Abstract
Mechanisms of inter-organ signaling have been established as hallmarks of nearly every pathophysiologic condition, where many exist as related and complex diseases. While significant work has been focused on understanding how individual cell types contribute and respond to specific perturbations related to common, complex disease, an equally-important but relatively less-explored question involves how relationships between organs are altered in the context of an integrated living organism. Current technical advances, such as proteomic analysis of plasma or conditioned media, have allowed for a more unbiased visualization and discovery of additional inter-tissue signaling molecules. However, one important feature which is lacking from these approaches is the ability to gain insight as to the function, mechanisms of action and target tissue(s) of relevant molecules. To begin to address these constraints, we initially developed a correlation-based bioinformatics framework which uses multi-tissue gene expression and/or proteomic data, as well as publicly available resources to statistically rank and functionally annotate endocrine proteins involved in tissue cross-talk. Using this approach, we identified many known and experimentally validated several novel inter-tissue circuits. This was this first study to directly link an endocrine-focused bioinformatics pipeline from population data directly to experimentally-validated mechanisms of inter-tissue communication. While these validations provide strong support for exploiting natural variation to discover new modes of communication, these serve as simple proof-of-principle studies and, thus, have promising potential for expansion. Some of these will be discussed during the presentation.
Presenter: Marcus Seldin, Ph.D. Assistant Professor, Biological Chemistry, University of California Irvine
Upcoming webinars schedule: https://dknet.org/about/webinar
Observational constraints on mergers creating magnetism in massive stars
dkNET Webinar: Population-Based Approaches to Investigate Endocrine Communication 02/26/2021
1. Population-based approaches to investigate
endocrine communication
Marcus Seldin
Department of Biological Chemistry
Center for Epigenetics and Metabolism
UCI
2. Outline
• Rationale for dissecting endocrinology
using population genetics approaches
• Can we identify known endocrine
interactions by surveying natural variation?
• Does these generalized principles allow us
to pinpoint new modes of inter-tissue
signaling?
• Can network-based approaches allow a
more “global” view of endocrine
interactions?
3. Complexity of tissue-tissue interactions
• Humans express ~3,000
secreted proteins, accounting
for 16% of the coding genome
• ~2000 have unknown functions
(UniProt Annotation Terms)
• Approach: Can natural variation
be used to deconvolute inter-
tissue communication?
Christian Rask-Madsen and C. Ronald Kahn, 2012
4. Natural variation in mice as a tool for discovery
Why study population
genetics in mice?
• Environment is restricted
• Penetrance of genetics
more apparent
• Mice are more honest on
questionnaires
• Tissues and traits more
accessible
• Rich datasets of targeted
experiments from
literature
• Same genetic
background can be
repeated across studies
7. Ranking system identifies many known (and unknown) endocrine axes
Bold – Human
SNP has been
associated with
relevant clinical
trait
Green
Background -
Peptide has
been shown to
act on target
tissue
Secreted Protein
Rank
8. Ranking system identifies many known (and unknown) endocrine axes
Bold – Human
SNP has been
associated with
relevant clinical
trait
Green
Background -
Peptide has
been shown to
act on target
tissue
Secreted Protein
Rank
9. Ranking system identifies many known (and unknown) endocrine axes
Bold – Human
SNP has been
associated with
relevant clinical
trait
Green
Background -
Peptide has
been shown to
act on target
tissue
Secreted Protein
Rank
10. Ranking system identifies many known (and unknown) endocrine axes
Bold – Human
SNP has been
associated with
relevant clinical
trait
Green
Background -
Peptide has
been shown to
act on target
tissue
42% of the top-ranked
interactions have
been validated by
previous studies
12. Ranking system identifies many known (and unknown) endocrine axes
Bold – Human
SNP has been
associated with
relevant clinical
trait
Green
Background -
Peptide has
been shown to
act on target
tissue
Secreted Protein
Rank
13. Adipose Lipocalin-5 is enriched for skeletal muscle
mitochondrial genes
LCN5 x Trait Correlations
LCN5 x Muscle Genes
16. Interaction score suggests human LCN6 as a functional
orthologue of mouse LCN5
In collaboration with Johan L. M. Björkegren and Simon Koplev
STARNET Population
17. Dissecting new circuits of physiologic communication using
pairwise comparisons
0 2 4 6
Erysipelotrichales
Erysipelotrichaceae
Clostridium
Bifidobacterium
Bifidobacteriales
Ruminococcus
Gammaproteobact…
-log(pvalue)
Microbiota Abundance
Microbiota Abundance
Fat Oxidation
TG Content
Intestine
ADAMDEC1
Glucose
Uptake
Time (min)
18. Dissecting new circuits of physiologic communication using
pairwise comparisons
0 2 4 6
Erysipelotrichales
Erysipelotrichaceae
Clostridium
Bifidobacterium
Bifidobacteriales
Ruminococcus
Gammaproteobact…
-log(pvalue)
Microbiota Abundance
Microbiota Abundance
Fat Oxidation
TG Content
Intestine
ADAMDEC1
Glucose
Uptake
19. Dissecting new circuits of physiologic communication using
pairwise comparisons
0 2 4 6
Erysipelotrichales
Erysipelotrichaceae
Clostridium
Bifidobacterium
Bifidobacteriales
Ruminococcus
Gammaproteobact…
-log(pvalue)
Microbiota Abundance
Microbiota Abundance
Fat Oxidation
TG Content
Intestine
ADAMDEC1
Glucose
Uptake
20. Perspectives (Part 1)
• Correlation-based models identify many known modes of endocrine interaction from mouse
population data
• New intertissue signaling mechanisms can be identified similarly
• Several novel regulators validate in vitro and in vivo
• Approach Advantages: Targeted, Simple to assign statistical thresholds, Generates immediate
testable hypotheses
• Approach Disadvantages: Subjective to spurious correlations or latent variable influence, fails to
