Abstract
Background:
The kynurenine (KYN) pathway has been implicated in many diseases associated with inflammation and aging (“inflammaging”). Targeting the kynurenine pathway to modify disease outcomes has been trialled pharmacologically, but the evidence of non-pharmacological means (ie, exercise) remains unclear.
Objective:
We aim to assess the evidence of the effects of exercise on the kynurenine pathway and psychological outcomes.
Methods:
Under Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines, a systematic literature search was performed in MEDLINE, EMBASE, EMCARE, and the Cochrane Central Registry of Controlled Trials. The main outcomes were changes in kynurenine pathway metabolite levels and psychological outcomes.
Results:
Six studies were analyzed (total n = 379) with exercise demonstrating significant concomitant effects on kynurenine pathway metabolite levels and associated psychological outcomes in domains of somatization, anxiety, and depression.
Conclusion:
Exercise has significant concomitant effect on kynurenine pathway metabolite levels and psychological outcomes. However, clear limitations exist in determining if the changes in the kynurenine pathway can fully explain the changes in psychological outcomes, or whether different diseases and exercise interventions act as confounding factors.
Keywords
Introduction
Aging is a process that every human goes through; our population is indeed living longer, with the median lifespan generally increasing.1,2 Despite its universal nature, the exact mechanisms of aging remain incompletely understood. One leading explanation, 3 however, explores the possibility that aging is associated with, or is a result of, continuous inflammatory insult throughout life. These inflammatory changes are associated with a number of diseases, including cardiovascular, 4 neurological,5,6 and musculoskeletal diseases, 7 and cancer. 8 Interestingly, while these diseases are generally called “age-related,” it seems likely that inflammatory changes start early and the accumulation of changes leads to disease progression. Nonetheless, given that each of these diseases have been shown to be age-related,9-12 it is worth exploring possible mechanisms of this “inflammaging” 13 process.
Inflammation is a highly complex process involving mediators14,15 with multiple effects on physiological and pathological processes. There are many pathways to explore that may be studied for potential therapeutic intervention.3,16 One pathway of recent interest linked to both inflammation and aging 13 is the kynurenine pathway of tryptophan degradation. The kynurenine pathway is involved in many physiological processes; well-known examples include tryptophan as the precursor for serotonin, an important neuroactive mediator, 17 and the role of gut microbiota in tryptophan metabolism. 18 It is becoming known that imbalances in tryptophan metabolism and the associated kynurenine pathway are involved in the process of aging and the progression of age-related diseases. 19 The kynurenine pathway has been implicated in many age-related disease categories including cardiovascular (eg, coronary artery disease), 20 neurological (eg, Alzheimer’s disease, depression, schizophrenia),21-23 musculoskeletal (ie, osteoporosis), 24 and cancer.25,26
There are various consequences of changes in the kynurenine pathway; for example, changes in serum levels of kynurenine, an intermediate metabolite of this pathway, have been shown to correlate with symptom severity in patients with Parkinson’s disease. 27 Some studies have even gone so far as to investigate individual metabolites, such as indoleamine 2,3-dioxygenase (IDO), and their relation to cancer-related fatigue. 28 Thus, there seems to be potential to target discrete elements of the pathway for therapeutic benefit in patients with age-related diseases, and there are many pharmacological treatments under current investigation to target the kynurenine pathway in different disease states. 25
In addition to pharmacological treatment, however, one must consider non-pharmacological interventions (ie, lifestyle) for disease. Of particular interest is exercise and its effect on the kynurenine pathway. As a lack of physical activity is increasingly common, 29 and linked to many chronic diseases associated with inflammaging, 30 exercise is an intervention worthy of consideration. However, exercise is an intervention that is highly variable in terms of frequency, intensity, time, and type. 31 It is especially important to distinguish between acute and chronic exercise bouts when considering the effects of exercise on the kynurenine pathway. It is known that both acute and chronic exercise alters cellular immune function in general.32,33 However, 2 reviews34,35 have suggested acute exercise may produce different kynurenine pathway outcomes than chronic exercise. The potential mechanism for these inflammatory pathway changes may involve accumulation of changes due to repetitive bouts of acute exercise. 34 While acute exercise does result in significant changes in cellular immunity (eg, white blood cell proportions), these changes tend to be temporary; in chronic exercise, however, these changes are not temporary, 32 as there are likely long-term adaptations of the immune system in response to such continuous exercise.
Therefore, the aims of this review were to assess the evidence of the effects of exercise on the kynurenine pathway and psychological outcomes in age-related disease.
