Prehabilitation during neoadjuvant chemotherapy results in an enhanced immune response in esophageal adenocarcinoma tumors: A randomized controlled trial

Highlights

• •16 weeks of prehabilitation exercise training during and after neoadjuvant chemotherapy for esophageal adenocarcinoma maintained relative peak cardiorespiratory fitness and significantly increased fitness before surgery compared to controls.
• •Prehabilitation exercise was associated with an altered tumor microenvironment and enhanced immune infiltration. Specifically, larger increases in relative peak cardiorespiratory fitness during neoadjuvant chemotherapy were associated with higher levels of CD8+ tumor infiltrating lymphocytes (TILs).
• •As well as more CD8+ TILs, exercise training was associated with higher levels of CD56+ NK TILs, suggesting an enhanced innate immune response to exercise.
• •Tertiary lymphoid structures that house dendritic cells and T cells showed an increased cell density (i.e., more cells per surface area) in the exercise group.

Abstract

Background

For patients with locally advanced esophagogastric cancer, the standard of care in the UK is neoadjuvant chemotherapy (NAC) followed by surgery. Prehabilitation exercise training can improve physiological function and fitness. If such improvements translate to increased immune infiltration of tumors, exercise could be prescribed as an immune adjuvant during NAC and potentially improve clinical outcomes. As such, we aimed to determine whether prehabilitation increased tumor infiltrating lymphocytes (TILs).

Methods

We assessed 22 patients with locally advanced esophageal cancer on a randomized control trial comparing 16 weeks of low-to-moderate intensity twice weekly supervised and thrice weekly home-based exercise (Prehab: n = 11) to no prehabilitation (Control: n = 11). Our primary outcome was to compare tumor-immune responses between Controls and Prehab. We compared formalin-fixed paraffin-embedded tumors by high-resolution multispectral immunohistochemistry (mIHC) and NanoString spatial transcriptomics. Secondarily, we determined relationships between changes in fitness to the exercise training and tumor-immune measures. Specifically, we assessed percentage changes in peak cardiorespiratory fitness as assessed by peak oxygen uptake (V̇O_2peak) before NAC (Baseline) and after 8 weeks of NAC (Post-NAC), and changes between Baseline and following 8 weeks of NAC recovery before surgery (Pre-surgery) and correlated changes in fitness with tumor-immune responses. Finally, as an exploratory aim, we assessed clinical outcomes between groups, including survival, therapy tolerance, and tumor regrading.

Results

We observed that Prehab had significantly more CD8+ lymphocytes in their tumors (mean difference (diff.) = 1.79, 95% confidence interval (95%CI): 0.76‒2.82, p < 0.001) and their stroma (mean diff. = 1.59, 95%CI: 0.66‒2.52, p < 0.001) than the Controls. When normalized to total numbers of TILs, Prehab had higher levels of CD56+ natural killer (NK) cells (median diff. = 0.87, 95%CI: 0.25‒2.18), p = 0.0274), consisting primarily of CD56^dim NK cells (median diff. = 0.48, 95%CI: 0.03‒2.53), p = 0.0464). Evaluation of the presence and localization of tumor-associated tertiary lymphoid structures (TLS) in the esophageal tumors revealed that most TLS were in the peritumoral regions. Prehab had a higher TLS cell density (cells/mm²; median diff. = 18,959, 95%CI: 13,518‒22,635), p < 0.001) and more clearly defined germinal centers indicative of mature TLS visually. We observed that Prehab maintained their V̇O_2peak during NAC while the Controls’ V̇O_2peak reduced by 9.0% ± 10.2% (mean ± SD) (Post-NAC: p = 0.018). Pre-surgery, Prehab V̇O_2peak was a clinically meaningful 3.27 ± 1.31 mL/kg/min higher than Controls (p = 0.022). Between Baseline and Post-NAC, where the Prehab maintained V̇O_2peak better than Controls, there were significant positive associations with percentage changes in V̇O_2peak and the frequencies of CD8+ TILs (r = 0.531, p = 0.016), programmed death-ligand 1+ (PDL1+) cells (r = 0.566, p = 0.009), and granzyme B+ (GrzB+) TILs (r = 0.582, p = 0.007). Similar relationships were observed for changes in V̇O_2peak from Baseline to Pre-Surgery only in the Prehab group. We observed no differences between groups regarding clinical outcomes such as survival, therapy tolerance, or tumor regrading.

Conclusion

We show that exercise training during NAC, which promotes higher levels of cardiorespiratory fitness than no exercise, is associated with increased frequencies of TILs and maturity of TLS. These data suggest that exercise during NAC enhances the immune system. Future studies are warranted to understand the clinical consequences of this.

Graphical abstract

1. Introduction

For patients in the UK (but not all countries) with locally advanced esophagogastric adenocarcinoma, the standard of care is neoadjuvant chemotherapy followed by surgery. Perioperative stress is associated with significant morbidity accentuated by reductions in functional capacity, muscle mass, and cardiopulmonary fitness.¹ Evidence is accumulating that prehabilitation exercise training is an effective strategy for preventing this loss in muscle mass and cardiopulmonary fitness.^2,3 Our randomized controlled trial (RCT), which this secondary analysis stems from, showed that prehabilitation maintained cardiopulmonary fitness better than no exercise but did not affect the primary outcome of anaerobic threshold.²

Although not yet a standard of care, a recent non-randomized trial suggests that exercise training may promote better responses to chemotherapy and enhanced circulating blood immune responses in esophageal cancer.³ However, whether this facilitates more significant esophagogastric tumor regression remains unclear. These findings add to the growing body of literature showing the benefits of exercise in patients with cancer, including improved quality of life, reduced fatigue, and therapy completion.^4,5 However, the mechanisms of exercise-induced benefits, such as tumor reduction or reduced recurrence, remain unclear.

