Altered hepatic metabolism mediates sepsis preventive effects of reduced glucose supply in infected preterm newborns

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors follow up on their published observation that providing a lower glucose parental nutrition (PN) reduces sepsis from a common pathogen [Staphylococcus epidermitis (SE)] in preterm piglets. Here they found that a higher dose of glucose could thread the needle and get the protective effects of low glucose without incurring significant hypoglycemia. They then investigate whether the change in low glucose PN impacts metabolism to confer this benefit. The finding that lower glucose reduces sepsis is important as sepsis is a major cause of morbidity and mortality in preterm infants, and adjusting PN composition is a feasible intervention.

Strengths:

(1) They address a highly significant problem of neonatal sepsis in preterm infants using a preterm piglet model.

(2) They have compelling data in this paper (and in a previous publication, ref 27) that low glucose PN confers a survival advantage. A downside of the low glucose PN is hypoglycemia which they mitigate in this paper by using a slightly high amount of glucose in the PN.

(3) The experiment where they change PN from high to low glucose after infection is very important to determine if this approach might be used clinically. Unfortunately, this did not show an ability to reduce sepsis risk with this approach. Perhaps this is due to the much lower mortality in the high glucose group (~20% vs 87% in the first figure).

(4) They produce an impressive multiomics data set from this model of preterm piglet sepsis which is likely to provide additional insights into the pathogenesis of preterm neonatal sepsis.

Weaknesses:

(1) The high glucose control gives very high blood glucose levels (Figure 1C). Is this the best control for typical PN and glucose control in preterm neonates? Is the finding that low glucose is protective or high glucose is a risk factor for sepsis?

(2) In Figure 1B, preterm piglets provided the high glucose PN have 13% survival while preterm piglets on the same nutrition in Figure 6B have ~80% survival. Were the conditions indeed the same? If so, this indicates a large amount of variation in the outcome of this model from experiment to experiment.

(3) Piglets on the low glucose PN had consistently lower density of SE (~1 log) across all time points. This may be due to changes in immune response leading to better clearance or it could be due to slower growth in a lower glucose environment.

(4) Many differences in the different omics (transcriptomics, metabolomics, proteomics) were identified in the SE-LOW vs SE-HIGH comparison. Since the bacterial load is very different between these conditions, could the changes be due to bacterial load rather than metabolic reprogramming from the low glucose PN?

Reviewer #2 (Public Review):

Summary:

The authors demonstrate that a low parenteral glucose regimen can lead to improved bacterial clearance and survival from Staph epi sepsis in newborn pigs without inducing hypoglycemia, as compared to a high glucose regimen. Using RNA-seq, metabolomic, and proteomic data, the authors conclude that this is primarily mediated by altered hepatic metabolism.

Strengths:

Well-defined controls for every time point, with multiple time points and biological replicates.

The authors used different experimental strategies to arrive at the same conclusion, which lends credibility to their findings.

The authors have published the negative findings associated with their study, including the inability to reverse sepsis-related mortality after switching from SE-high to SE-low at 3h or 6h and after administration of hIAIP.

Weaknesses:

(1) The authors mention, and it is well-known, that Staph epi is primarily involved in late-onset sepsis. The model of S. epi sepsis used in this study clearly replicates early-onset sepsis, but S. epi is extremely rare in this time period. How do the authors justify the clinical relevance of this model?

(2) The authors find that the neutrophil subset of the leukocyte population is diminished significantly in the SE-low and SE-high populations. However, they conclude on page 10 that "modulations of hepatic, but not circulating immune cell metabolism, by reduced glucose supply..." and this is possible because the authors have looked at the entire leukocyte transcriptome. I am curious about why the authors did not sequence the neutrophil-specific transcriptome.

(3) The authors use high (30g/k/d) and low (7.2g/k/d) glucose regimens. These translate into a GIR of 21 and 5 mg/k/min respectively. A normal GIR for a preterm infant is usually 5-8, and sometimes up to 10. Do the authors have a "safe GIR" or a threshold they think we cannot cross? Maybe a point where the metabolism switch takes place? They do not comment on this, especially as GIR and glucose levels are continuous variables and not categorical.

(4) In Figures 2B and C the authors show that SE-high and SE-low animals have differences in the oxphos, TCA, and glycolytic pathways. The authors themselves comment in the Supplementary Table S1B, E-F that these same metabolic pathways are also different in the Con-Low and Con-high animals, it is just the inflammatory pathways that are not different in the non-infected animals. How can they then justify that it is these metabolic pathways specifically which lead to altered inflammatory pathways, and not just the presence of infection along with some other unfound mechanism?

(5) The authors mention in Figure 1F that SE-low animals had lower bacterial burdens than SE-high animals, but then go on to infer that the inflammatory cytokine differences are attributed to a rewiring of the immune response. However, they have not normalized the cytokine levels to the bacterial loads, as the differences in the cytokines might be attributed purely to a difference in bacterial proliferation/clearing.

(6) The authors mention that switching from SE-high to SE-low at 3 or 6 h time points does not reduce mortality. Have the authors considered the reverse? Does hyperglycemia after euglycemia initially, worsen mortality? That would really conclude that there is some metabolic reprogramming happening at the very onset of sepsis and it is a lost battle after that.

Reviewer #3 (Public Review):

Summary:

Baek and colleagues present important follow-up work on the role of serum glucose in the management of neonatal sepsis. The authors previously showed high glucose administration exacerbated neonatal sepsis, while strict glucose control improved outcomes but caused hypoglycemia. In the current report they examined the effect of a more tailored glucose management approach on outcomes and examined hepatic gene expression, plasma metabolome/proteome, blood transcriptome, as well as the the therapeutic impact of hIAIP. The authors leverage multiple powerful approaches to provide robust descriptive accounts of the physiologic changes that occur with this model of sepsis in these various conditions.