Methods
This review was registered at PROSPERO (University of York) with registration number CRD42020204035 and conducted according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines.
Search strategy
A systematic literature search was performed in MEDLINE, EMBASE, EMCARE, and the Cochrane Central Registry of Controlled Trials, from the first available date of the database to September 2020. The search strategy created and executed is shown in Table 1.
Search strategy developed for database search using ovid interface.
Inclusion and exclusion criteria
COVIDENCE was used for reference importing, title and abstract screening, and full text review.
Inclusion criteria:
– Experimental trials or randomized controlled trials comparing exercise intervention with placebo/baseline exercise control in persons with known age-related disease.
– Age-related disease can include, but are not limited to, cardiovascular, neurological, or oncological disease.
– Exercise can include, but is not limited to, aerobic, resistance, mixed, high-intensity interval training, passive, certain forms/sports of exercise (running/jogging, swimming).
– Physiological/metabolic outcomes can include, but are not limited to, kynurenine pathway measures of activity (eg, kynurenine/tryptophan ratio), kynurenine pathway metabolites (eg, kynurenic acid), inflammatory markers (eg, IL-1, neopterin).
– If psychological outcomes are present, these can include, but are not limited to, quality of life, symptom burden/relief, mood.
Exclusion criteria:
– Reviews of primary studies.
– Non-human studies, which can include, but are not limited to, animals (eg, mice), cell culture lines, bacteria.
– Persons without current age-related disease or risk factors for age-related disease.
– Studies in which exercise was not compared against a valid control group, for example, both groups exercising, but the intervention was actually a protein supplement versus placebo.
Studies were screened by 2 reviewers (AL, CH), and studies without consensus were given a final decision by a third reviewer (GD).
Data extraction
COVIDENCE and Google Sheets were used to compile study data and assess risk of bias in studies. Risk of bias was assessed using either Cochrane RoB 2.0 36 or ROBINS-I. 37 Bias was assessed in 2 rounds: first round separately by AL and CH, second round for consensus.
Results
Included studies
The search strategy identified 360 studies (see Figure 1). Bibliographic search outside the database search identified 7 studies. Eighty-six duplicates were removed.
PRISMA flowchart.
Table 2 provides a summary of studies and their effects of intervention on the kynurenine pathway. Included disease pathologies included cardiovascular (diabetes), cancer (pancreatic, breast, gastroesophageal), neurological disease (stroke), psychiatric disease (major depressive disorder, dysthymia, somatization syndrome).
Included study characteristics.
Abbreviations:
-Biochemical: BCAAs, branched chain amino acids; f-TRP, free-tryptophan; KMO, kynurenine 3-monooxygenase; Kynurenine, kynurenine; KYNA, kynurenic acid; QUINA, quinolinic acid; TRP, tryptophan.
-Psychiatric: BDI, Beck Depression Inventory; HADS, Hospital Anxiety and Depression Scale; MDD, major depressive isorder; SSI-8, Somatoform Symptom Index-8; SOMS7, Screening for Somatoform Disorders 7.
Risk of bias assessment
Herrstedt et al 41 was at “critical” (ROBINS-I) risk of bias, owing to the use of an ill-described and unsupervised control group with no record-keeping of types of exercise in the control group, missing outcome data with a high proportion of non-adherence, and omission of one data set deemed of sufficient importance.
Pal et al 42 was at “high” (Cochrane RoB 2.0) risk of bias, owing to randomization that actually started using living distance as a condition (introduces confounders), missing outcome data likely to differ between intervention arms, and different supervision levels between intervention arms.
Hennings et al 40 was at “high” (Cochrane RoB 2.0) risk of bias, owing to the use of many self-reported questionnaires (including a demographics questionnaire and self-reported food intake log) in outcomes measures.
Zimmer et al 43 was evaluated using both the Cochrane RoB 2.0 tool for the randomized controlled trial portion comparing breast cancer intervention and control groups, and the ROBINS-I tool for the observational study comparing breast cancer (intervention + control) versus healthy intervention groups. Zimmer et al 43 was at “unclear/some concerns” (Cochrane RoB 2.0) risk of bias, owing to the lack of information of how the 24 healthy volunteers were allocated for inclusion in the study (eg, was it the first 24 who volunteered, or 24 randomly assigned from a larger volunteer pool). Zimmer et al 43 was at “moderate” (ROBINS-I) risk of bias, owing to possible confounders such as radiotherapy and different stages/types of breast cancer; however, the study did measure baseline differences and made clear the limitations of these confounders.