Recent preclinical studies have reported that exercise training reduces tumor burden in several solid tumors, including melanoma, lung, and liver cancers.^6,7 Exercise training increases cancer cell apoptosis, reduces intratumoral hypoxia levels, and increases the frequency of tumor-infiltrating lymphocytes (TILs).^6,7 Increased TILs (specifically CD56+ natural killer (NK) cells and CD8+ T cells) appear to be mediated by exercise-induced release of muscle-derived cytokines (i.e., myokines) and their adrenaline-mediated mobilization into the blood, which allows them to traffic to tumors.⁶ However, in humans, few studies have assessed exercise-induced TIL responses.^8,9 In adults with localized prostate cancer who did not receive chemotherapy, neither an acute bout of exercise⁸ nor 4–30 total sessions of exercise training⁹ before prostatectomy resulted in higher frequencies of TILs compared to controls. These studies suggest that short-duration exercise may not be sufficient to develop a TIL response. Alternatively, it is plausible that chemotherapy is required to facilitate a rapid immune response by facilitating tumor damage, which stimulates an immunological response.³ More recently, in patients at a very high risk of colorectal and endometrial cancer, 12 months of exercise training increased the frequency of NK cells and CD8+ T cells in the colon mucosa associated with tumor development.¹⁰ Subsequently, to the best of our knowledge, there is no evidence in humans of the TIL response to exercise training during neoadjuvant chemotherapy.

Therefore, this exploratory secondary analysis of our previously conducted RCT aimed to evaluate the TIL response to 16 weeks of prehabilitation exercise training compared to no exercise training in locally advanced esophageal adenocarcinoma. Additionally, we aimed to determine possible mechanisms by assessing differences in cell phenotype, spatial transcriptomics, and gene expression within the tumor microenvironment. Finally, we wanted to determine relationships between exercise-induced physiological responses and levels of TILs.

2. Materials and methods

This secondary analysis of our recently completed parallel-group RCT compares prehabilitation exercise training with standard of care during neoadjuvant chemotherapy (NAC) before esophagogastric surgery. Patient recruitment, study design, intervention details, and endpoints are detailed elsewhere.² Briefly, patients were recruited from The Royal Surrey NHS Foundation Trust, a tertiary referral center, where 153 patients were screened and 54 provided informed written consent, of which 48 had adenocarcinoma. A total of 48 patients (24 Controls and 24 Prehabilitation) with locally advanced esophagogastric cancer planned for neoadjuvant therapy plus esophagogastrostomy or total gastrectomy completed the study. We randomly chose 22 patients with adenocarcinoma for this sub-study using a random number generator. Our primary trial and protocol were registered on ClinicalTrials.gov (NCT02950324.) and the study was approved by the London-Bromley Research Ethics Committee, meeting guidelines of the NHS Health Research Authority (16 November, 2016).

As such, this secondary analysis consists of a subgroup of 22/48 patients who completed the study, and our work has been reported in line with the Consolidated Standards of Reporting Trials (CONSORT) criteria (Supplementary Fig. 1).¹¹ Inclusion criteria for this cohort were those patients who had undergone esophagogastric resection with a histological subtype of esophageal adenocarcinoma and had Siewert classification I or II tumors. Given the complexity of Siewert III tumors and that less than 6% of the primary study patients had Siewert III tumors, this was deemed a reasonable inclusion criteria. Although the demographics between this cohort and the main study are similar (data not shown), we have detailed the demographics and relevant outcome measures for this specific cohort (Table 1 and Fig. 1). All clinical data, including survival, tumor grading, and chemotherapy responses were collected from medical records accessed by the clinical members of the research team.

Table 1.

Baseline demographics of trial patients between July 2016 and July 2019.

	Overall (n = 22)	Prehab (n = 11)	Control (n = 11)
Demographics & fitness
Age (year)	62.8 ± 12.8	62.0 ± 6.4	63.6 ± 7.2
Sex (male/female)	20/2	10/1	10/1
Weight (kg)	88.9 ± 12.8	86.7 ± 11.3	91.0 ± 14.3
BMI (kg/m2)	28.7 ± 4.3	28.5 ± 4.5	29.0 ± 4.5
Overweight/obese	18 (82)	10 (91)	9 (82)
Smoking status
Current	4 (18)	2 (18)	2 (18)
Ex-smoker	12 (55)	5 (45)	7 (64)
Non	6 (27)	4 (36)	2 (18)
Cancer characteristics
Adenocarcinoma	22 (100)	11 (100)	11 (100)
Neoadjuvant chemotherapy
ECX	18 (82)	10 (91)	8 (73)
FLOT	4 (18)	1 (9)	3 (27)
Aerobic fitness (V̇O2peak)
Relative (mL/kg/min)	20.4 ± 3.3	21.0 ± 3.8	19.8 ± 3.0
VO2 as a % of predicted (%)a	83.8 ± 18.6	85.3 ± 23.0	82.2 ± 13.9
Absolute (L/min)	1.8 ± 0.3	1.8 ± 0.4	1.8 ± 0.3
AT (mL/kg/min)	13.4 ± 2.8	13.5 ± 2.9	13.3 ± 2.9

Notes: Data are mean ± SD or n (%) unless otherwise indicated. Percentages might not add up to 100% due to rounding.

Abbreviations: AT = anaerobic threshold; BMI = body mass index; ECX = epirubicin, cisplatin, capecitabine; FLOT = fluorouracil, leucovorin, oxaliplatin, docetaxel; VO₂ = oxygen consumption; V̇O_2peak = peak oxygen uptake.

^aPredicted values were calculated using the Wasserman/Hansen calculation, which accounts for weight, height, and age.

Fig. 1.

Prehabilitation exercise maintains cardiopulmonary fitness through neoadjuvant chemotherapy. (A) V̇O_2peak relative to body mass; (B) absolute V̇O_2peak; (C) oxygen consumption at anaerobic threshold. Data are means with standard deviation error bars. * p < 0.05 different from Baseline; ^# p < 0.05 different from Post-NAC; ^§ p < 0.05 different from Control. NAC = neoadjuvant chemotherapy; V̇O_2peak = peak oxygen uptake.

