People living with obesity have elevated levels of the peptide hormone leptin. This can be attributed to the amount of leptin in circulation being proportional to the amount of adipose tissue that an individual has (Obradovic et al., 2021). After being secreted by fat cells in adipose tissue, leptin predominantly acts in the hypothalamus as a major regulator of energy balance (Morrison, 2009). Likewise, both neural and adipose tissues are known to play an important role in the molecular etiology of body mass index (BMI), which is conventionally used to clinically diagnose obesity. However, despite being a cost-effective approach to routinely measure adiposity at scale, BMI is a construct that captures multiple heterogeneous subcomponents. This is epitomized by previous investigations into the functional genes that exert their effects on adiposity via their expression in brain and adipose tissues (Timshel et al., 2020; Rask-Andersen et al., 2019), highlighting the divergent pathways that exist between BMI and downstream complex traits.

Mendelian randomization (MR) is a causal inference technique that can exploit the random segregation of genetic variants within a population to evaluate the genetically predicted effects of modifiable exposures on complex outcomes (Davey Smith and Ebrahim, 2003; Richmond and Davey Smith, 2022). We recently extended the principles of MR to evaluate the separate effects of molecular subcomponents of BMI using genetic variants partitioned by their impact on gene expression derived from subcutaneous adipose tissue (SAT) and brain tissue (known as ‘tissue-partitioned MR’) (Leyden et al., 2022). Using this approach suggested that the brain-derived variants are predominantly responsible for driving the effect of BMI on cardiometabolic disease outcomes (Leyden et al., 2022), which we postulate is due to their involvement in appetite regulation and energy expenditure. These in turn influence the timelines and sites of adipose tissue distribution, which have more adverse consequences at the same level of BMI as do the processes influenced by the SAT-derived variants. Conversely, the SAT-derived set of instruments were predominantly responsible for the effect of BMI on outcomes such as endometrial cancer (Leyden et al., 2023). Differences in components of adiposity at the same level of BMI have previously been demonstrated with these instruments (Leyden et al., 2022), preserving the gene–environment equivalence assumption in MR (Davey Smith, 2012).

In this study, we sought to dissect the causal pathway between BMI and circulating leptin using tissue-partitioned MR by leveraging these brain- and adipose tissue-derived sets of instruments. However, investigating this hypothesis is made even more challenging given that associations between BMI and leptin levels have been reported as early in life as childhood (Shalitin and Phillip, 2003). Therefore, to further develop insight into the relationship between BMI and circulating leptin, we additionally applied another extension that we have developed in recent years known as ‘lifecourse MR’ (Richardson et al., 2020). This approach allows the independent effects of childhood and adult adiposity to be simultaneously estimated on an outcome that can also be measured at separate timepoints in the lifecourse (Richardson et al., 2022b; Sanderson et al., 2022), such as childhood (mean age: 10 y) and adulthood (mean age: 55 y) measures of circulating leptin levels in this study. Taken together, we aimed to conduct a proof-of-concept study for these novel extensions of the conventional MR approach.

In this study, we applied two recent approaches in the MR paradigm, known as lifecourse and tissue partitioned MR, to investigate the influence of adiposity on circulating leptin levels as an exemplar relationship to demonstrate the value of these techniques. Whilst there is irrefutable evidence that having a higher BMI increases circulating levels of leptin, these two extensions of MR methodology provide insight into lifecourse-specific effects and the tissues of action underlying this etiological relationship. We propose that these approaches can be conducted to build upon the findings by conventional MR analyses, which use naturally occurring genetic variants as causal anchors to help establish and estimate the causal effect of modifiable exposures on complex outcomes and disease endpoints.

Our application of lifecourse MR in this work supports a breadth of previous research using this technique, which highlights the importance of taking into consideration the age at which data from participants for exposure and outcome datasets are analyzed. Findings from this study provide evidence that adiposity in childhood exerts a direct effect on leptin levels when measured in early life, corroborating findings from observational studies (Shalitin and Phillip, 2003). Conversely, evidence of an indirect effect along the causal pathway involving adulthood adiposity was found when analyzing leptin measured in later life. This provides a powerful proof of concept for this approach, which has previously found evidence of a direct effect of increased childhood adiposity on lower risk of breast cancer (Richardson et al., 2020), elevated risk of type 1 diabetes (Richardson et al., 2022a), and an influence on cardiac structure (O’Nunain et al., 2022). Evidence supporting indirect effects using this approach, as in this study for adulthood leptin, has been found previously for various cardiovascular (Power et al., 2021) and site-specific cancer endpoints (Mariosa et al., 2022; Papadimitriou et al., 2023).

Additionally, in the multivariable MR analysis of childhood adiposity on childhood leptin, we found that the effect of adulthood adiposity, despite attenuating in comparison to univariable MR estimates, still provided some evidence of an indirect influence. However, within a lifecourse model an effect of adulthood adiposity on childhood leptin levels is impossible given that an adulthood risk factor cannot influence a trait measured at an earlier stage in the lifecourse. We therefore investigated this residual effect of the adulthood adiposity genetic score on in-depth measures of childhood adiposity using data from the age 10 clinic of the ALSPAC study. Similar evidence of a residual effect of the adulthood genetic score on childhood measured DXA-assessed fat mass was found in these analyses, whereas a direct effect of childhood adiposity was found on skinfold measures. These analyses further complement the study design of leveraging both individual- and summary-level data within an MR framework to develop more granular insight into the underlying etiological pathways between risk factors and complex traits. This residual influence of the adulthood BMI score on childhood leptin suggests that the aspects of childhood adiposity that relate to the adulthood genetic score in the multivariable analyses independently of the childhood score importantly influence leptin in childhood. In similar analyses for childhood vitamin D no such relationship was seen (Richardson et al., 2022b). This suggests that different aspects of childhood body composition influence circulating vitamin D and leptin levels, and demonstrates how cognisance of the gene-environment equivalence assumption in MR can motivate improved understanding of genotype to phenotype associations.