Mudry et al 38 was at “moderate” (ROBINS-I) risk of bias, owing to the single-intervention arm treatment design with possible confounders unaccounted for. However, the exclusion criteria were adequately defined to exclude most confounders of diabetes comorbidities.
Baek et al 39 was at “moderate” (ROBINS-I) risk of bias, owing to the confounders that could not be eliminated due to matching for physical demographics. However, the exclusion criteria were adequately defined to exclude most confounders (eg, overlapping psychiatric conditions) of post-stroke depression.
Discussion
In this review, we considered the current evidence of exercise and its effects on the kynurenine pathway concomitantly with psychological outcomes. In particular, this review included studies with patients with known chronic/age-related diseases (and healthy controls, where appropriate). See Table 2 for all outcome data discussed.
Our findings suggest exercise has significant effects on multiple parts of the kynurenine pathway 25 from the start (tryptophan), to quinolinic acid—the precursor to the NAD+ pathway, and other branch pathways such as that leading to anthranilic acid. However, mixed results were found. Table 2 contains all biochemical outcome data; as there were numerous measurements and comparisons made, the discussion on biochemical outcomes will focus on broader comparisons.
Comparing post-exercise versus pre-exercise within the same group (ie, not vs healthy controls), for example, serum tryptophan could have significantly decreased, 38 increased,39,41 or showed no significant difference.40,42,43 Each study implemented different exercise programs (eg, aerobic, resistance, mixed), making it hard to predict differential biochemical effects. With the lack of studies on this topic, it becomes more challenging to ascertain the true direction of change in certain kynurenine pathway metabolites.
Comparing post-exercise versus pre-exercise between groups (ie, diseased vs healthy controls), for example, serum tryptophan (as compared previously), could have been significantly higher, 39 or showed no statistically significant difference38,40,42,43; Herrstedt et al 41 made no intergroup comparison. Post-exercise intergroup differences may suggest a different pre-exercise baseline, as was the case in Zimmer et al 43 (significantly higher kynurenine/TRP ratio in breast cancer patients vs healthy controls), or a true effect of exercise producing a differential kynurenine pathway outcome (ie, a significant group × time interaction as was the case with kynurenine in Pal et al. 42 )
While there are extensive comparisons made between post-exercise versus pre-exercise, another comparison alluded to in the Introduction section is the acute versus chronic exercise conditions. Only one study (Mudry et al 38 ) qualifies as acute exercise, as this was a single session of cycle ergometry with no other follow-up exercise interventions. Importantly, Mudry et al 38 reported that in response to acute exercise in both groups (normal glucose tolerance, type 2 diabetes), there was a significant decrease in serum tryptophan and kynurenine, and a significant increase in kynurenic acid. However, for chronic exercise studies included in our review, the post-exercise versus pre-exercise results were mixed. Serum tryptophan increased, 39 decreased, 41 or showed no significant difference; 40 Pal et al 42 and Zimmer et al 43 did not report on the direction of change for serum tryptophan, but noted a significant time effect post versus pre-exercise. Serum kynurenine showed no significant difference40,41; Pal et al 42 and Zimmer et al 43 did not report on the direction of change for serum kynurenine, but noted a significant time effects post versus pre-exercise. Serum kynurenic acid showed no significant difference; 41 Zimmer et al 43 did not report on the direction of change for serum kynurenic acid, but noted a significant time effects post versus pre-exercise. These mixed results, while difficult to directly compare to the acute exercise study by Mudry et al, 38 do not rule out the suggestion that there may be differential activation of the kynurenine pathway between acute and chronic exercise. 34
Further complicating the discussion of acute versus chronic exercise is the importance of obtaining the sample at the correct time, especially for measuring inflammatory-associated pathways. 44 Ideally, to measure the acute/short-term effects of an intervention, a measurement should be taken immediately after the intervention is completed. In the acute exercise study by Mudry et al, 38 the samples for serum metabolite analysis were appropriately obtained “immediately” after both exercise (“EXERCISE” time period) and a 3-hour sitting period (“RECOVERY” time period). The immediate sampling ensures that the measurements accurately reflect the acute effects of exercise on the kynurenine pathway. The sampling after the 3-hour sitting period is useful to compare to both exercise and baseline and show there is measurable change before and after exercise. For chronic exercise, however, the measurement process must be done more carefully. After any exercise bout, there will be short-term changes (eg, inflammatory 45 ) relative to pre-exercise. Thus, it is important not to sample too quickly after the exercise bout, even though it may be session number 10 out of 20 total sessions, since the goal is to measure change over all multiple or all 20 sessions, rather than just the single session. The chronic exercise studies had variable measurement protocols, with samples being obtained either “immediately” after exercise 39 or during some unspecified time after exercise.40-43 One study 44 —while not specifically about the kynurenine pathway—has suggested 2 cut-off points for acute exercise: up to 4 hours post-exercise for measuring inflammatory-associated outcomes; 24 hours post-exercise is not useful. This seems to suggest that to study chronic exercise effects, samples should be taken after 24 hours post-exercise. However, given all the chronic exercise studies in our review did not provide any evidence of such a rigorous sampling time period, there may be considerable difficulty in interpreting results. Future studies must not only account for the exercise intervention and outcomes themselves, but also the measurement and sampling thereof.