2.1. Physical testing

Before starting NAC (Baseline), patients completed a cardiopulmonary exercise test (CPET). The CPET involved a graded incremental cycling test until volitional exhaustion, with starting workloads determined by the clinical exercise team to ensure the test lasted no longer than 15 min. During this, we determined peak cardiorespiratory fitness via V̇O_2peak using a graded cycling test to maximal exhaustion or symptom-limited discontinuation. We assessed breath-by-breath analysis and a 12-lead electrocardiogram (ECG) to assess heart function (Ergoline; LoveMedical, Manchester, UK). All patients completed a valid peak test as indicated by a respiratory exchange ratio of over 1.1 and a peak heart rate (HR) of less than 5% of age-predicted maximum. Patients completed the same CPET 2 weeks after completion of NAC (Post-NAC) and 2‒3 days before surgery (Pre-surgery).

2.2. Intervention

Stage 1 of prehabilitation began at the commencement of NAC and involved 8 weeks of a tailored program based on participants’ baseline CPET. Exercise intensities were calculated using V̇O_2reserve (V̇O_2reserve = V̇O_2peak –V̇O_2rest) and monitored using heart rate reserve (HRR).¹² We calculated HR zones that corresponded with the percentage of V̇O_2peak by ((HR_peak –HR_rest) × intensity percentage) + HR_rest), which has been shown to correlate highly with VO_2reserve. The program consisted of twice-weekly, 1-h supervised sessions of low-to-moderate intensity aerobic exercises (40%–60% HRR), free-weight and resistance band strengthening exercises, and flexibility exercises to develop a range of motion capabilities. Aerobic exercise consisted of 25 min of cycling (Ergoline; LoveMedical) with increasing electromagnetic resistance levels to achieve the required target percentage of HRR. Following a 5 min “very light” intensity warm up (BORG rating of perceived exertion (RPE) 9/20), training commenced for 20 min at intensities that increased from 40% to 60% HRR (RPE 11-14/20; “fairly light” to “somewhat hard/hard”) across the 15 weeks, as tolerated by the participant. Resistance exercises focused on 6 key muscle groups, with sessions involving 2 sets of 12 repetitions and aiming for an RPE of 12–14 (“somewhat hard”) using resistance bands and free weights. These exercises were accompanied by dynamic stretching exercises to improve flexibility and maintain range of motion. Additionally, patients completed 3 × 1-h home-based resistance and core stability exercises. After completion of NAC, patients’ V̇O_2peak and training zones were re-calculated, and they completed a similar 7- or 8-week (depending on surgical availability) exercise program before completing a final bicycle CPET in the few days before surgery (pre-surgery). We provided the Control group with physical activity guidance as per the standard esophagogastric pathway at that time. We gave all patients a Fitbit Flex 2 activity monitor (Fitbit Limited, London, UK) to assess physical activity levels during the study.

2.3. Tumor sampling/processing

We fixed surgically resected esophagogastric cancer specimens in neutral-buffered formalin (NBF), and formalin-fixed paraffin-embedded (FFPE) sample blocks were prepared. From each block, 4-µm serial tissue sections were prepared and placed on FLEX immunohistochemistry (IHC)-coated Microscope Slides (DAKO, Glostrup, Denmark). For all tumor sampling and analyses, researchers were blinded to patient grouping. Additionally, multispectral IHC (mIHC) and NanoString were conducted from samples acquired from the same FFPE block to ensure they were closely localized.

2.4. Fluorescent mIHC

We applied optimized fluorescent mIHC 8-plex panels plus the nuclear marker 4’, 6-diamidino-2-phenylindole, dihydrochloride (DAPI) to the FFPE tissues to simultaneously detect multiple markers on a single section. We designed panels (Supplementary Table 1) to (a) characterize generically important immune cells and factors that are known to influence the progression of solid tumors and (b) detect the cell types characteristic of tertiary lymphoid structures (TLS). Panel 1 consisted of primary mouse anti-human antibodies for CD8 (clone 4B11; Bio-Rad, Hercules, CA, USA), CD4 (clone 4B12; Leica Biosystems, Nussloch, Germany), CD68 (clone PG-M1; DAKO), Granzyme B (clone 11F1; Leica Biosystems), PDL1 (clone E1L3N(R); Cell Signaling, Danvers, MA, USA), CD57 (cone HNK-1; BD Pharmingen, San Diego, CA, USA), FOXP3 (clone 236A/E7; Abcam, Cambridge, UK), and PanCK (clone AE-1/AE-3; Novus Biologicals, Minneapolis, MN, USA). Panel 2 consisted of primary mouse anti-human antibodies for CD8, CD4, CD68, FOXP3, and PanCK, which were the same as Panel 1, plus CD208 (clone EPR24265-8; Abcam), CD19 (clone EPR5906; Abcam), and CD20 (clone L26, DAKO). Full methodological details can be found in Supplementary Table 1 text.

2.5. Image acquisition and analysis

For multispectral image acquisition of sample tissues, we performed low-resolution (4×) whole-slide scans of the stained specimens and their Haematoxylin and Eosin (H&E)-stained equivalent using the PhenoImager HT (Akoya Biosciences, Marlborough, MA, USA). Evaluation of the H&E slides was performed by a senior consultant histopathologist (IB) at the Royal Surrey Hospital using Phenochart 1.0.12 (Akoya Biosciences), and areas of interest were identified within the tumor tissue. We then identified these areas on the whole-slide scan images of the multiplex-stained tissue sections to allow the acquisition of 20× multispectral images. We exported images that had been spectrally unmixed using InForm (Akoya) and analyzed in the open-source software package QuPath 0.3.1.¹³ We then performed tissue annotation using PanCK-Opal 780 as the tumor marker, with cell segmentation utilizing the DAPI channel. Phenotyping of the images was then completed using a machine learning classifier, which was trained across multiple images from each batch. Once each marker had been classified, they were combined into a single classifier, which was applied hierarchically to identify cell phenotypes and 2-dimensional spatial relationships.

