent of these changes varied with the pro-
gression of the disease (see example: Supplememental
Figs. S6–S8). However, some proteins exhibited opposite
expression patterns (absolute log2FC > 0.5) (Fig. 11), with only
10 of these protein changes (ALDH1A7, COX6B1, DNAJC19,
ENSA, IDE, MFGE8, NDUFAB1, OPN1MW, PCP4, ZNF219)
0 2 46810 12 14
Loading...
GLB1L2
TTC23L
GNAI1
PLCXD2
OPN1MW
ABCA4
ZNF143
H2-D1
H2-D1
GFAP
S100A6
RHO
FGF2
GNAT1
ROM1
CNTF
FABP 7
GNG11
S100A13
PBXIP1
INPPL1
TNFAIP3
GNGT1
MYOM1
ARL13B ISG15
A2M
GPNMB
IFITM3
MLKL
S100A16
PLPP2
TRAF3IP3
GBP2
AGTPBP1
IKBIP
H2-D1
SDC4
DDX58
GNA14
-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4
0246 8 10 12 14
WT_vs_rd10_-log(q-value)
WT_vs_P23H_-log(q-value)
Loading...
B2M
H2-D1
A2M
S100A6
GFAP
FGF2
COX6A1
MVP
LCP1
IQGAP1
PDLIM4
TGM2
ANK2
ASPH
CNTF
CD44
RCVRN
BBS4
PLPP2
RGS9BP
ABCA4
H1-1
SLC4A7
RGS9
PPEF2
ENSA
RHO
GUCY2E
GNGT1
ROM1
CNGB1
ANKRD33B
NXNL1
PCP4L1
CNGA1
PDE6A
PDE6B
GUCY2F
PDE6G
-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 11.5 22.5 33.5 4
(rd10-WT)_logFC
(P23H-WT)_logFC
rd10 retina
P23H retina
Upregulated (151 proteins)
Upregulated (197 proteins)
Downregulated (41 proteins)
Downregulated (25 proteins)
A
B
FIG.8. Pathologic upregulation of proteins is more prevalent than downregulation