Tissue-partitioned MR analyses in this work suggest that the subcomponent of BMI proxied using genetic variants whose effects are putatively mediated via gene expression in brain tissue are predominantly responsible for driving effects of adiposity on leptin during both childhood and adulthood. We postulate that these findings highlight a role for appetite regulation and energy expenditure mechanisms as of fundamental importance in the effect of adiposity on leptin levels. This corroborates findings from the literature, which suggest that being overweight has a downstream consequence on elevated leptin levels due to leptin resistance occurring in the central nervous system, particularly the hypothalamus (Gruzdeva et al., 2019). This is influenced by factors such as blood–brain barrier permeability, which results in leptin failing to suppress appetite and consequently leads to circulating hyperleptinemia amongst patients with obesity (Izquierdo et al., 2019). Taken together with the evidence of an indirect effect highlighted by our lifecourse MR analysis, these findings suggest that leptin may have long-term consequences for appetite suppression. This in turn has an influence on excess body weight, which may start in childhood and then can be sustained into later life. Future research in this space should investigate large-scale genome-wide association studies (GWAS) outcomes related to subcutaneous adipose tissues (e.g. skinfold thickness), particularly given that this is where leptin is primarily produced (Russell et al., 1998).

Fractionating genetic instruments for the same exposure using molecular datasets has important considerations for the (often overlooked) gene–environment equivalence assumption in MR (Davey Smith, 2012). This states that the effect of germline genetic perturbations should have the same downstream consequence on outcomes as if they were caused by the modifiable exposures themselves. For the adipose- and brain tissue-partitioned sets of instruments used in this study, we found previously that the manner in which they relate to downstream disease and complex traits can drastically differ despite having almost identical average effect estimates on BMI as an exposure. In particular, this finding underlines the heterogeneous nature of BMI as a lifestyle risk factor and highlights that this human-derived construct likely captures various causal pathways underlying the relationship between anthropometry and complex outcomes. For example, we previously found evidence that the brain tissue-derived variants are predominantly responsible for the effect of BMI on both cigarette smoking and lung cancer (Leyden et al., 2023). We emphasize that future application of tissue-partitioned MR should carefully consider both the exposure and functionally relevant tissue types being investigated, as well as ensuring that the derived instrument sets are equally predictive of the exposure (Leyden et al., 2023). In particular, applications of a combined approach using both lifecourse and tissue-partitioned MR will likely be most powerful for exposures that have been analyzed by GWAS in large samples at distinct timepoints in the lifecourse (e.g. childhood and adulthood) where the functionally important tissue types have been well characterized by previous studies.

We note that both lifecourse and tissue-partitioned MR have important caveats. For instance, the childhood and adult body size instruments used to disentangle direct and indirect effects in this study do not provide insight into other timepoints over the lifecourse (e.g. adolescence). Future efforts should focus on deriving instruments to separate effects of age-specific adiposity at more granular windows over the lifecourse. Moreover, the childhood adiposity instruments are based on recall data, which is why they required validation in three independent cohorts as described in the ‘Methods’ section. In addition, tissue-partitioned instruments were derived using bulk tissue in this work due to the availability of data and therefore do not take into account cell-type heterogeneity (Glastonbury et al., 2019; Prince et al., 2021). Furthermore, as with all estimates derived from MR, triangulating findings from other orthogonal lines of evidence derived using different study design and datasets provides the most robust conclusions for the approaches applied in this work (Munafò and Davey Smith, 2018).

In summary, our findings highlight a putative role for genes expressed in neural tissues in the etiology of adiposity and leptin levels during both childhood and adulthood. Furthermore, this innovative study provides a proof of concept into how the principles of MR can be adapted to investigate the hypotheses outside the scope of how this causal inference technique was originally conceived (Davey Smith and Ebrahim, 2003).

All individual level data analysed in this study was obtained from the ALSPAC study which is not allowed to be deposited in a public repository. However, all data can be accessed via an application to ALSPAC which requires approval from executive committee (http://www.bristol.ac.uk/alspac/researchers/access/). Summary-level data on adulthood leptin levels were provided by the deCODE Health study which can be found at (https://download.decode.is/form/folder/proteomics). All other data analysed in this study is based on summary-level results as referenced throughout.

1. 1. Agrawal S
   2. Wang M
   3. Klarqvist MDR
   4. Smith K
   5. Shin J
   6. Dashti H
   7. Diamant N
   8. Choi SH
   9. Jurgens SJ
   10. Ellinor PT
   11. Philippakis A
   12. Claussnitzer M
   13. Ng K
   14. Udler MS
   15. Batra P
   16. Khera AV

   (2022) Inherited basis of visceral, abdominal subcutaneous and gluteofemoral fat depots

   Nature Communications 13:3771.

   https://doi.org/10.1038/s41467-022-30931-2

   • PubMed
   • Google Scholar
2. 1. Bowden J
   2. Davey Smith G
   3. Burgess S

   (2015) Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression

   International Journal of Epidemiology 44:512–525.

   https://doi.org/10.1093/ije/dyv080

   • PubMed
   • Google Scholar
3. 1. Bowden J
   2. Davey Smith G
   3. Haycock PC
   4. Burgess S

   (2016) Consistent estimation in mendelian randomization with some invalid instruments using a weighted median estimator

   Genetic Epidemiology 40:304–314.

   https://doi.org/10.1002/gepi.21965

   • PubMed
   • Google Scholar
4. 1. Boyd A
   2. Golding J
   3. Macleod J
   4. Lawlor DA
   5. Fraser A
   6. Henderson J
   7. Molloy L
   8. Ness A
   9. Ring S
   10. Davey Smith G

   (2013) Cohort Profile: the ’children of the 90s’--the index offspring of the Avon Longitudinal Study of Parents and Children

   International Journal of Epidemiology 42:111–127.

   https://doi.org/10.1093/ije/dys064

   • PubMed
   • Google Scholar
5. 1. Brandkvist M
   2. Bjørngaard JH
   3. Ødegård RA
   4. Åsvold BO
   5. Davey Smith G
   6. Brumpton B
   7. Hveem K
   8. Richardson TG
   9. Vie GÅ

   (2021) Separating the genetics of childhood and adult obesity: a validation study of genetic scores for body mass index in adolescence and adulthood in the HUNT Study

   Human Molecular Genetics 29:3966–3973.

   https://doi.org/10.1093/hmg/ddaa256

   • Google Scholar
6. 1. Burgess S
   2. Butterworth A
   3. Thompson SG

   (2013) Mendelian randomization analysis with multiple genetic variants using summarized data

   Genetic Epidemiology 37:658–665.

   https://doi.org/10.1002/gepi.21758

   • PubMed
   • Google Scholar
7. 1. Davey Smith G
   2. Ebrahim S

   (2003) Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease?*

   International Journal of Epidemiology 32:1–22.

   https://doi.org/10.1093/ije/dyg070

   • Google Scholar
8. 1. Davey Smith G

   (2012) Epigenesis for epidemiologists: does evo-devo have implications for population health research and practice?