2.7. NanoString gene expression profiling

For RNA extraction, we took five 20-µm thick FFPE tissue curls from the esophageal tumor from each patient’s tissue blocks. We extracted RNA using the Norgen Biotek FFPE RNA purification kit (Cat. 25300; Norgen Biotek) as per the manufacturer’s instructions. We quantified RNA quality and content by Nanodrop spectrophotometry (#ND-1000; ThermoFisher Scientific, Waltham, MA, USA), and samples were stored at –80°C before batch analysis. We hybridized RNA with the NanoString nCounter IO360 Panel Human Codeset (NanoString, Seattle, WA, USA). RNA was quantified using the nCounter Digital Analyzer (NanoString), and data were processed with nSolver Analysis Software (NanoString) using the Advanced Analysis module. Due to low quality RNA, we acquired data from n = 8 Controls and n = 9 Prehab. No statistical differences in mIHC were observed between those measured and not measured for NanoString.

2.8. Statistical analyses

We assessed group differences using t tests, χ² tests, and multivariate two-way analyses of variance (ANOVA) depending on data normality (GraphPad Prism v9.3.1.; San Diego, CA, USA). We used the Kaplan-Meier method and log-rank (Mantel-Cox) method to compare group mortality and the Kaplan-Meier plot to illustrate group survival (v9.3.1.; GraphPad Prism). Aerobic fitness outcomes were analyzed using a repeated measures ANOVA with group and time as fixed factors and Bonferroni post hoc analysis where required (SPSS Statistics v28.01.1; IBM Corp., Armonk, NY, USA). Group × time interactions were resolved with simple main effects examining the group response at each time point. Associations between percentage changes in relative V̇O_2peak (Baseline to Post-NAC and Baseline to Pre-surgery) and the frequency of TILs were analyzed using Spearman’s correlations. Cell type deconvolution analysis was performed using the advanced analysis nSolver software tool (NanoString) to generate an immune cell abundance for each cohort, which were then plotted as line graphs showing the relative expression change between exercise and control cohorts and as individual box plots. Given the exploratory nature of this analysis, statistical significance was set at p < 0.05 for main effects and p < 0.10 for group × time interactions.

3. Results

Most patients (mean age = 62.8 years; range: 52.7–75.5 years) included in the tumor analyses were male (90%), overweight or obese (82%), had a history of smoking (73%), and had low aerobic fitness, with only 2 patients (9%) fitter than their age- and sex-predicted (Wasserman equation) values (Table 1). All patients were diagnosed with adenocarcinoma. We observed no differences between groups in terms of chemotherapy received, chemotherapy tolerance, tumor histology, or mortality. Due to the publication of the FLOT4 trial¹⁴ and subsequent change of care during the study, slightly more Prehab patients (91%) were treated with epirubicin, carboplatin, and capecitabine (ECX) as per the MAGIC protocol compared to the Control group (73%). The rest of the patients were treated with FLOT (fluorouracil, leucovorin, oxaliplatin, and docetaxel). However, this did not result in differences in dose modifications, with 3 patients (27%) in Prehab and 5 patients (48%) in Controls having dose reductions.

3.1. Effects of prehabilitation on clinical outcomes

We present clinical information and outcomes after NAC and surgery in Table 2. Following surgery, no patients had complete tumor regression (tumor regression grade (TRG) 1) or rare residual cell detection (TRG 2). The majority of patients (68%) in both groups (Prehab: 64% vs. Controls: 73%) presented with TRG 4, which represents residual cancer that is outgrowing the fibrosis. Most patients in the Prehab (82%) and Control (64%) groups presented with a Stage 3 tumor, while 45% of Prehab presented with no detectable lymph node infiltration (N0) and 36% of Controls presented with low localized lymph node infiltration (N1). A total of 8 patients received dose reductions during NAC, but no differences were observed between groups. Overall survival was similar between groups (Supplementary Fig. 2) with Kaplan‒Meier analysis suggesting that the estimated survival probability at 1 year was 82% (95% confidence interval (95%CI): 69%–95%) in both groups and at 3 years was 73% (95%CI: 57%–89%) in Prehab and 68% (95%CI: 54%–83%) in the Control Group. The median survival time in the Control group was 1.5 years while the median survival in the Prehab group was not reached. The log-rank test showed that survival was not different between the groups (p = 0.600).

Table 2.

Clinical characteristics of patients following neoadjuvant chemotherapy and surgery.

	Overall (n = 22)	Prehab (n = 11)	Control (n = 11)	p
Clinical data
Number of patients who completed chemotherapy at full dose	14 (64)	8 (73)	6 (54)	0.375
Post-operative tumor histology
Adenocarcinoma	22 (100)	11 (100)	11 (100)	1.000
AJCC tumor stage
T1	3 (14)	1 (9)	2 (18)	0.709
T2	3 (14)	1 (9)	2 (18)
T3	16 (73)	9 (82)	7 (64)
T4	0	0	0
N0	8 (36)	5 (45)	3 (27)	0.355
N1	7 (32)	2 (18)	5 (45)
N2	3 (14)	2 (18)	1 (9)
N3	4 (18)	2 (18)	2 (18)
Mandard TRG				0.875
1	0	0	0
2	0	0	0
3	2 (9)	1 (9)	1 (9)
4	15 (68)	7 (64)	8 (73)
5	5 (23)	3 (27)	2 (18)
Survival
Alive at analysis of tumors	10 (45)	6 (54)	4 (36)	0.670

Notes: Data are n (%). Percentages might not add up to 100% due to rounding.

Abbreviations: AJCC = American Joint Committee on Cancer; N = node; T = tumor; TRG = tumor regression grade.

