Write a comment

In this manuscript, Nicholson et al. investigated whether hyperglycemia caused by decreased insulin vs insulin resistance drives AD pathology. The interest in addressing this question stems from the observation that type 2 diabetes and people with metabolic syndrome (MetS) show early onset of Alzheimer’s disease (AD). To address this question, the authors performed two types of interventions, separately, in a double knock-in clinical-relevant AD mouse model: 1) the administration of streptozotocin (STZ) to destroy beta cells and mimic type 1 diabetes, and 2) a high-fat, high sugar (HFHS) diet to mimic type 2 diabetes. This design allows them to parse out the respective contributions of hyperglycemia and metabolic dysfunction to AD pathology. The authors then performed an extensive comparison across these models to define the role of metabolic dysfunction in contributing to the cognitive deficits independent of hyperglycemia. Using single-nucleus RNAseq, they identified transcriptional changes in glial populations that they call metabolic impairment in neurodegeneration (MinD) state which is really interesting, and this provided some insight on a few pathways (including Nrg3-pathway) in metabolic dysfunction that could be contributing to neurodegeneration. We have several recommendations for the authors to strengthen the conclusions drawn from the evidence presented and improve data presentation.

Major:

1. The conclusion regarding slower learning in STZ-treated mice (Ext fig.1) in the Morris Water Maze test is unclear. The authors state that “STZ-treated WT and AD mice exhibited longer latencies to locate the hidden platform during initial acquisition and reversal learning (Extended Data Fig. 1d, g).” However, in Ext Fig 1d, g, the latency difference is mainly observed between WT (blue and yellow) and AD mice (green and pink). Based on the figure, STZ treatment does not appear to induce changes in either WT mice or AD mouse model. To help clarify, the authors are encouraged to evaluate their statistical approach to see if it is the best fit for the data presented. A repeated measures ANOVA may be a good fit.

2. The authors conclude that “cognitive outcomes are shaped by metabolic stress driven by insulin resistance, rather than hyperglycemia being the primary factor” based on the observations that mice with HFHS diet showed increased body weight but mice with STZ did not, although both groups exhibited hyperglycemia (Fig 1a, ext Fig 1a). While the results show that hyperglycemia alone doesn’t induce cognitive deficits, claiming that metabolic stress (rather than hyperglycemia) drives cognitive decline requires more direct evidence. Specifically, the conclusion that STZ mice do not exhibit metabolic stress needs evidence. The authors could consider either revise their wording to make the claim more precise or provide additional evidence demonstrating that STZ mice did not experience metabolic stress despite having hyperglycemia.

3. The HFHS diet interaction with Trem2 is quite interesting, since Trem2 has been shown as PD risk factor. However, the absence of significant difference in synaptic density between DKI-HFHS vs DKI-lean mice raises some concern about statistical power. Since the effect size between WT and DKI appear small in this case, could the authors estimate or perform a power calculation to see how many samples would be needed to detect such a small difference between either DKI-lean vs WT or DKI-lean vs DKI-HFHS.

4. In both Fig3 f and Ext. Fig 7d, when the authors showed DEGs for DKI-HFHS and MinD State, the DKI-HFHS changes are going in the opposite direction of changes in either WT-HFHS or DKI-lean mice. These data raise the question of whether these MinD genes play a protective or disease-exacerbating role. The authors should consider discussing what role these genes played.

5. When the authors look into the Meis2_ IN population, the data support an enrichment of Meis2+ neurons (Fig 6e). However, when examining the specific role that these neurons contribute to the AD pathology, it is important to compare the unique changes detected in DKI-HFHS vs WT-HFHS mice to gain insight into whether and how this population is altered between healthy and disease states in response to HFHS challenge.

6. The authors mentioned that it is known that MEIS2 is a transcription factor that regulates gene expression upon glucose dysregulation. To strengthen their conclusion, the authors can assess some of the target genes that are highly regulated by MEIS2. The authors can analyze the binding motif for MEIS2 to identify targets.

7. Gene ontology analysis cannot conclude that excitability is a defining feature of InN3 neurons under metabolic stress without additional experimental data. To address this point, the authors can either comment on this limitation in the discussion or support it with electrophysiological data to show the excitability change.

8. The argument regarding MG3 cluster increase needs to be supported with a statistical test (for example: propeller method) for the percentage of MG3 in the microglial population. This method would account for high sample-to-sample variability from single cell data.