account for cumulative interactions
22. Itih5 is expressed by mature adipocytes and correlated
with clinical traits in mice and humans
23. What is known about Itih5?
• Member of larger family of inter-alpha trypsin inhibitors (serine-type
endopeptidase inhibitors)
• Adipose expression is upregulated by HFD in mice and humans (Anveden Å.,
Obesity (Silver Spring). 2012)
• Promotes cell growth independent of protease inhibition (Rose M. Mol
Carcinog. 2018)
• Found covalently liked to bikunin in skin ECM, where it regulates HA signaling
Inflammatory skin disease (Huth S., Exp Dermatol. 2015)
• Implicated as a biomarker of several disease:
• BMI and HbA1c (Rönn T., Hum Mol Genet. 2015)
• Adenocarcinoma (Dötsch MM., Epigenetics. 2014)
• Colon cancer (Kloten V., Epigenetics. 2014)
• Breast cancer (Kloten V., Breast Cancer Res. 2013)
33. ITIH5 overexpression in HFD increases energy expenditure
proportional to amount of fat mass
In collaboration with Selma Masri (UCI), analysis using CalR
34. ITIH5 expression in HMDP adipose tissue negatively
correlates with heart AMPK/Foxo signaling pathways
35. ITIH5 reduces cardiac output in a diet-specific manner
LV Mass Ejection
Fraction
Diastolic
Vol
Heart
Rate
* *
*
*
• Lily Mott
• Casey Johnson
36. Physiologic mechanism of ITIH5
Genetic Variation
& HFD
ITIH5
Adipose energy
consumption
Local immune
recruitment
Other metabolic
consequences (ex.
cardiac output)
Adipose Vascular
and ECM
homeostasis
Systemic Insulin
resistance
37. Perspectives (Part 2)
• Network construction from different tissues highlighted adipose signaling as a central component
• Itih5 suggested as a key driver of this adipose tissue central module
• Expression of Itih5 correlates with various metabolic traits in a gene-by-diet fashion
• ITIH5 signals primarily in autocrine/paracrine in fat, which is blocked by high-fat diet
• Overexpression of ITIH5 alter adipose tissue morphology, glucose homeostasis and cardiac
output in a gene-by-diet manner
38. Outline
• Rational for studying endocrinology
• We can identify known endocrine interactions by
surveying “omics” data in natural variation
• This approach allow us to pinpoint new modes of
intertissue signaling
• Network-based approaches allow a more “global”
view of endocrine interactions (Itih5 maintains
adipose tissue homeostasis)
39. Future directions
• Method expansion to define endocrine
communication (Bayesian networks,
ICA decomposition, CNNs etc.)
• Web portal viewer for mouse and
human proteins (coming soon!)
• Get wet! (in vitro and in vivo models
using AAV and recombinant protein)
40. • Casey Johnson
• Leandro Velez
• Lilly Mott
• Jorge Luis Jr. Gorguet
• Cassandra Van
• Maggie Myers
• Daryn Chau
• Diana Quach
Acknowledgements
UC Irvine (Lab)
University of Eastern Finland
• Markku Laakso
Icahn-MSSM / Karolinska Institute
• Johan L. M. Björkegren
University of Wisconsin
• Alan D Attie
• Federico Rey
University of Sydney
• David E. James
• Jake Lusis
• Andrea Hevener
• Peter Tontonoz
• Arjun Deb
• Karen Reue
UCLA
UC Irvine
• Paolo Sassone-Corsi
• Selma Masri
• Cholsoon Jang
• Mike Zaragoza
• Anya Grosberg
University of Pittsburgh
• Erin Kershaw
• dkNET Pilot program in bioinformatics • K99/R00