3.2. Effects of prehabilitation on aerobic fitness

Similar to the main trial, adherence to the supervised and home-based program was high (>70% of sessions were completed), resulting in a main effect of time (F (2,40) = 6.394, p = 0.004, η² = 0.242) and a group × time interaction (F (2,40) = 3.445, p = 0.042, η² = 0.147) for relative V̇O_2peak (Fig. 1A). This was characterized by a 9.0% ± 10.2% (mean ± SD) reduction at Post-NAC (p = 0.018) for the Controls, while the Prehab group maintained V̇O_2peak at Post-NAC (p = 1.000) and increased by 9.4% ± 7.6% from Post-NAC to Pre-surgery (p = 0.010). At Pre-surgery, the Prehab group’s V̇O_2peak was 3.27 ± 1.31 mL/kg/min higher than Controls (p = 0.022). Similarly, there was a main effect of time (F (2,40) = 6.907, p = 0.003, η² = 0.257) and a group × time interaction (F (2,40) = 3.544, p = 0.038, η² = 0.151) for absolute V̇O_2peak (Fig. 1B). This was characterized by a 10.5% ± 11.4% reduction at Post-NAC (p = 0.016) for Controls, while the Prehab group maintained V̇O_2peak at Post-NAC (p = 1.000) and increased by 9.6% ± 7.6% from Post-NAC to Pre-surgery (p = 0.007). For the V̇O_2peak at anaerobic threshold (Fig. 1C), there was a main effect of time (F (2,40) = 4.780, p = 0.015, η² = 0.219) that was characterized by an increase from Post-NAC to Pre-surgery (p = 0.037). Although we did not observe a group × time interaction, for full transparency we report that the anaerobic threshold was lower in the Controls at Post-NAC (mean difference (diff.) = –2.45 mL/kg/min, 95%CI: –4.75 to –0.15, p = 0.038) and Pre-surgery (mean diff. = –2.82 mL/kg/min, 95%CI: –4.70 to –0.93, p = 0.006).

3.3. Effects of prehabilitation on tumor cell frequencies

To determine differences in the immune cell infiltrate in esophageal tumor samples, we performed multispectral IHC on tumor tissues (Fig. 2A and 2B). We used a generic immune cell multiplex panel of markers (CD8, PDL1, CD57, FOXP3, CD68, GrzB, CD4, and Pan-CK). Prehab had significantly more CD8+ cells in the tumors (mean diff. = 1.79, 95%CI: 0.76‒2.82, p < 0.001; Fig. 2B) and the stroma (mean diff. = 1.59, 95%CI: 0.66‒2.52, p < 0.001; Fig. 2C) than the Controls. We did not observe any significant differences between groups for tumor or stroma frequencies of CD4+, CD68+, PDL1+, CD57+, or GrzB+ cells CD8+/PDL1+, CD8+/GrzB+, CD4+/FOXP3+ (all p > 0.05).

Fig. 2.

Prehabilitation exercise training is associated with an increased frequency of CD8+ tumor infiltrating lymphocytes (TILs) in esophageal tumors in an aerobic capacity-mediated effect. Multiplex immunohistochemistry using an 8-plex immune panel plus DAPI was applied to the tumor tissues and imaged using Akoya’s PhenoImager HT. InForm® automated image analysis software was used to visualize and quantify the immune infiltrate. Percentages of TILs are presented relative to the total frequency of nucleated cells. (A) Representative images of tumor tissue from patients randomized to the Control (n = 11) or Prehab group (n = 11), with 2 patients from the Control group (top and bottom left) and 2 patients from the Prehab group (top and bottom right); (B) Tumor infiltrating cell marker frequencies between Prehab and Controls. (C) Stroma-infiltrating cell marker frequencies between Prehab and Controls. Relationships between the percentage change in peak oxygen consumption (mL/kg/min) from Baseline to Post-NAC and the frequency of (D) CD8+ TILs, (E) PDL1+ cells, and (F) Granzyme B+ TIL. In (A): CD8 (red), CD4 (white), CD68 (maroon), granzyme B (orange), PDL1 (green), CD57 (yellow), cytokeratin (cyan), DAPI (blue). Data are means and standard error of the mean. NAC = neoadjuvant chemotherapy.

3.4. Relationships between changes in VO_2peak and the frequency of tumor infiltrating cells

To determine the relationships between Prehab- and Control-induced changes in aerobic capacity and TILs, we correlated percentage changes of relative V̇O_2peak with TILs frequencies for all participants together (Fig. 2). Between Baseline and Post-NAC where the Prehab group maintained V̇O_2peak better than Controls, there were significant positive associations, with changes in V̇O_2peak and the frequencies of CD8+ TILs (r = 0.531, p = 0.016; Fig. 2D), PDL1+ cells (r = 0.566, p = 0.009; Fig. 2E), and GrzB+ TILs (r = 0.582, p = 0.007; Fig. 2F). We then ran correlations independently for each group and found relationships in only the Prehab group for changes in V̇O_2peak and the frequencies of CD8+ TILs (r = 0.727, p = 0.011), PDL1+ cells (r = 0.615, p = 0.044), and GrzB+ TILs (r = 0.774, p = 0.005). For the whole cohort, no relationships existed between changes in Baseline and Pre-surgery V̇O_2peak or any TIL subsets. However, only the Prehab group showed relationships between changes in V̇O_2peak and the frequencies of CD8+ TILs (r = 0.706, p = 0.015) and GrzB+ TILs (r = 0.683, p = 0.020) but not PDL1+ cells (r = 0.477, p = 0.138).

3.5. Effects of prehabilitation on gene expression levels in tumors

To gain further insight into the immune gene expression levels in response to Prehab, we employed the NanoString platform (NanoString) to examine the expression of 770 highly annotated genes with the PanCancer IO360 panel (NanoString). Immune cell deconvolution of the IO360 panel transcriptome data suggested that Prehab was associated with an altered tumor microenvironment compared to the Control group (Fig. 3A). RNA quality was sufficient for gene expression analyses on 8 Controls and 9 Prehab. When normalized to total numbers of TILs (Fig. 3B), Prehab was associated with higher levels of CD56+ NK cells (median diff. = 0.87, 95%CI: 0.25‒2.18), p = 0.0274; Fig. 3C), of which CD56^dim NK cells were highest (median diff. = 0.48, 95%CI: 0.03‒2.53; Fig. 3D), p = 0.0464). We did not observe differences in levels of CD8+ T cells (median diff. = 0.19, 95%CI: –0.37 to 1.00, p = 0.2766; Fig. 3E). Similarly, we observed no differences (all p > 0.05) for levels of other immune cells, including cytotoxic cells (Fig. 3F), exhausted T cells (Fig. 3G), CD3+ T cells (Fig. 3H), B cells (Fig. 3I), macrophages (Fig. 3J), neutrophils (Fig. 3K), dendritic cells (Fig. 3L), or mast cells (Fig. 3M).

