December 2021

Liver metastasis limits the efficacy of immune checkpoint blockade, in part through sequestration and elimination of tumor antigen-specific T cells by hepatic macrophages. This negative impact of liver metastasis can be reversed in animal models through macrophage depletion and radiotherapy.

Summary:
The liver is one of the most common sites of metastasis and is specially equipped to promote immune tolerance in the context of autoimmune diseases, viral infections and organ transplantation. However, the importance of the liver’s ability to promote immune tolerance in the context of cancer is poorly understood. This is especially true for immunotherapy which is highly reliant on tumor infiltrating CD8+ T cells and is FDA approved for many solid tumors. Using both retrospective patient data and animal models, Yu et al explored the role of liver metastasis in systemic immunity and immunotherapy efficacy.

First, the authors used retrospective cohorts to demonstrate that patients with liver metastasis treated with checkpoint inhibitors had worse outcomes compared to patients with metastasis to other sites. In a melanoma cohort, patients with liver metastasis treated with checkpoint blockade had a 40-month overall survival of ~20% compared to ~62% for patients with other sites of metastasis (P<0.0001). These findings remained significant even after controlling for other important clinicopathologic variables such as age, gender, prior therapy, PD-L1 expression and tumor burden.

Next, the authors used preclinical models to further explore the mechanisms underlying the findings in patient cohorts. They first demonstrate that SubQ MC38 flank tumors (a mouse colorectal cancer cell line) will regress after anti-PD-L1 treatment (IgG ~0.3g vs anti-PD-L1 ~0.1g, P=0.0013). However, if the mice have concurrent MC38 tumors in their livers, then anti-PD-L1 therapy loses its efficacy (IgG ~0.35 vs anti-PD-L1 ~0.3, P=0.75). Interestingly, anti-PD-L1 remains effective for mice concurrently bearing MC38 tumors in both their lungs and SubQ tissue. These findings were also true for a mouse melanoma cell line (B16F10). Finally, the authors show that their phenotype is rescued and anti-PD-L1 regains effectiveness in mice bearing SubQ B16F10 tumors with MC38 liver tumors and vice-versa, suggesting that liver metastasis abolishes systemic immunotherapy efficacy in a tumor-specific manner.

Using flow cytometry and mass spectrometry, the authors then show that mice harboring SubQ tumors with liver metastases sequester tumor antigen-specific T cells in the liver. This finding was not true for mice with tumors in their lungs or SubQ tissue alone. Interestingly, the tumor antigen-specific T cells compartmentalized in the liver had significantly higher rates of cleaved caspase-3 and other markers of apoptosis, suggesting that liver metastasis uniquely siphons activated antigen specific T cells and subsequently causes T cell deletion. Importantly, the authors noted an increase in liver CD11b+, F4/80+ macrophages in mice bearing liver tumors. Depleting macrophages with clodronate and anti-CSF-1 diminished T cell apoptosis and restored anti-PD-L1 efficacy. These findings suggested that the loss of anti-PD-L1 therapy in mice with liver tumors was due to hepatic macrophages inducing T cell apoptosis.

Myeloid cells and macrophages can induce apoptosis through direct cell contact, often via a Fas-FasL mechanism. Flow cytometry results showed that hepatic macrophages expressed MHC-I presenting tumor antigens as well as high levels of FasL. Ex-vivo experiments showed that hepatic isolated CD11b+F4/80+ macrophages could induce apoptosis in co-cultured tumor antigen-specific T cells through a Fas-FasL mechanism. These experiments were supported by single cell RNA-seq data that showed an increased proportion of immunosuppressive macrophages and reductions in T cells in livers harboring metastases. CD8 T cell clusters also showed higher expression of apoptosis markers in mice with liver tumors compared to SubQ tumors alone.