   International Journal of Epidemiology 41:236–247.

   https://doi.org/10.1093/ije/dys016

   • PubMed
   • Google Scholar
9. 1. Ferkingstad E
   2. Sulem P
   3. Atlason BA
   4. Sveinbjornsson G
   5. Magnusson MI
   6. Styrmisdottir EL
   7. Gunnarsdottir K
   8. Helgason A
   9. Oddsson A
   10. Halldorsson BV
   11. Jensson BO
   12. Zink F
   13. Halldorsson GH
   14. Masson G
   15. Arnadottir GA
   16. Katrinardottir H
   17. Juliusson K
   18. Magnusson MK
   19. Magnusson OT
   20. Fridriksdottir R
   21. Saevarsdottir S
   22. Gudjonsson SA
   23. Stacey SN
   24. Rognvaldsson S
   25. Eiriksdottir T
   26. Olafsdottir TA
   27. Steinthorsdottir V
   28. Tragante V
   29. Ulfarsson MO
   30. Stefansson H
   31. Jonsdottir I
   32. Holm H
   33. Rafnar T
   34. Melsted P
   35. Saemundsdottir J
   36. Norddahl GL
   37. Lund SH
   38. Gudbjartsson DF
   39. Thorsteinsdottir U
   40. Stefansson K

   (2021) Large-scale integration of the plasma proteome with genetics and disease

   Nature Genetics 53:1712–1721.

   https://doi.org/10.1038/s41588-021-00978-w

   • PubMed
   • Google Scholar
10. 1. Fraser A
    2. Macdonald-Wallis C
    3. Tilling K
    4. Boyd A
    5. Golding J
    6. Davey Smith G
    7. Henderson J
    8. Macleod J
    9. Molloy L
    10. Ness A
    11. Ring S
    12. Nelson SM
    13. Lawlor DA

    (2013) Cohort profile: the avon longitudinal study of parents and children: ALSPAC mothers cohort

    International Journal of Epidemiology 42:97–110.

    https://doi.org/10.1093/ije/dys066

    • PubMed
    • Google Scholar
11. 1. Glastonbury CA
    2. Couto Alves A
    3. El-Sayed Moustafa JS
    4. Small KS

    (2019) Cell-type heterogeneity in adipose tissue is associated with complex traits and reveals disease-relevant cell-specific eQTLs

    American Journal of Human Genetics 104:1013–1024.

    https://doi.org/10.1016/j.ajhg.2019.03.025

    • PubMed
    • Google Scholar
12. 1. Gruzdeva O
    2. Borodkina D
    3. Uchasova E
    4. Dyleva Y
    5. Barbarash O

    (2019) Leptin resistance: underlying mechanisms and diagnosis

    Diabetes, Metabolic Syndrome and Obesity 12:191–198.

    https://doi.org/10.2147/DMSO.S182406

    • PubMed
    • Google Scholar
13. 1. Hemani G
    2. Zheng J
    3. Elsworth B
    4. Wade KH
    5. Haberland V
    6. Baird D
    7. Laurin C
    8. Burgess S
    9. Bowden J
    10. Langdon R
    11. Tan VY
    12. Yarmolinsky J
    13. Shihab HA
    14. Timpson NJ
    15. Evans DM
    16. Relton C
    17. Martin RM
    18. Davey Smith G
    19. Gaunt TR
    20. Haycock PC

    (2018) The MR-Base platform supports systematic causal inference across the human phenome

    eLife 7:e34408.

    https://doi.org/10.7554/eLife.34408

    • PubMed
    • Google Scholar
14. 1. Izquierdo AG
    2. Crujeiras AB
    3. Casanueva FF
    4. Carreira MC

    (2019) Leptin, obesity, and leptin resistance: where are we 25 years later?

    Nutrients 11:2704.

    https://doi.org/10.3390/nu11112704

    • PubMed
    • Google Scholar
15. 1. Leyden GM
    2. Shapland CY
    3. Davey Smith G
    4. Sanderson E
    5. Greenwood MP
    6. Murphy D
    7. Richardson TG

    (2022) Harnessing tissue-specific genetic variation to dissect putative causal pathways between body mass index and cardiometabolic phenotypes

    American Journal of Human Genetics 109:240–252.

    https://doi.org/10.1016/j.ajhg.2021.12.013

    • PubMed
    • Google Scholar
16. 1. Leyden GM
    2. Greenwood MP
    3. Gaborieau V
    4. Han Y
    5. Amos CI
    6. Brennan P
    7. Murphy D
    8. Davey Smith G
    9. Richardson TG

    (2023) Disentangling the aetiological pathways between body mass index and site-specific cancer risk using tissue-partitioned Mendelian randomisation

    British Journal of Cancer 128:618–625.

    https://doi.org/10.1038/s41416-022-02060-6

    • Google Scholar
17. 1. Mariosa D
    2. Smith-Byrne K
    3. Richardson TG
    4. Ferrari P
    5. Gunter MJ
    6. Papadimitriou N
    7. Murphy N
    8. Christakoudi S
    9. Tsilidis KK
    10. Riboli E
    11. Muller D
    12. Purdue MP
    13. Chanock SJ
    14. Hung RJ
    15. Amos CI
    16. O’Mara TA
    17. Amiano P
    18. Pasanisi F
    19. Rodriguez-Barranco M
    20. Krogh V
    21. Tjønneland A
    22. Halkjær J
    23. Perez-Cornago A
    24. Chirlaque M-D
    25. Skeie G
    26. Rylander C
    27. Borch KB
    28. Aune D
    29. Heath AK
    30. Ward HA
    31. Schulze M
    32. Bonet C
    33. Weiderpass E
    34. Davey Smith G
    35. Brennan P
    36. Johansson M

    (2022) Body size at different ages and risk of 6 cancers: a mendelian randomization and prospective cohort study

    Journal of the National Cancer Institute 114:1296–1300.

    https://doi.org/10.1093/jnci/djac061

    • PubMed
    • Google Scholar
18. 1. Morrison CD

    (2009) Leptin signaling in brain: a link between nutrition and cognition?