Fig. 3.

Transcriptional analyses reveal that Prehabilitation is associated with increased natural killer (NK) cell tumor infiltration. (A) We clustered gene expression values into panels to decipher and estimate the immune cell abundance for Controls (n = 8) and Prehab (n = 9). (B) Differences between groups for total tumor infiltrating lymphocytes (TIL) and log₂ transformed normalized to levels of TIL for (C) CD56+ NK cells, (D) CD56^dim NK cells, (E) CD8+ T cells, (F) cytotoxic cells, (G) exhausted T cells, (H) CD3+ T cells, (I) B cells, (J) macrophages, (K) neutrophils, (L) dendritic cells, (M) mast cells. Boxplots represent the median and interquartile range with error bars reflecting the minimum and maximum values. Prehab = prehabilitation.

As we were underpowered for gene expression data, pathway analysis of the gene expression profiles revealed no significant differences between groups (Fig. 4). However, there were trends in keeping with known effects of exercise, including a reduction in hypoxia and transforming growth factor-β (TGFβ)-signaling pathways, while antigen presentation, cytotoxicity, cytokine, and chemokine signaling and the lymphoid and myeloid compartment pathways were all elevated in the Prehab cohort (Fig. 4A).

Fig. 4.

Pathway analysis and gene expression reveal that Prehabilitation is associated with altered tumor gene expression. (A) Clustered gene pathway analyses differences between groups. (B) Volcano plot of differential gene expression plotted showing log₂ fold expression change (X axis) vs. ‒log₁₀ (p value) (Y axis). The genes with Wald test p ≤ 0.05 are represented by blue circles (lower expression in Prehab) or red circles (higher expression in Prehab). Prehab = prehabilitation.

Differential gene expression analysis comparing Prehab to the Controls revealed 16 non-false discovery rate (FDR) adjusted genes with a Wald test p ≤ 0.05 (Fig. 4B). Of these, 7 genes had ≤–1 or ≥1 log₂ fold differences between groups. Relative to Controls, Prehab was associated with higher expression of 3 genes primarily involved in inflammatory signaling (TNFSF13), matrix remodeling (MMP1), and MAP-kinase signaling (MET). Conversely, Prehab was associated with lower expression of 13 genes involved primarily in matrix remodeling and metastasis (PLOD2, ITGA6, VCAM1) and metabolic stress (GOT2, MAP3K5, RAD50, ENO1, GOT1, ERO1A, EGFR) pathways. Table 3 shows the top 20 non-FDR adjusted differentially expressed genes between Prehab and Controls.

Table 3.

Top differentially expressed genes and related pathways.

Gene (mRNA)	Log2 fold change	Wald test p	Pathway
GBP2	–1.25	0.0099	Interferon signaling
MMP1	2.68	0.0141	Matrix remodeling and metastasis, myeloid compartment
TNFSF13	0.641	0.0179	NF-κB signaling
PLOD2	–1.69	0.0214	Matrix remodeling and metastasis
DTX3L	–0.588	0.0218	Antigen presentation, notch signaling
MET	2.09	0.0238	MAPK, metabolic stress, PI3K‒Akt
GOT2	–0.752	0.0285	Metabolic stress
MAP3K5	–1.01	0.0328	MAPK, metabolic stress
CDH2	–0.803	0.0329	Immune cell adhesion and migration
RAD50	–1.07	0.0363	Cell proliferation, DNA damage repair, metabolic stress
ENO1	–0.906	0.0369	Hypoxia, metabolic stress
GOT1	–1.32	0.0369	Metabolic stress
IRF1	–1.12	0.0427	Cytotoxicity, interferon signaling
CCL21	–0.848	0.0439	Cytokine and chemokine signaling
VCAM1	–1.39	0.0465	Immune cell adhesion and migration, interferon signaling, Matrix remodeling and metastasis
ITGA6	–1.15	0.0480	Immune cell adhesion and migration, matrix remodeling and metastasis, PI3K‒Akt
IFNAR1	–0.631	0.0549	Interferon signaling, JAK‒STAT signaling, PI3K‒Akt
ERO1A	–0.826	0.0566	Metabolic stress
EGFR	–0.721	0.0574	Hypoxia, MAPK, metabolic stress, PI3K‒Akt
FCGRT	–0.369	0.0578	No pathway identified

Abbreviations: DNA = deoxyribonucleic acid; JAK–STAT = Janus kinase-signal transducers and activators of transcription; MAPK = mitogen-activated protein kinase; NF-κB = nuclear factor kappa-light-chain-enhancer of activated B cells; PI3K–Akt = phosphatidylinositol 3-kinase – protein kinase B.

3.6. Maturation of tertiary lymphoid structures may be influenced by prehabilitation

Having identified increased infiltrating immune cells and transcriptional changes within the tumors of patients who had undergone prehabilitation, we were interested in determining whether this resulted in the increased organization of TLS. A further multiplex IHC panel was optimized to allow for the identification of TLS as well as their cellular composition and maturation. Eight immune cell markers (CD8, CD4, CD20, CD19, FOXP, CD68, CD208 (dendritic cell lysosomal associated membrane glycoprotein (DCLAMP), and Pan-Cytokeratin (as a tumor marker)) were multiplexed on the tumor samples (Fig. 5A). Multispectral images were processed by machine learning software InForm and QuPath, which were trained to automatically segment the TLS and provide phenotyping of the cellular composition (Fig. 5A and 5B).

Fig. 5.

