results reported in text

Fig. 1a shows the “relative mRNA” of brown-fat-selective genes between wild-type and transgenic mice in two different fat depots. The authors write: “There were no significant alterations in the expression of brown-fat-selective genes in the interscapular brown adipose tissue or in the visceral (epididymal) white adipose tissue (Fig. 1a)”

Fig. 1b is the the relative mRNA in a third fat depot. The authors believe something different is going on here between the wild-type and transgenic mice: “However, the subcutaneous fat layer (inguinal), a white adipose tissue that is particularly prone to ‘browning’ (that is, formation of multilocular, UCP1-positive adipocytes), had significantly increased levels of Ucp1 and Cidea messenger RNAs (Fig. 1b).”

Fig. S1a “Similar to what has been reported, a twofold increase in Ucp1 mRNA expression was observed in the visceral, epididymal fat with 3 weeks of wheel running (Supplementary Fig. 1). However, a much larger change (approximately 25 fold) was seen in the same mice in the subcutaeneous inguinal fat depot.”

Fig S1b,c - “Similarly, a small increase in Ucp1 mRNA expression was seen in the epididymal fat with repeated bouts of swimming in warm (32 uC) water (Supplementary Fig. 1); however a very large increase (65 fold) was observed in the inguinal white depot (Supplementary Fig. 1). Thus, muscle-specific expression of PGC1a drives browning of subcutaneous white adipose tissue, possibly recapitulating part of the exercise program.”

My summary

1. authors want to show that some adipose tissue is upregulating “browning” proteins in response to pgc1-a (the (MCK)-PGC1-a transgenic).
2. Similar response occurs in exercised mice (replication of other studies)
3. Similar response in “cultured primary subcutaneous adipocytes with serum-free media conditioned by myocytes expressing PGC1-a” with “cells expressing green fluorescent protein (GFP)” as a control.* (Fig 1e)

1 and 2 are patterns that suggest something is 3 is an experiment to support inference from 1 and 2.

comments

• ucp1 effect is consistent in swimming and running mice

What the statistics are doing?

The authors are arguing that transgenic mice differ in “browning” protein expression in inguinal white fat but not epididymal white fat or brown fat. The evidence is 2/6 p < 0.05 values in inguinal fat but 0/10 in the other two depots. There is no attempt to measure effect sizes and uncertainty. An increase in expression occurs for all proteins in all three tissues, except adipoq in inguinal fat, but p < 0.05 in only 2 of the proteins in the inguinal fat. Cumulative probability of 2/16 p if null is true is 0.19.

What pattern would induce them to conclude something different from what they concluded? Does it even matter if the inguinal fat differs? What if no p-values were significant? There is a consistent effect. Is this enough to move forward? Does it matter that 4 of the p-values are not significant (and difference is in wrong direction)?

1. Fig 1A, the authors are using a p-value from a t-test to show no differences in expression of the five genes between white and brown adipose tissue in both wild type and transgenic mice. This isn’t what p-values do.
2. Fig 1B - Note the effect size in the subcutaneous depot (1B) are similar to that in the other two depots. The authors are inferrering something different in subcutaneous as opposed to epidydimal or BAT depots based on pattern of p-values - but “a difference in statistical significance is not (necessarily) statistically significant” and the similarity in effect sizes suggests this is especially so here.
3. Fig 1A,B How does use of t-test effect \(p\) given that the outcome is probably not approximately normally distributed? What is the distribution in these kinds of data? In general, it looks from both 1a and 1b that the variance increases with the mean.
4. Fig S1a - Clearly the SE is not appropriate (what is the mean - 2SE in the UCP1, wheel running, subcutaneou treatment?) nor a t-test given the extreme heterogenous variance.