    Biochimica et Biophysica Acta 1792:401–408.

    https://doi.org/10.1016/j.bbadis.2008.12.004

    • PubMed
    • Google Scholar
19. 1. Munafò MR
    2. Davey Smith G

    (2018) Robust research needs many lines of evidence

    Nature 553:399–401.

    https://doi.org/10.1038/d41586-018-01023-3

    • PubMed
    • Google Scholar
20. 1. Obradovic M
    2. Sudar-Milovanovic E
    3. Soskic S
    4. Essack M
    5. Arya S
    6. Stewart AJ
    7. Gojobori T
    8. Isenovic ER

    (2021) Leptin and obesity: role and clinical implication

    Frontiers in Endocrinology 12:585887.

    https://doi.org/10.3389/fendo.2021.585887

    • PubMed
    • Google Scholar
21. 1. O’Nunain K
    2. Park C
    3. Urquijo H
    4. Leyden GM
    5. Hughes AD
    6. Davey Smith G
    7. Richardson TG

    (2022) A lifecourse mendelian randomization study highlights the long-term influence of childhood body size on later life heart structure

    PLOS Biology 20:e3001656.

    https://doi.org/10.1371/journal.pbio.3001656

    • PubMed
    • Google Scholar
22. 1. Papadimitriou N
    2. Bull CJ
    3. Jenab M
    4. Hughes DJ
    5. Bell JA
    6. Sanderson E
    7. Timpson NJ
    8. Smith GD
    9. Albanes D
    10. Campbell PT
    11. Küry S
    12. Le Marchand L
    13. Ulrich CM
    14. Visvanathan K
    15. Figueiredo JC
    16. Newcomb PA
    17. Pai RK
    18. Peters U
    19. Tsilidis KK
    20. Boer JMA
    21. Vincent EE
    22. Mariosa D
    23. Gunter MJ
    24. Richardson TG
    25. Murphy N

    (2023) Separating the effects of early and later life adiposity on colorectal cancer risk: a Mendelian randomization study

    BMC Medicine 21:5.

    https://doi.org/10.1186/s12916-022-02702-9

    • PubMed
    • Google Scholar
23. 1. Pietzner M
    2. Wheeler E
    3. Carrasco-Zanini J
    4. Cortes A
    5. Koprulu M
    6. Wörheide MA
    7. Oerton E
    8. Cook J
    9. Stewart ID
    10. Kerrison ND
    11. Luan J
    12. Raffler J
    13. Arnold M
    14. Arlt W
    15. O’Rahilly S
    16. Kastenmüller G
    17. Gamazon ER
    18. Hingorani AD
    19. Scott RA
    20. Wareham NJ
    21. Langenberg C

    (2021a) Mapping the proteo-genomic convergence of human diseases

    Science 374:eabj1541.

    https://doi.org/10.1126/science.abj1541

    • PubMed
    • Google Scholar
24. 1. Pietzner M
    2. Wheeler E
    3. Carrasco-Zanini J
    4. Kerrison ND
    5. Oerton E
    6. Koprulu M
    7. Luan J
    8. Hingorani AD
    9. Williams SA
    10. Wareham NJ
    11. Langenberg C

    (2021b) Synergistic insights into human health from aptamer- and antibody-based proteomic profiling

    Nature Communications 12:6822.

    https://doi.org/10.1038/s41467-021-27164-0

    • PubMed
    • Google Scholar
25. 1. Power GM
    2. Tyrrell J
    3. Frayling TM
    4. Davey Smith G
    5. Richardson TG

    (2021) Mendelian randomization analyses suggest childhood body size indirectly influences end points from across the cardiovascular disease spectrum through adult body size

    Journal of the American Heart Association 10:e021503.

    https://doi.org/10.1161/JAHA.121.021503

    • PubMed
    • Google Scholar
26. 1. Prince C
    2. Mitchell RE
    3. Richardson TG

    (2021) Integrative multiomics analysis highlights immune-cell regulatory mechanisms and shared genetic architecture for 14 immune-associated diseases and cancer outcomes

    American Journal of Human Genetics 108:2259–2270.

    https://doi.org/10.1016/j.ajhg.2021.10.003

    • PubMed
    • Google Scholar
27. 1. Rask-Andersen M
    2. Karlsson T
    3. Ek WE
    4. Johansson Å

    (2019) Genome-wide association study of body fat distribution identifies adiposity loci and sex-specific genetic effects

    Nature Communications 10:339.

    https://doi.org/10.1038/s41467-018-08000-4

    • PubMed
    • Google Scholar
28. 1. Richardson TG
    2. Sanderson E
    3. Elsworth B
    4. Tilling K
    5. Davey Smith G

    (2020) Use of genetic variation to separate the effects of early and later life adiposity on disease risk: mendelian randomisation study

    BMJ 369:m1203.

    https://doi.org/10.1136/bmj.m1203

    • PubMed
    • Google Scholar
29. 1. Richardson TG
    2. Mykkänen J
    3. Pahkala K
    4. Ala-Korpela M
    5. Bell JA
    6. Taylor K
    7. Viikari J
    8. Lehtimäki T
    9. Raitakari O
    10. Davey Smith G

    (2021) Evaluating the direct effects of childhood adiposity on adult systemic metabolism: a multivariable Mendelian randomization analysis

    International Journal of Epidemiology 50:1580–1592.