Radiotherapy is known to reshape the tumor microenvironment, and the authors demonstrate that its use in their animal model reduced liver myeloid cell number and increased expression of effector CD8 T cell markers (IFNg and granzyme B). The authors go on to show that mice with MC38 SubQ and liver tumors treated with anti-PD-L1 and radiotherapy had significantly decreased tumor volume and prolonged overall survival compared to mice treated with monotherapy alone. These results suggest that radiotherapy can alter the liver tumor microenvironment to improve the efficacy of systemic immunotherapy.  

This prospective study did not find an improved quality of life for patients undergoing early closure of temporary ileostomy, but rather demonstrated an increase in adverse outcomes, and as such was terminated early. However, there are limitations in the study design which could explain these results. 

Summary:
It has long been established that creation of a temporary diverting stoma, typically with diverting loop ileostomy (DLI), is an effective method to protect low colorectal anastomoses following surgical resection for rectal cancer. The presence of a stoma typically has a significant negative impact on a patient’s quality of life. As such, early closure of DLI (8-15 days) has been proposed as an alternative to the typical stoma closure timeline, which typically occurs after 12 weeks. Multiple previous randomized controlled studies have demonstrated safety in early stoma closure. This multicenter, prospective, and randomized study aimed to evaluate early vs. late closure (2 vs. 12 weeks) on quality of life, intra-operative feasibility, and post-operative morbidity and mortality. 

This study utilized patients >18 years of age who underwent LAR between 2007 and 2014 at three different surgical sites, and were randomized to early vs. late ileostomy closure. The study was terminated early after 71 patients (37 vs. 34) were accumulated due to safety concerns in the early closure group. Specifically, patients who had an early ileostomy closure had significantly more intraoperative bleeding (p=0.011), more adhesions (p=0.034), higher rates of colorectal anastomotic leakage (19% vs. 0%, p=0.012), and a higher reoperation rate (16% vs. 0%, p=0.026). Thus, it was concluded that early stoma closure should be avoided.

Six previous studies demonstrated safety in early ileostomy closure, so this study begs the question of why this group showed differing results? One major limitation of this study which could explain the high leak rate of 19% is that the integrity of the colorectal anastomosis with a contrast enema was not evaluated prior to randomization and stoma reversal, whereas in previous trials only patients with a negative contrast enema were included in the randomization. Additionally, this study’s choice of 2 weeks (14 days) as “early closure” differs in the 8-15 days chosen in 5 out of the 6 previous trials. More adhesions are likely to form by 14 days as compared to 8 days, which could explain the higher rates of intraoperative bleeding and adhesions seen in the early reversal group in this study.

In conclusion, the results of this study published in Diseases of the Colon and Rectum should not dissuade surgeons from consideration of early stoma closure in appropriate patients, but rather should highlight the importance of patient selection and operative timing. For patients with no anastomotic leak visualized on contrast study, early stoma reversal 8 days following surgery for rectal cancer is a safe option, as demonstrated by multiple previous randomized studies.  

Non-trauma patients who underwent damage-control laparotomy were more likely to develop post operative abdominal sepsis and a need for dialysis but less likely to develop delirium. More work needs to be done examining the outcomes of damage-control laparotomy versus initial closure within the non-trauma emergency surgery population.

Summary:
In this study, the authors present data from an EAST multicenter trial in which outcomes were compared between patients with open abdomens for trauma vs non-trauma indications. Patients who underwent damage-control laparotomy (DCL) were enrolled from 15 centers and data was gathered retrospectively over a 2-year period spanning 2017-2018. Data on patients’ underlying comorbidities, demographics, and injury severity were recorded, as were amount/duration of sedating agents and standardized measures of delirium. Primary endpoints were mortality and proportion of ICU days free of delirium and coma. This was calculated by tabulating the number of days with a RASS>=-3 and a negative CAM-ICU , then dividing this by the total number of ICU days (censoring at 30 days).