Prehabilitation is associated with the development of more mature tertiary lymphoid structures (TLS). (A) Representative structure of a tumor associated TLS in esophageal adenocarcinoma; regions of interest containing TLS were stamped on the whole tissue scan, imaged at high resolution followed by image analysis in InForm to identify cells using machine learning algorithmic segmentation and phenotypic identification based on all stained channels. Composite fluorescent multiplex image along with single marker fluorescent images showing the phenotyping of different cell types within a TLS using InForm software (CD19, yellow; CD20, pink; CD68, orange; CD4, cyan; FOXP3, magenta; CD8, red; DCLAMP, green; Pan-CK, white; DAPI, blue). (B) Representative images of TLS from tumors from the Control vs. the Prehab cohorts. High-power field images of each of the TLS per tumor case were then used for automated trained tissue segmentation and the total cell counts within each TLS were calculated along with the TLS surface area. (C) The proportions of each of the immune cell subsets within TLS were determined for tumors from the Control and the Prehab cohorts.

Evaluation of the presence and localization of tumor-associated TLS in the esophageal tumors revealed that most TLS were in the peritumoral regions (Supplementary Fig. 3A). We observed no differences in the numbers of TLS normalized to the tumor area between groups (Supplementary Fig. 3B). Further analysis determined the cell density and surface area of the TLS as a measure of the maturity of the TLS. Prehab was associated with a higher TLS cell density (median diff. = 18,959, 95%CI: 13,518‒22,635, p < 0.001; Fig. 5B), and although surface area appeared visually smaller and less diffuse, this was not significantly different between groups (median diff. = –0.47, 95%CI: –1.54 to 0.33, p = 0.513; Fig. 5B). Additionally, Prehab tumors had visually more clearly defined germinal centers indicative of mature TLS, which we are unable to quantify (Fig. 5B). We further assessed the cellular composition of the TLS and showed that the proportion of component cell types differed between the exercise and control tumor cohorts (Fig. 5C). Specifically, there was a significantly larger proportion of DCLAMP+ dendritic cells (p = 0.0157) and non-significant trends towards more CD4+, CD8+, and CD68+ cells (all p > 0.05) in the tumor-associated TLS from the Prehab group compared to the Controls.

4. Discussion

This proof-of-concept study shows that low-to-moderate intensity prehabilitation exercise during neoadjuvant chemotherapy is associated with changes in the tumor microenvironment of patients with esophageal adenocarcinoma. We show that prehabilitation promotes an increased frequency of CD8+ tumor infiltrating immune cells, while gene expression data suggest higher levels of NK cells. Additionally, we show that prehabilitation is associated with the development of mature TLS in the peritumoral regions. We characterized these TLS by increased cell density and the frequency of CD4+, CD8+, macrophages, and dendritic cells. Critically, the frequency of CD8+ tumor infiltrating immune cells was associated with prehabilitation-induced changes in aerobic fitness after neoadjuvant chemotherapy, suggesting a relationship between aerobic fitness and immune responses. Although we observed no significant differences in clinical outcomes, our data highlight the role and potential of prehabilitation exercise that improves aerobic capacity during neoadjuvant chemotherapy as a non-pharmacologic immune modifier in esophageal adenocarcinoma.

Of the TIL subsets, we found that exercise training was associated with higher levels of CD8+ lymphocytes and trends for CD4+ lymphocytes. In a retrospective analysis of patients with esophageal adenocarcinoma mostly undergoing NAC (59.4%), Noble et al.¹⁵ showed increased levels of intratumoral CD3+, CD8+, and CD4+ lymphocytes were associated with reduced pathological N stage and better disease-free survival. Although it is unclear whether these TILs are present before NAC (and, as such, whether those patients are predisposed to a better response), increasing TILs frequencies may provide similar beneficial outcomes. The role of exercise training in cancer control and the tumor microenvironment has focused chiefly on preclinical studies.^6,16, 17, 18 Pedersen and colleagues⁶ observed that 30 days of voluntary running reduces murine tumor burden by increasing CD56+ NK cell infiltration in lung, liver, and melanoma tumors. In agreement, we show gene expression profiles of TILs that consist primarily of CD56+ NK cells. However, our mIHC panel did not assess the co-expression of CD3, CD8, and CD56, making it difficult to confirm exactly which cells were present. Further, as we could only determine gene expression on 17 samples (8 Control and 9 Prehab), we cannot be certain that CD8+ T cells were not different between groups. Although we did not assess NK cell function, around 30% of NK cells expressed CD8.¹⁹ CD8+ NK cells could explain our immunohistochemistry findings, and importantly, these CD8+ NK cells have increased cytolytic activity and reduced activation-induced apoptosis.¹⁹

Although our results are promising, we were unable to assess diagnostic biopsy TIL levels. Although there is a possibility that more Prehab patients had higher levels of TILs, which would translate to higher TILs at surgery, we believe our randomization should have taken care of this. However, future studies should endeavor to compare both to confirm the effects of exercise. Subsequently, others have not shown similar results. Djurhuus and colleagues^8,9 observed that neither a single bout of exercise nor 4–30 sessions of total exercise training sessions before prostatectomy increased tumor infiltrating CD56+ cells. Although their cancer population differed greatly from ours (e.g., no neoadjuvant therapies), they observed increased CD56+ cell infiltration with more completed exercise sessions. This was derived from their per-protocol analyses, which suggest that the more exercise completed, the better the tumor-immune response.⁹ In agreement, we observed higher frequencies of TILs expressing cytotoxic markers, including CD8+ and granzyme B+, with larger changes in cardiorespiratory fitness.

Higher cardiorespiratory fitness has been robustly associated with better surgical outcomes and overall survival in several solid tumors.²⁰ In our study, the exercise program consisted of low-to-moderate intensity aerobic exercise training that, coupled with resistance and flexibility training, maintained fitness during NAC and increased fitness before surgery. However, more significant increases in cardiorespiratory fitness are evident with higher intensities of aerobic exercises. Moderate-to-vigorous and high-intensity interval training (HIIT) consistently produce more prominent changes in V̇O_2peak.²¹ HIIT is generally accepted to increase V̇O_2peak in a shorter period than other intensities and may provide a more efficient means to increase fitness, and possibly TILs, in esophageal cancer. We and others have shown that HIIT is feasible in patients undergoing cancer therapies and can increase V̇O_2peak.^10,22 Future studies should determine whether higher intensities of exercise during a time-constrained treatment window do increase V̇O_2peak to a larger magnitude and whether this results in more TILs. We now add to this literature and suggest that the more we increase cardiorespiratory fitness during NAC, the more chance we have of inducing a TIL response.