    https://doi.org/10.1093/ije/dyab051

    • PubMed
    • Google Scholar
30. 1. Richardson TG
    2. Crouch DJM
    3. Power GM
    4. Morales-Berstein F
    5. Hazelwood E
    6. Fang S
    7. Cho Y
    8. Inshaw JRJ
    9. Robertson CC
    10. Sidore C
    11. Cucca F
    12. Rich SS
    13. Todd JA
    14. Davey Smith G

    (2022a) Childhood body size directly increases type 1 diabetes risk based on a lifecourse Mendelian randomization approach

    Nature Communications 13:2337.

    https://doi.org/10.1038/s41467-022-29932-y

    • PubMed
    • Google Scholar
31. 1. Richardson TG
    2. Power GM
    3. Davey Smith G

    (2022b) Adiposity may confound the association between vitamin D and disease risk - a lifecourse Mendelian randomization study

    eLife 11:e79798.

    https://doi.org/10.7554/eLife.79798

    • PubMed
    • Google Scholar
32. 1. Richmond RC
    2. Davey Smith G

    (2022) Mendelian randomization: concepts and scope

    Cold Spring Harbor Perspectives in Medicine 12:a040501.

    https://doi.org/10.1101/cshperspect.a040501

    • PubMed
    • Google Scholar
33. 1. Russell CD
    2. Petersen RN
    3. Rao SP
    4. Ricci MR
    5. Prasad A
    6. Zhang Y
    7. Brolin RE
    8. Fried SK

    (1998) Leptin expression in adipose tissue from obese humans: depot-specific regulation by insulin and dexamethasone

    The American Journal of Physiology 275:E507–E515.

    https://doi.org/10.1152/ajpendo.1998.275.3.E507

    • PubMed
    • Google Scholar
34. 1. Sanderson E
    2. Davey Smith G
    3. Windmeijer F
    4. Bowden J

    (2019) An examination of multivariable Mendelian randomization in the single-sample and two-sample summary data settings

    International Journal of Epidemiology 48:713–727.

    https://doi.org/10.1093/ije/dyy262

    • PubMed
    • Google Scholar
35. 1. Sanderson E
    2. Richardson TG
    3. Morris TT
    4. Tilling K
    5. Davey Smith G

    (2022) Estimation of causal effects of a time-varying exposure at multiple time points through multivariable mendelian randomization

    PLOS Genetics 18:e1010290.

    https://doi.org/10.1371/journal.pgen.1010290

    • PubMed
    • Google Scholar
36. 1. Shalitin S
    2. Phillip M

    (2003) Role of obesity and leptin in the pubertal process and pubertal growth--a review

    International Journal of Obesity and Related Metabolic Disorders 27:869–874.

    https://doi.org/10.1038/sj.ijo.0802328

    • PubMed
    • Google Scholar
37. 1. Timshel PN
    2. Thompson JJ
    3. Pers TH

    (2020) Genetic mapping of etiologic brain cell types for obesity

    eLife 9:e55851.

    https://doi.org/10.7554/eLife.55851

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Time-varying and tissue-dependent effects of adiposity on leptin levels: a Mendelian randomization study" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, including Edward D Janus as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Matthias Barton as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Despoina Manousaki (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this letter to help you prepare a revised submission.

Essential revisions:

1) It might not fit directly with the proof of concept aim of the manuscript, but this study could be a nice opportunity to test the opposite association, ie if leptin affects adiposity at different life stages. The relationship between BMI and leptin levels is complex as nicely depicted in lines 248-253 of the discussion. In the introduction, it is stated that leptin acts in the hypothalamus to regulate appetite. If the retroaction works, the regulation of appetite is expected to decrease appetite and food intake, and maintain BMI to a normal level. The knock-outed mice for leptin or congenital leptin deficiency in humans result in uncontrolled satiety and obesity. A bidirectional MR of Steiger filtering in the forward MR could answer this question.

2) In the discussion could the authors say if there are any other examples of already analysed or potentially interesting indirect life course effects ie expand on reference 26? Can they provide other examples of tissue-partitioned effects either already shown or potentially relevant? Can they suggest examples of other issues that this combined approach could address? This could give readers an even better idea of how these new approaches could be used.

Reviewer #1 (Recommendations for the authors):

I don't find a lot that needs to be addressed.

I could not find Supplementary notes 1 and 2.

In the discussion could the authors say if there are any other examples of already analysed or potentially interesting indirect life course effects ie expand on reference 26?

Can they provide other examples of tissue-partitioned effects either already shown or potentially relevant?

Any suggestions/examples for other issues that this combined approach could address?

This could give readers an even better idea of how these new approaches could be used.

Reviewer #2 (Recommendations for the authors):

This is an interesting study exploring lifecourse effects of adiposity on leptin levels, and using tissue-partitioned MR instruments to study the effects of BMI on leptin levels. While the study question is appealing, the methods are novel and they have been nicely applied in the context of the present study, and the results are interesting, I find that certain parts of the manuscript could be further clarified.

Also, it might not fit directly with the proof of concept aim of the manuscript, but this study could be a nice opportunity to test the opposite association, ie if leptin affects adiposity at different life stages. The relationship between BMI and leptin levels is complex as nicely depicted in lines 248-253 of the discussion. In the introduction, it is stated that leptin acts in the hypothalamus to regulate appetite. If the retroaction works, the regulation of appetite is expected to decrease appetite and food intake, and maintain BMI to a normal level. The knock-outed mice for leptin or congenital leptin deficiency in humans result in uncontrolled satiety and obesity. A bidirectional MR of Steiger filtering in the forward MR could answer this question.

Below are my point-by-point comments:

Abstract:

Methods: When the authors say in the results: "along the causal pathway", do they mean that they adjusted for adult body size in multivariable MR? If yes, please add this to the methods section of the abstract.

The units of change in exposure that confer the betas in the leptin outcome are not exposed in the abstract or in the manuscript (see additional comment on this below). This raises a challenge in directly comparing the two betas (ie in the childhood vs adulthood MR analysis).

In the conclusion, the phrase "as well as raising implications for the gene-environment equivalence assumption in MR". It is not clear how this phrase is justified by the analyses presented in the abstract. It is explained in the discussion of the paper, but I would suggest omitting it from the abstract conclusion.

Some specifications are needed in regards to the exposure and the outcome of GWAS sources.