The number of trauma patients was nearly triple the number of non-trauma patients in this study (411 vs 143). Non-trauma patients were older, more female, and with more comorbidities. There were no differences in number of takebacks, time to first takeback, and time to fascial closure, among other measures of surgical management. The non-trauma group did suffer a higher incidence of intra-abdominal sepsis and need for dialysis. On the other hand, the trauma group had a lower proportion of coma- and delirium-free days (51.1% vs 73.7%, p=0.029). This increased incidence of delirium persisted in a multivariable regression analysis. There was no difference in mortality.

This is an interesting question, but I think that more than a comparison between trauma and non-trauma DCL patients, the better question is how emergency general surgery (EGS) patients undergoing DCL fare compared to those who are closed at the index operation. While the use of DCL is well-established and backed by evidence in the trauma population, this is less true for EGS. Furthermore, a comparison between EGS and trauma patients is a comparison between apples and oranges. As the authors note, EGS patients often suffer from an underlying problem that is infectious in nature, as opposed to the trauma population, in whom the etiology of shock is often hemorrhagic. This may explain the higher incidence of abdominal sepsis in the EGS group in this study. The increased incidence of kidney injury in the EGS group may be related to the increased age and comorbidities in this group. Nonetheless, this is an interesting contribution to the literature on damage control laparotomy, particularly with respect to the under-studied non-trauma population.

The FOXO class of transcription factors seem to be a dominant driver of stress-induced hyperglycemia in a mouse model of trauma and hemorrhage, potentially involving cross-talk between liver and adipose tissue. This highlights a novel mechanism that could potentially be utilized to combat the acute hyperglycemia and insulin resistance commonly seen in the stress-state of surgical patients.

Summary:
Stress-induced hyperglycemia is a common metabolic derangement found in a wide variety of critical illnesses and injury, including trauma and even post-operative states. More than just a clinical marker of stress severity, there is ample data showing control of hyperglycemia leads to improved morbidity and mortality outcomes in trauma as well as surgical patients. In this basic science paper authored by Dr. Anna Whitlock and the Titchenell lab, the role of forkhead box O (FOXO) transcription factors, which have been previously implicated in chronic insulin resistance, was examined in the setting of stress-induced hyperglycemia.

One of the underappreciated aspects of basic science is understanding the models on which experiments are performed and conclusions are made. In this manuscript it is worth pointing out the mice used and the model of inducing stress-induced hyperglycemia. Two specific mice were used – L-FOXO1 (which can be induced to have liver specific knockout of FOXO1) and L-FOXOTKO (which can be induced to have liver specific knockout of FOXO1, FOXO3, and FOXO4). To reliably and consistently simulate stress-induced hyperglycemia, an established model of trauma and hemorrhage was utilized whereby a laparotomy is made and the bilateral femoral arteries were identified and catheterized to allow “controlled hemorrhage” to a set meant arterial pressure.

The first thing the group demonstrated was the effect of trauma and hemorrhage model on glucose and insulin signaling in normal mice. Both insulin and glucose blood levels, as expected were higher in the trauma and hemorrhage mice compared to both baseline and surgery alone (without the trauma and hemorrhage). Furthermore, IGFBP-1, a transcriptional target of FOXO1, was shown to be increased in the liver by both gene expression and protein analysis. When performing this experiment in L-FOXO1KO mice, which have no FOXO1 expression, hyperinsulinemia, but not hyperglycemia, was reduced. Finally, the same conditions in L-FOXOTKO mice, which have no FOXO1, FOXO3, or FOXO4 expression, reduced both hyperglycemia and hyperinsulinemia. This reduction in hyperglycemia in this mouse was also associated with an inhibition of peripheral lipolysis, which is usually increased in states with increased insulin resistance such as type II diabetes and stress-induced hyperglycemia.

Overall, this work supports the increased scrutiny of FOXO transcription factors as well as adipocyte lipolysis as a means of combatting stress-induced hyperglycemia. While more research is required to translate this into the clinical setting, this is undoubtedly a good initial step in combatting a problem that involves all surgery patients across all surgical specialties.