The mechanisms by which exercise promotes enhanced immune responses in cancer are multifactorial and may differ between individual patients and cancer types.^16,18 Exercise training promotes potent anti-inflammatory effects.²³ However, each bout of exercise performed stimulates acute muscle-induced inflammatory and metabolic signaling that promotes fitness adaptations. The release of muscle-specific cytokines (i.e., myokines) such as interleukin-6 (IL-6), IL-15, and IL-7 also promotes enhanced immune surveillance by circulating NK and T cells.²⁴ This, coupled with increased chemokine signaling, allows effector immune cells to traffic from the circulating blood to the tumor.⁶ In their non-randomized control trial, Zylstra and colleagues³ observed that exercise training during NAC for esophageal cancer significantly improved tumor regression. Tumor regression was associated with increased circulating blood CD3+ and CD8+ lymphocyte counts and reduced TNFα concentrations in the exercise group. Although it was unclear what the average duration of exercise training was, most patients in the control (84%) and exercise (75%) groups completed NAC, suggesting that exercise was at least 8 weeks long. Although neither we nor Zylstra et al.³ assessed myokine changes, our data indicates that exercise training provides systemic and tumor immune enhancements through known exercise-mediated pathways.

To further explore the mechanisms, we performed NanoString gene expression analyses to identify differentially regulated gene ontology pathways in tumors. Similar to murine tumor samples after exercise,⁶ we observed differential gene expression pathways associated with better immune function, inflammation, and signatures for an increased lymphoid compartment. To address this lymphoid compartment, additional staining of tumor samples revealed that exercise was associated with increased maturity of TLS. The presence of mature TLS in esophageal squamous cell carcinoma was recently associated with better recurrence-free and overall survival.²⁵ In agreement with the findings from Nakamura et al.,²⁵ we also found higher levels of cytotoxic immune cells and antigen-presenting dendritic cells in TLS after exercise training. However, we did not observe better overall survival in the exercise group, most likely due to the low sample size. To the best of our knowledge, our study is the first to show that TLS maturation in human tumors is influenced by exercise training. However, it is unknown whether prehabilitation exercise training in esophageal adenocarcinoma provides survival benefits similar to those observed by Noble et al.¹⁵ and Nakamura et al.²⁵

Our study has several strengths, including the randomization of patients to either prehabilitation exercise or the standard of care at that time. Additionally, our robust and comprehensive analysis of tumor samples has provided novel information about the role of exercise in human tumor biology. We have previously discussed the limitations of our study, which included the small sample size and its single-center design.² Although this suggests that results may not be generalizable to the wider population, we are confident that our population was representative of patients with advanced esophageal adenocarcinoma at the time. Of course, with changes in current treatment and new diagnostic tools, such patients may differ and, therefore, the effects of exercise may also differ. Specific limitations of this secondary analysis are that, as a post hoc analysis, the findings require further corroboration through studies designed to examine this effect specifically. While there were no differences in completion rates or type of chemotherapy delivered between our cohorts, there was a change of standard of care chemotherapy used in the neoadjuvant setting during the trial. This has led to the majority of patients now receiving FLOT rather than Epirubicin, Cisplatin and Capecitabine (ECX) chemotherapy.²⁶ Docetaxel is known to induce “immunogenic modulation” of tumor cells and increase their susceptibility to cytotoxic CD8⁺ T cell lymphocytes.²⁷ Therefore, enhancing CD8⁺ T cell infiltration with exercise alongside FLOT chemotherapy may lead to improved tumor destruction. We further analyzed TILs between FLOT and ECX and those with dose modifications but found no differences. However, our small sample size was underpowered for this analysis. We were also limited by not having access to diagnostic biopsies to determine whether patients in the exercise group had higher levels of TILs to start with. Although this may have occurred, it is unlikely as patients were randomly assigned but clinically similar at baseline. Furthermore, our whole cohort had tumor regression grades of 3 or higher, indicative of those who did not respond well to chemotherapy. Although this is similar to the larger initial study and the general population, we are limited in understanding how exercise works in those who respond well to chemotherapy. Future studies are warranted to understand the optimal dose of prehabilitation exercise that elicits the largest increase in V̇O_2peak and, hopefully, produces the best circulating- and tumor-immune responses when accompanied by chemotherapy, which may provide new clinical pathways for these patients.

5. Conclusion

Our study demonstrates for the first time that prehabilitation exercise training is capable of altering the tumor microenvironment of patients undergoing neoadjuvant chemotherapy and surgery for esophageal adenocarcinoma, albeit without significant clinical outcome differences. The larger changes in cardiorespiratory fitness were associated with increased frequency of TILs. This gives further evidence of the relationships between cardiorespiratory fitness and better cancer control. Additionally, our novel findings of increased maturation of TLS provide evidence of a change in the tumor that may facilitate better responses to checkpoint immunotherapies.

Authors’ contributions

CJR conceived the tumor study, carried out IHC analysis, and drafted the manuscript; DBB conceived the tumor study, analyzed the fitness data, participated in the statistical analysis and drafted and finalized the manuscript; SKA conceived the exercise study, recruited patients and participated in the surgery; TW and TS participated in the IHC and tumor assays; SS optimized the IHC panels and oversaw the data acquisition; JH and DK completed the exercise training and fitness assessments of patients; IB conducted the pathologhical assessment of tumors; JS and SRP recruited patients and oversaw the surgical procedures; AEF conceived the tumor study, participated in the clinical analysis of patients, and drafted the manuscript; NEA conceived the tumour study, oversaw all the biological works, confirmed the data robustness, and drafted the manuscript; NAG conceived the tumor study, recruited patients, oversaw the surgical procedures, and participated in the clinical analysis of patients and drafted the manuscript. All authors have read and approved the final version of the manuscript, and agree with the order of presentation of the authors.

Competing interests

The authors declare that they have no competing interests.