For the exposure GWAS: In the BMJ paper cited as the source of the SNPs for adult and childhood body size, it is described that adult body size was measured based on the measured body mass index during adulthood (mean age 56.5) while child body mass was self-reported perceived body size at age 10. This is important information that should appear in the methods section of the manuscript. This leads to the question, could the fact that the IVs for measuring adult vs child body mass were based on a different method of estimating body size explain the difference we see in the results of the lifecourse MR? I understand that these IVs successfully predicted adult vs child BMI in 3 cohorts, but the measure of the exposure conferred by the genetic instrument in adults vs children is different. This should be highlighted in the methods, Results section and as a limitation of this study.

In terms of the outcome GWAS:

In line 105, could the authors specify how were leptin levels measured in ALSPAC? Was the method the same as that in DECODE? If the method was different, could this as well induce any bias in the MR estimates?

In terms of comparison between ALSPAC and DECODE, the Icelandic population of DECODE may have a distinct genetic architecture from ALSPAC or other European cohorts. This could introduce bias due to population stratification when using instruments for the exposure that comes from European meta-analyses. Is there another adult proteomic GWAS with available leptin levels which can be used to replicate the results in DECODE?

Line 138: in the multivariable MR analyses, which were the other variables tested as exposures (ie other than the adiposity variants for children or adults?) -this part should be clarified in both methods and Results section. I understand that adulthood and childhood body size were the two exposure variables, based on the DAGs in Figure 1. It would be helpful to put the DAGs in the two boxes entitled "Genetic variants" the GWAS source of the genetic variants and specify: " genetic variants for adulthood adiposity" or "genetic variants for childhood adiposity" depending on the analysis. Also, could the authors explain in the legend the interpretation of the colours of the arrows (ie black vs red?).

Line 144: Could the authors specify if in the tissue partitioned MR analysis the IVs were for adult BMI?

Line 167: in the phrase: "We firstly estimated the total effect of childhood adiposity on circulating leptin using data measured in ALSPAC participants at mean age 9.9 years in the lifecourse", please specify that you are referring to a univariate MR analysis.

In Tables S3 and S5, please provide the p-value of the intercept of the MR-Egger, as well as the Cohrane Q p-values for IVW and MR-Egger, and based on these results, please comment on the presence or absence of evidence of pleiotropy in the IVs used in the MR analyses.

Discussion: It would be nice to further elaborate on the hypothesis explaining the finding in regards to the indirect effect of childhood adiposity on adult leptin levels. In line 239, the authors explain this finding by "individuals in a population remaining overweight for many years in the lifecourse". The effect of early-life adiposity remains after adjustment for adult adiposity. Can the results of the tissue-partitioned MR (in terms of the sole effect of brain BMI SNPs on leptin levels in childhood vs a composite effect of adipose and brain tissue IVs on adult leptin levels) inform us on a possible hypothesis explaining the above phenomenon? Could early-life adiposity be associated with specific hypothalamic responses persisting in adulthood?

In regards to the sensitivity analysis on the effect of tissue-partitioned BMI on visceral vs subcutaneous fat, the authors could provide further explanation on the relationship between leptin secretion and fat distribution providing a rationale for this analysis.

Essential revisions:

1) It might not fit directly with the proof of concept aim of the manuscript, but this study could be a nice opportunity to test the opposite association, ie if leptin affects adiposity at different life stages. The relationship between BMI and leptin levels is complex as nicely depicted in lines 248-253 of the discussion. In the introduction, it is stated that leptin acts in the hypothalamus to regulate appetite. If the retroaction works, the regulation of appetite is expected to decrease appetite and food intake, and maintain BMI to a normal level. The knock-outed mice for leptin or congenital leptin deficiency in humans result in uncontrolled satiety and obesity. A bidirectional MR of Steiger filtering in the forward MR could answer this question.

Many thanks for this suggestion. We have now conducted an additional analysis to evaluate the effect of circulating leptin levels on adulthood BMI using a cis-acting protein quantitative trait loci (pQTL) located in the LEP gene as an instrumental variable:

This provided comparatively weaker effect estimates between circulating leptin levels in adulthood on BMI. pQTL data from the deCODE study was used for this analysis as there were no genome-wide corrected pQTL (i.e. P<5x10-8) in the Fenland study at this locus.

However, as the reviewer mentions this work is outside the scope of this current manuscript. Further research investigating this hypothesis is therefore necessary to robustly evaluate time-varying and tissue-dependent relationships once the relevant data become accessible.

2) In the discussion could the authors say if there are any other examples of already analysed or potentially interesting indirect life course effects ie expand on reference 26? Can they provide other examples of tissue-partitioned effects either already shown or potentially relevant? Can they suggest examples of other issues that this combined approach could address? This could give readers an even better idea of how these new approaches could be used.

We have added references to site-specific cancer endpoints where our lifecourse approach has previously provided evidence of an indirect effect of childhood adiposity (page 15):

“Evidence supporting indirect effects using this approach, as in this study for adulthood leptin, have been found previously for various cardiovascular (30) and site-specific cancer endpoints (31, 32).”

As recommended by Reviewer #2, we have also added some discussion regarding interpretation of findings from both approaches together (page 16):

“Taken together with the evidence of an indirect effect highlighted by our lifecourse MR analysis, these findings suggest that leptin may have long term consequences for appetite suppression. This in turn has an influence on excess body weight which may start in childhood and then can be sustained into later life.”

We have also expanded on further applications of this approach to page 17:

“In particular, applications of a combined approach using both lifecourse and tissue-partitioned MR will likely be most powerful for exposures which have been analysed by GWAS in large samples at distinct timepoints in the lifecourse (e.g. childhood and adulthood) where the functionally important tissue types have been well characterized by previous studies.”

Reviewer #1 (Recommendations for the authors):

I don't find a lot that needs to be addressed.

I could not find Supplementary notes 1 and 2.

Many thanks for your comments to help refine this manuscript. We have now incorporated Supplementary Notes 1 and 2 into the main article file as requested by the editorial team at eLife.

In the discussion could the authors say if there are any other examples of already analysed or potentially interesting indirect life course effects ie expand on reference 26?