Key to figure

1. (MCK)-PGC1-a : These MCK-PGC1-alpha transgenic mice express mouse peroxisome proliferative activated receptor, gamma, coactivator 1 alpha under the direction of the mouse muscle creatine kinase promoter. Muscle fibers from transgenic mice exhibit a more type II oxidative phenotype than wild-type. This mutant mouse strain may be useful in studies of muscle physiology and disease, exercise and oxidative capacity, and metabolic homeostasis (https://www.jax.org/strain/008231)
2. Ucp1 – uncoupling protein 1. Uncouples H+ gradient from ATP synthesis in mitochondria. Expressed in Brown Adipose Tissue (BAT)
3. Pgc1a – Peroxisome proliferator-activated receptor gamma coactivator 1-alpha. PGC-1α is a transcriptional coactivator that regulates the genes involved in energy metabolism. It is the master regulator of mitochondrial biogenesis.[6][7][8] This protein interacts with the nuclear receptor PPAR-γ, which permits the interaction of this protein with multiple transcription factors. This protein can interact with, and regulate the activities of, cAMP response element-binding protein (CREB) and nuclear respiratory factors (NRFs)[citation needed]. It provides a direct link between external physiological stimuli and the regulation of mitochondrial biogenesis, and is a major factor causing slow-twitch rather than fast-twitch muscle fiber types.[9] Endurance exercise has been shown to activate the PGC-1α gene in human skeletal muscle.[10] Exercise-induced PGC-1α in skeletal muscle increases autophagy [11] and unfolded protein response.[12] PGC-1α protein may be also involved in controlling blood pressure, regulating cellular cholesterol homoeostasis, and the development of obesity.[13]
4. Prdm – PR domain containing 16, also known as PRDM16, is a protein which in humans is encoded by the PRDM16 gene.[5][6] PRDM16 acts as a transcription coregulator that controls the development of brown adipocytes in brown adipose tissue.[7] Previously, this coregulator was believed to be present only in brown adipose tissue, but more recent studies have shown that PRDM16 is highly expressed in subcutaneous white adipose tissue as well.[7]
5. Cidea – Cell death activator CIDE-A is a protein that in humans is encoded by the CIDEA gene.[5][6][7] Cidea is an essential transcriptional coactivator regulating mammary gland secretion of milk lipids.[8] This gene encodes the homolog of the mouse protein Cidea that has been shown to activate apoptosis. This activation of apoptosis is inhibited by the DNA fragmentation factor DFF45 but not by caspase inhibitors. Mice that lack functional Cidea have higher metabolic rates, higher lipolysis in brown adipose tissue and higher core body temperatures when subjected to cold. These mice are also resistant to diet-induced obesity and diabetes. This suggests that in mice this gene product plays a role in thermogenesis and lipolysis. Two alternative transcripts encoding different isoforms have been identified.[7]
6. Adipoq – Adiponectin (also referred to as GBP-28, apM1, AdipoQ and Acrp30) is a protein hormone which is involved in regulating glucose levels as well as fatty acid breakdown. In humans it is encoded by the ADIPOQ gene and it is produced in adipose tissue.[5]
7. Ndufs1 – NADH-ubiquinone oxidoreductase 75 kDa subunit, mitochondrial (NDUFS1) is an enzyme that in humans is encoded by the NDUFS1 gene.[5] The encoded protein, NDUFS1, is the largest subunit of complex I, located on the inner mitochondrial membrane, and is important for mitochondrial oxidative phosphorylation. Mutations in this gene are associated with complex I deficiency.[6]

comments

• upward response of prdm16, ucp1, cidea and downward response of adipoq are p<0.05 and consistent with in vivo inquinal fat. The upward responses are concisistent with epididymal and BAT fat.
• responses about the same as in vivo but the errors are smaller.

What the statistics are doing?

4/5 are p<0.05 so the p-value is a marker that something is going on, to move forward. Would the plot itself be convincing? Probably no need of effect sizes because there is no theory to interpret this. But what if n were large and the effect were small, then would the experiment move on? what is the minimum effect size to move on?