9. Since the authors use mouse data to understand human disease, it would be helpful for the authors to provide context from the existing literature in the discussion about the link between T2DM and AD risk.

10. Lastly, the authors conclude that the HFHS diet compromises cognition independent of Abeta pathology and that cognitive loss is purely secondary to metabolic stress (in the second result and conclusion). This conclusion should be qualified to acknowledge the limitation of mouse model. In humans with T2DM or metabolic syndrome, patients are known to have small vessel disease that could either aggravate the effects or might lead to APP pathology only in the human context (PMID:30106209, PMID: 29686024). The authors could discuss how these human-specific vascular pathologies may limit the translatability of their findings and revise their conclusion to reflect that the observed independence from Aβ pathology may not fully capture the complexity of metabolic syndrome-associated cognitive decline in human patients.

Minor:

1. Can the authors clarify the “homozygous App^NL-F/Mapt^hMAPT” notation? From what I interpreted, this represents a double knock-in. If the mice are homozygous for both knock-in alleles, consider specifying the full genotype as App^NL-F/NL-F;Mapt^hMAPT/hMAPT to improve clarity.

2. There are a lot of places where figures are referenced incorrectly throughout the manuscript. The authors need to review and correct these references. Specific instances are given below:

a. In ext Fig1, the label for panel a indicates “WT-lean, WT-HFHS, DKI-lean and DKI HFHS”. Based on the text, it seems that STZ manipulation is independent of the HFHS diet manipulation. Could the authors clarify whether this is a typo in the legend, or whether these mice were indeed treated with both STZ + HFHS diet?

b. In ext Fig1 d, the pink legend. Should this be App/hTau STZ rather than WT? Please double check this labeling.

c. The authors state that “All groups performed similarly in the visible platform test, confirming intact visual acuity and motor function (Fig. 1d).” However, Fig 1d shows blood glucose levels, not behavioral performance data.

d. My comments are indicated in brackets within the quoted text:

“We next assessed the phenotypic effects of diet-induced insulin resistance in the HFHS cohort (Fig. 1a [Figure reference issue: 1a shows the graphic design only, the citation should likely be entire Fig. 1] and Extended Data Fig. 2). Unlike STZ-treated mice, both WT and homozygous AppNL-G-F/MapthMAPT (DKI) mice on a chronic HFHS diet exhibited significant weight gain compared to lean-diet-fed controls (Fig. 1b [Fig.1b is just a picture, 1c is the data]). Within four weeks, body weights were significantly different from those of lean-diet mice and continued to rise over 16 weeks, accompanied by elevated resting glucose levels (Fig. 1c [1d is the glucose, 1c is body weight]), indicating insulin resistance-induced hyperglycemia. Glucose tolerance tests showed impaired glucose clearance in HFHS-fed mice compared to lean-fed mice (Fig.1e), confirming the establishment of a type 2 diabetes phenotype in HFHS diet mice” [lacking statistical comparison].

e. Fig 1 i and j are not mentioned in the text.

f. “First, we focused on microglia to identify transcriptional changes that may underlie the observed Trem2 alterations, despite the equal number of microglial nuclei in the DKI-Lean and DKI-HFHS groups (Extended Data Fig. C).” Which Figure? Ext Fig 6C? I also suggest referencing Ext Fig. 6C when first mentioning the number of glial population frequencies for clarity.

g. “We next cross-referenced gene expression within the MG3 cluster against a curated list of microglial state markers33, which were stratified by WT and DKI groups based on disease-associated activation (Extended Data Fig. d).”

Please specify the figure number.

3. In different experiments, either APPNL-F mice or APPNL-G-F mice were chosen to combine with the MapthMAPT mice. For clarity, a sentence mentioning the rationale for their mouse model choice would be helpful.

4. When the authors discuss the MapMyCell results (Page4), the referenced supplemental table (supplemental table 1) is incorrect. Please double check the reference and link the correct table.

5. When showing DEG numbers (where pink refers to upregulated genes) and in heatmaps (where pink refers to downregulation), it would be more consistent if the authors could use the same color scheme throughout.

6. There is no mention of whether the experimenter is blinded to the genotype when performing the Sholl analysis in Fig. 2d-e; Extended Data Fig. 3a.

7. Moreover, the snRNAseq data argue against Trem2 being different due to the diet (The MG3 is low in Trem2 expression). Could the authors elaborate further if they believe the microglial morphology and the shift in Trem2 location are independent of the Trem2 level?

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.