While all studies measured markers of the kynurenine pathway (and other inflammatory markers), only 3 studies (Baek et al, 39 Hennings et al, 40 Herrstedt et al 41 ) measured psychological outcomes. In this review, the included studies measured 3 psychological domains: somatization, anxiety, depression. Baek et al 39 and Hennings et al 40 reported on depressive symptoms using the Beck Depression Inventory (BDI) 46 score, and Hennings et al 40 reported on somatoform symptoms using the Screening for Somatoform Symptoms-7 (SOMS-7) 47 score. Herrstedt et al 41 reported on depressive and anxiety symptoms using the Hospital Anxiety and Depression Scale (HADS). 48
In Baek et al, 39 depressive symptoms were significantly decreased in the final week of exercise (vs first day). In Hennings et al 40 depressive and somatoform symptoms were significantly decreased in all groups when considering post- versus pre-exercise. In Herrstedt et al, 41 there was a significant decrease in HADS depression and anxiety scores when considering post- versus pre-exercise. Notably, in Hennings et al, 40 there was a significant decrease in depressive and somatoform symptoms in the control group (post- vs pre-exercise) with no known psychiatric conditions; there was a trend of significant group × time interaction that may have justified a “stronger effect” in the groups with major depression and somatization, but for reasons unknown (eg, small sample size), this finding was not conclusive. Similarly, in Herrstedt et al, 41 there was a significant decrease in HADS depression (but not anxiety), in the control group (post- vs pre-exercise). Baek et al, 39 however, reported no significant difference in depressive symptoms post-exercise in the control group. A possible explanation, as noted in Herrstedt et al 41 may be the effect of supervision of the exercise program, rather than the exercise itself, on depressive symptoms; this may apply to somatoform and anxiety symptoms as well. While studies could remove the supervision element from exercise, for example, home-based exercise programs (Pal et al 42 ) there is a trade-off of lower adherence/proper implementation of the program. Therefore, we note that for psychological outcomes, higher-powered studies with appropriate supervision levels are necessary to justify that exercise has a more pronounced effect on psychological outcomes in patients with psychiatric conditions, versus healthy controls.
Clinically, pharmacological modulation of the kynurenine pathway has proven challenging. Many drugs targeting the kynurenine pathway are under investigation. 25 However, studies on pharmacological intervention have stalled, 49 revealed appreciable amounts of side effects, 50 or have been demonstrated to be ineffective or have failed. 51 In addition, these drugs are likely to be for very specific use cases, for example, breast cancer, and may not be cost-effective. Exercise, however, can be implemented in more populations and is likely to be more cost-effective. As mentioned earlier, it is no surprise that pharmacological approaches will almost always carry the risk of side effects. 52 Given that many chronic diseases will still require pharmacological approaches, it is useful to know that exercise may attenuate these side effects53,54 that have a significant impact on people with chronic disease.
Some limitations of this review include the small number of studies, sample size, intervention variation, and the confounder of exercise itself. Among the small number of studies available for analysis, 3 (Herrstedt et al, 41 Pal et al, 42 Zimmer et al 43 ) were actually analyses of data obtained in previous clinical trials. This introduces elements of bias, and for our review, particularly unmeasured confounding 55 for temporally distant outcomes, and reporting of secondary analyses without mentioning of uncertainty. 56 As mentioned earlier, the controversy of acute versus chronic exercise resulting in differential kynurenine pathway activation, and the need to obtain samples within an appropriate time period, must not be understated. Given only the acute exercise study (Mudry et al 38 ) obtained samples within an appropriate time period, while the chronic exercise studies did not provide evidence of appropriate measurement protocols, it is not possible to conclude if long-term exercise has a long-term effect on the kynurenine pathway, and thus outcomes for age-related disease.