Strengths:

(1) Use of preterm piglet model.

(2) Robust, multi-pronged approach to address both hepatic and systemic implications of sepsis and glucose management.

(3) Trial of therapeutic intervention - glucose management (Figure 6), hIAIP (Figure 7).

Weaknesses:

(1) The translational role of the model is in question. CONS is rarely if ever a cause of EOS in preterm neonates. The model. uses preterm pigs exposed at 2 hours of age. This model most likely replicates EOS.

(2) Throughout the manuscript it is difficult to tell from which animals the data are derived. Given the ~90% mortality in the experimental CONS group, and 25% mortality in the intervention group, how are the data from animals "at euthanasia" considered? Meaning - are data from survivors and those euthanized grouped together? This should be clarified as biologically these may be very different populations (ie, natural survivor vs death).

(3) With limited time points (at euthanasia ) for hepatic transcriptomics (Figure 2), plasma metabolite (Figure 3) blood transcriptome (Figure 4), and plasma proteome (Figure 5) it is difficult to make conclusions regarding mechanisms preceding euthanasia. Per methods, animals were euthanized with acidosis or clinical decompensation. Are the reported findings demonstrative of end-organ failure and deterioration leading to death, or reflective of events prior?

(4) Data are descriptive without corresponding "omics" from interventions (glucose management and/or hIAIP) or at least targeted assessment of key differences.

Author response:

We thank the reviewers for their thorough comments on our manuscript. We appreciate their recognition of the strengths in our study, including addressing the significant problem of neonatal sepsis in preterm infants using a preterm piglet model, the robustness of our multi-omics dataset, and our multi-pronged approach to examining the physiological changes under different glucose management regimens.

This document addresses our initial responses to the main concerns of the 3 reviewers. We will provide more detailed responses to their comments and revise the manuscript at a later date.

In response to Reviewer #1, we acknowledge the concern about high blood glucose levels in the control group. This work is a follow-up from our previous work (Muk et al, JCI insight 2022) where we explored different PN glucose regimens. Taken together, our experiments suggest a linear relationship between glucose provision and infection severity, indicating increased glucose may heighten mortality risk, while radical reduction could reduce mortality due to sepsis, but cause hypoglycemia and brain damage. As for the discrepancy in survival rates between Figures 1B and 6B, this is due to a shortened follow-up time in the follow-up experiment. This was done to minimize animal suffering because relevant differences in immune-responses were detectable within 12 hours in the primary experiment. As for the relationship between bacterial burdens and glucose, we agree that lower bacterial density in piglets receiving the reduced glucose PN may result from slower bacterial growth. However, we analyzed the relationship between bacterial burdens and mortality and found that it did not correlate within each of the treatment groups. This finding inspired us to further explore the relationship between bacterial burdens and infection responses in our model which has resulted in our recent preprint: Wu et at. Regulation of host metabolism and defense strategies to survive neonatal infection. BioRxiv 2024.02.23.581534; doi: https://doi.org/10.1101/2024.02.23.581534

For Reviewer #2, The distinction between early (EOS) and late onset sepsis (LOS) in the time cut-off makes sense clinically because they are likely to be caused by different organisms and origins (EOS with maternal origin and LOS with postnatal origin) and therefore require different empirical antibiotics regimes. However, it is also important to acknowledge that the pathophysiology of “sepsis” may be similar despite timing and pathogen and depends on the degree of immune activation. Therefore, even though the infection in our model is initiated on the first day after birth the organism that we use, Staphylococcus epidermidis (most common bacteria detected in LOS), makes it a better model for LOS. As for neutrophil specific transcripts, we only collected the whole blood transcript during the experiments, which reflects the transcriptomic profile of all the leucocytes. Since we did not do single cell RNA sequencing during the experiment there is no possibility of isolating the neutrophil transcriptome at this time. As for the question of a “safe glucose infusion rate”, there likely is none as the immune responses to glucose intake do not seem binary but increase with glucose intake. Our reduced glucose PN was chosen as it corresponded with the low end of recommended guidelines for PN glucose intake. However, the reduced glucose intervention still resulted in significant morbidity and a 25% mortality within 22 hours. There is therefore still vast room for improvement, but even though further reduction in PN glucose intake would probably provide further protection, it would entail dangerous hypoglycemia. The findings in this paper have prompted us to explore several alternative strategies to both reduce infection-related mortality and maintain glucose homeostasis. Thus, the optimal PN for infected newborns would probably differ from standard PN in all macronutrients compartments and will require much more pre- and clinical research.

For Reviewer #3, we acknowledge the variability in data collected from animals at euthanasia. These endpoints represent snapshots of the animals' states at euthanasia, which is a clear limitation of our method. Therefore, we do not know what metabolic processes precede the development of lethal sepsis, although the increases in plasma lactate suggest a higher rate of glycolysis in animals on high glucose PN. However, we believe the data still heavily imply a causal relationship between energy metabolic processes, especially glycolytic breakdown of glucose, and the pro-inflammatory responses leading to sepsis. In our recent preprint mentioned above we further explored the metabolic responses in pigs that succumbed to sepsis, compared to those that survived and found that survival was strongly associated with increases in mitochondrial metabolism and reduction in glycolysis.

We hope these clarifications and our commitment to further research address your concerns satisfactorily. Thank you for your valuable feedback.