We have now added examples of indirect lifecourse effects as suggested to page 15 of the manuscript:

“Evidence supporting indirect effects using this approach, as in this study for adulthood leptin, have been found previously for various cardiovascular (30) and site-specific cancer endpoints (31, 32).”

Can they provide other examples of tissue-partitioned effects either already shown or potentially relevant?

Other examples of tissue-partitioned effects have now been described on page 17:

“For example, we previously found evidence that the brain tissue-derived variants are predominantly responsible for the effect of BMI on both cigarette smoking and lung cancer (8).”

Any suggestions/examples for other issues that this combined approach could address?

This could give readers an even better idea of how these new approaches could be used.

We have expanded on further applications of this approach to page 17:

“In particular, applications of a combined approach using both lifecourse and tissue-partitioned MR will likely be most powerful for exposures which have been analysed by GWAS in large samples at distinct timepoints in the lifecourse (e.g. childhood and adulthood) where the functionally important tissue types have been well characterized by previous studies.”

Reviewer #2 (Recommendations for the authors):

This is an interesting study exploring lifecourse effects of adiposity on leptin levels, and using tissue-partitioned MR instruments to study the effects of BMI on leptin levels. While the study question is appealing, the methods are novel and they have been nicely applied in the context of the present study, and the results are interesting, I find that certain parts of the manuscript could be further clarified.

Also, it might not fit directly with the proof of concept aim of the manuscript, but this study could be a nice opportunity to test the opposite association, ie if leptin affects adiposity at different life stages. The relationship between BMI and leptin levels is complex as nicely depicted in lines 248-253 of the discussion. In the introduction, it is stated that leptin acts in the hypothalamus to regulate appetite. If the retroaction works, the regulation of appetite is expected to decrease appetite and food intake, and maintain BMI to a normal level. The knock-outed mice for leptin or congenital leptin deficiency in humans result in uncontrolled satiety and obesity. A bidirectional MR of Steiger filtering in the forward MR could answer this question.

Many thanks for this suggestion. We have now conducted an additional analysis to evaluate the effect of circulating leptin levels on adulthood BMI using a cis-acting protein quantitative trait loci (pQTL) located in the LEP gene as an instrumental variable:

This provided comparatively weaker effect estimates between circulating leptin levels in adulthood on BMI. pQTL data from the deCODE study was used for this analysis as there were no genome-wide corrected pQTL (i.e. P<5x10-8) in the Fenland study at this locus.

However, we believe this work is outside the scope of this current manuscript. Further research investigating this hypothesis is therefore necessary to robustly evaluate time-varying and tissue-dependent relationships once the relevant data become accessible.

Below are my point-by-point comments:

Abstract:

Methods: When the authors say in the results: "along the causal pathway", do they mean that they adjusted for adult body size in multivariable MR? If yes, please add this to the methods section of the abstract.

We have clarified this point on page 2 as suggested:

“Genetic instruments for childhood and adult adiposity were incorporated into a multivariable MR (MVMR) framework to estimate lifestage specific effects on leptin levels measured during early life (mean age:10 years) in the Avon Longitudinal Study of Parents and Children (ALSPAC) and in adulthood (mean age:55 years) using summary-level data from the deCODE Health study.”

The units of change in exposure that confer the betas in the leptin outcome are not exposed in the abstract or in the manuscript (see additional comment on this below). This raises a challenge in directly comparing the two betas (ie in the childhood vs adulthood MR analysis).

Units of change have now been added to the abstract (1 standard deviation change in circulating leptin levels). The reviewer makes a valid point that using SD change makes comparisons between childhood and adulthood estimates challenging given that the measures of leptin derived from these cohorts will differ in terms of how they vary across populations. We have emphasized this point on page 8:

“Although effect estimates from these summary-level data can be interpreted as a 1-SD change in normalized plasma leptin levels, we note that direct comparisons between estimates derived from childhood measures of leptin in ALSPAC should not be drawn given the various sources of heterogeneity between these measures.”

In the conclusion, the phrase "as well as raising implications for the gene-environment equivalence assumption in MR". It is not clear how this phrase is justified by the analyses presented in the abstract. It is explained in the discussion of the paper, but I would suggest omitting it from the abstract conclusion.

We have omitted this section of the abstract as recommended.

Some specifications are needed in regards to the exposure and the outcome of GWAS sources.

For the exposure GWAS: In the BMJ paper cited as the source of the SNPs for adult and childhood body size, it is described that adult body size was measured based on the measured body mass index during adulthood (mean age 56.5) while child body mass was self-reported perceived body size at age 10. This is important information that should appear in the methods section of the manuscript. This leads to the question, could the fact that the IVs for measuring adult vs child body mass were based on a different method of estimating body size explain the difference we see in the results of the lifecourse MR? I understand that these IVs successfully predicted adult vs child BMI in 3 cohorts, but the measure of the exposure conferred by the genetic instrument in adults vs children is different. This should be highlighted in the methods, Results section and as a limitation of this study.

We have added further detail to the childhood and adult body size instruments section as suggested on page 6:

“Briefly, genetic instruments derived in the UK Biobank study based on self-reported body size at age 10 and measured adulthood body mass index (11) have been shown to separate clinically measured childhood and adult BMI in 3 independent cohorts (ALSPAC (11), the Young Finns Study (14) and the Trøndelag Health (HUNT) study (15)).”

We have also added details to the methods section to describe analyses conducted in our original paper which suggested that recall bias related to the childhood measure is unlikely to have influenced the conclusions of our lifecourse MR approach. This point has also been highlighted in our discussion on page 17:

“Moreover, the childhood adiposity instruments are based on recall data, which is why they required validation in 3 independent cohorts as described in our methods section.”

In terms of the outcome GWAS:

In line 105, could the authors specify how were leptin levels measured in ALSPAC? Was the method the same as that in DECODE? If the method was different, could this as well induce any bias in the MR estimates?