Small sample size, while convenient and cost-effective, presents problems. In Pal et al, 42 65 patients were eligible for the study, but only 32 were included for analysis due to blood sample quantity requirements. There are 2 important points to consider here: (1) future studies must measure exercise effects on the kynurenine pathway as a primary outcome (vs leaving the duty to a secondary investigator and analysis) and (2) when the sample size is already small and further decreased, the difference in absolute number (ie, number of participants) of outcomes makes it difficult to determine whether the effect is due to the intervention or simply chance. 57
Either a limitation in theory or a fortunate element in practice, “exercise” is highly variable. Most exercise studies create reproducible and objectively-measured regimens. However, unlike pharmacological interventions, it is inherently difficult to tailor specific exercise interventions with variable populations. Though most, if not all, of the studies included in this review reported significant results post- versus pre-exercise, or significant differences post-exercise between chronically ill and healthy/other groups, the studies have inherent variations between exercise and control groups that are hard to measure. In the case of Herrstedt et al, 41 the referenced study design’s control group was actually allowed to participate in any standard hospital/community-based exercise programs; while this may be interpreted as a loosely-defined control group that introduces too many confounding factors (which may have been prevented had all participant exercise programs in the control group been recorded), it may represent a new way to compare highly-structured exercise programs with exercise programs available in the community. Of course, we cannot overstate the importance of defining the methods pre-intervention and robust data collection of the exercise regimen (frequency, intensity, time, and type). 31 Without this data, outcomes comparisons between studies is invalid, if not nearly impossible. Future investigators may wish to make such similar comparisons, implement and record a set number of exercise programs to participants in one control group, and collect and analyse data in a pooled/unpooled setting all in one study. This would increase the number of exercise program comparisons, without introducing extra cost/time burden of designing multiple studies.
Perhaps the most pressing limitation, however, is the complexity of the relationships associated with the kynurenine pathway (see Figure 2). All our included studies reported significant effects of exercise on the kynurenine pathway, that is, evidence does exist for at least a unidirectional relationship between exercise and the kynurenine pathway. Only Hennings et al 40 and Herrstedt et al 41 measured psychological outcomes. However, it is widely known that exercise produces benefits on the measured mental health outcomes in our included studies: depressive symptoms, 58 somatoform symptoms, 59 anxiety symptoms. 60 Thus, the current evidence is unclear whether exercise affects psychological outcomes through the kynurenine pathway, or through some other mediator (see Figure 2), for example, supervision, 35 some other inflammatory pathway (eg, TNFɑ 22 ).
Schematic of kynurenine pathway and effects of exercise on kynurenine pathway elements and end outcomes.
We recommend that a future study be constructed with multiple intervention/control arms to elucidate these relationships backed up by robust evidence: multiple arms including kynurenine pathway pharmacological intervention (with placebo control) and exercise intervention (including “non-exercise” control), with outcomes including kynurenine pathway metabolites and psychological outcomes. Differences between kynurenine pathway pharmacological interventions and exercise interventions could then be assessed to determine if psychological outcomes are due to kynurenine pathway modulation or other confounders present during exercise interventions.
Conclusion
There are few studies investigating the effects of exercise on the kynurenine pathway and/or psychological outcomes associated with the kynurenine pathway. Of the studies that have been performed, it is easier to find studies performed in healthy volunteers34,61,62 without documented age-related disease versus studies conducted in people with age-related disease. In addition, this review did not explore other age-related diseases, particularly high-burden ones such as cardiovascular disease 63 and osteoporosis. 64 Currently, there are few reviews34,35,65 on the effects of exercise on the kynurenine pathway and potential mechanisms. Importantly, one review 65 mentions the possibility that long-term exercise interventions may only have measurable effects if age-related disease is present. Thus, further studies and reviews, particularly with a “prolonged duration” element (age-related disease and exercise intervention), are needed to establish a guideline for a combination 22 of lifestyle modification and pharmacological treatment that can be prescribed for diseases of inflammaging.
Footnotes
Funding:
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests:
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Author Contributions
AL - first investigator, paper writing
CH - second investigator, paper writing
SV - assistance with following: data analysis/interpretation, risk of bias assessment
GG - supervision, assistance with following: paper edits
GD - primary supervision, assistance with following: paper edits