Further detail to leptin measures analysed in this study are now reported on page 7 and 8. We have also emphasized that the SomaLogic measure of circulating leptin previously provided a high degree of concordance with an alternative measure using the Olink platform, reinforcing its reliability as a measure of this protein given that Olink measures have previously been validated using the same technology as the one used in the ALSPAC study (immunosorbent assay (ELISA)):

“Leptin was measured by an in-house enzyme-linked immunosorbent assay (ELISA) validated against commercial methods. Analyses in ALSPAC were undertaken on a final sample size of 4,155 individuals after removing those without genetic data and withdrawn consent.”

“Genetic estimates on adulthood circulating leptin levels were obtained from a study of 10,708 individuals enrolled in the Fenland study. Full details have been described elsewhere (24). Leptin was measured using the SomaScan v4 assay which applies single-stranded oligonucleotides aptamers with specific binding affinities to protein targets. This assay has also been reported to provide a very highly correlated measure of circulating leptin using the antibody based Olink platform (r=0.95) (25). The Fenland study was approved by the National Health Service (NHS) Health Research Authority Research Ethics Committee (NRES Committee-East of England Cambridge Central, ref. 04/Q018/19). All participants provided written informed consent.”

In terms of comparison between ALSPAC and DECODE, the Icelandic population of DECODE may have a distinct genetic architecture from ALSPAC or other European cohorts. This could introduce bias due to population stratification when using instruments for the exposure that comes from European meta-analyses. Is there another adult proteomic GWAS with available leptin levels which can be used to replicate the results in DECODE?

Many thanks for this suggestion. Since this manuscript was submitted, a protein GWAS study conducted in the Fenland study has made their full summary statistics available (Pietzner et al., Science (2021)). We have there now undertaken an additional analysis using Leptin data derived from the UK based Fenland cohort which we have included as our primary analysis, with deCODE analyses still included as a further validation analysis.

Line 138: in the multivariable MR analyses, which were the other variables tested as exposures (ie other than the adiposity variants for children or adults?) -this part should be clarified in both methods and Results section. I understand that adulthood and childhood body size were the two exposure variables, based on the DAGs in Figure 1. It would be helpful to put the DAGs in the two boxes entitled "Genetic variants" the GWAS source of the genetic variants and specify: " genetic variants for adulthood adiposity" or "genetic variants for childhood adiposity" depending on the analysis. Also, could the authors explain in the legend the interpretation of the colours of the arrows (ie black vs red?).

We have now clarified in the figure legends that multivariable MR analyses requires genetic variants for all exposures to be modelled simultaneously (as well as interpretation regarding the colours of arrows).

Line 144: Could the authors specify if in the tissue partitioned MR analysis the IVs were for adult BMI?

We have now clarified on page 6 that these IVs were based on adult BMI.

Line 167: in the phrase: "We firstly estimated the total effect of childhood adiposity on circulating leptin using data measured in ALSPAC participants at mean age 9.9 years in the lifecourse", please specify that you are referring to a univariate MR analysis.

We have updated this sentence accordingly (page 9):

“We firstly conducted univariable MR to estimate the total effect of childhood adiposity on circulating leptin using data measured in ALSPAC participants at mean age 9.9 years in the lifecourse (Β=1.28 per 1-standard deviation (SD) change in leptin levels per change in body size category, 95% CI=1.10 to 1.46, P=2x10^-41).”

In Tables S3 and S5, please provide the p-value of the intercept of the MR-Egger, as well as the Cohrane Q p-values for IVW and MR-Egger, and based on these results, please comment on the presence or absence of evidence of pleiotropy in the IVs used in the MR analyses.

These additional statistics have been added to Supplementary Tables and referenced on page 12:

“Cochran’s Q-statistics for these analyses, as well as those derived in Lifecourse MR analyses, can be found in Supplementary File 1l, along with intercept terms for MR-Egger analyses which did not suggest horizontal pleiotropy may be biasing conclusions.”

Discussion: It would be nice to further elaborate on the hypothesis explaining the finding in regards to the indirect effect of childhood adiposity on adult leptin levels. In line 239, the authors explain this finding by "individuals in a population remaining overweight for many years in the lifecourse". The effect of early-life adiposity remains after adjustment for adult adiposity. Can the results of the tissue-partitioned MR (in terms of the sole effect of brain BMI SNPs on leptin levels in childhood vs a composite effect of adipose and brain tissue IVs on adult leptin levels) inform us on a possible hypothesis explaining the above phenomenon? Could early-life adiposity be associated with specific hypothalamic responses persisting in adulthood?

We have added some further discussion regarding this point to page 16:

“Taken together with the evidence of an indirect effect highlighted by our lifecourse MR analysis, these findings suggest that leptin may have long term consequences for appetite suppression. This in turn has an influence on excess body weight which may start in childhood and then can be sustained into later life.”

In regards to the sensitivity analysis on the effect of tissue-partitioned BMI on visceral vs subcutaneous fat, the authors could provide further explanation on the relationship between leptin secretion and fat distribution providing a rationale for this analysis.

Discussion regarding this point can be found on page 8:

“We postulate that these findings highlight a role for appetite regulation and energy expenditure mechanisms as of fundamental importance in the effect of adiposity on leptin levels.”

Author details

1. Tom G Richardson

   MRC Integrative Epidemiology Unit (IEU), Population Health Sciences, Bristol Medical School, University of Bristol, Oakfield House, Oakfield Grove, Bristol, United Kingdom

   Contribution

   Data curation, Formal analysis, Visualization, Methodology, Writing - original draft, Writing - review and editing

   For correspondence

   Tom.G.Richardson@bristol.ac.uk

   Competing interests

   TGR is an employee of GlaxoSmithKline outside of this work

   "This ORCID iD identifies the author of this article:" 0000-0002-7918-2040

2. Genevieve M Leyden

   MRC Integrative Epidemiology Unit (IEU), Population Health Sciences, Bristol Medical School, University of Bristol, Oakfield House, Oakfield Grove, Bristol, United Kingdom

   Contribution

   Data curation, Formal analysis, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

3. George Davey Smith

   MRC Integrative Epidemiology Unit (IEU), Population Health Sciences, Bristol Medical School, University of Bristol, Oakfield House, Oakfield Grove, Bristol, United Kingdom

   Contribution

   Conceptualization, Supervision, Funding acquisition, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1407-8314

• 253

  Page views

• 36

  Downloads

• 0

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.