Targeting Rac1 for the prevention of atherosclerosis among U.S. Veterans with inflammatory bowel disease

ABSTRACT

Evidence suggests that Ras-related C3 botulinum toxin substrate 1 (Rac1) might be a target in atherosclerotic disease (AD). We hypothesize that due to their ability to inhibit Rac1, thiopurines are associated with a lower risk of AD. We fit a time-dependent cox proportional hazards model estimating the hazard of AD among a national cohort of US veterans with inflammatory bowel disease. Patients exposed to thiopurines had a 7.5% lower risk of AD (HR = 0.925; 95% CI = (0.87–0.984)) compared to controls. The propensity score weighted analysis reveals thiopurine exposure reduces the risk of AD by 6.6% (HR = 0.934; 95% CI = (0.896–0.975)), compared to controls. Further exploration and evaluation of Rac1 inhibition as a target for AD is warranted.

KEYWORDS:

• Rac1
• thiopurines
• atherosclerosis
• cohort
• inflammatory bowel disease

Introduction

Ras‐related C3 botulinum toxin substrate 1 (Rac1), a member of the Rho‐GTPases family of proteins, is implicated in a variety of diseases including neurologic conditions such as Alzheimer’s disease and depression. Recent evidence suggests that Rac1 might be a potential target in atherosclerotic disease (AD). Rac1 is an important signalling molecule in cardiovascular disease [1]. Specifically, Rac1 is activated by various cardiovascular agonists, and is involved with smooth muscle cell processes including migration, proliferation, and differentiation, all potentially associated with the development of atherosclerotic plaque [2–6]. Bandaru et al. 2020 studied atherosclerosis in vivo by infecting mice with AdPCSK9 (adenoviral vector overexpressing proprotein convertase subtilisin/kexin type 9) [7]. They found that Rac1 deficient mice had lower macrophage IL-6 and TNF-α levels, both of which are effective targets in cardiovascular disease [7,8].

Thiopurines (azathioprine (AZA) and mercaptopurine (6-MP)) are medications indicated for use in inflammatory bowel conditions, such as Crohn’s disease (CD) and ulcerative colitis (UC). Thiopurines suppress the immune system via inhibition of Rac1 [9–12]. Specifically, thiopurines inhibit proliferative and inflammatory processes in macrophages through both Rac1 independent and dependent mechanisms. The inflammatory response of macrophages is reduced in a Rac1 dependent manner in two ways. First, Rac1 is required for CD4⁺ T cell differentiation and activation with antigen presenting cells. AZA and 6-MP inhibit Rac1 activity in CD4⁺ T cells by blocking their activation and consequently minimizing inflammation [13]. Second, 6-MP reduces inducible nitric oxide synthase (iNOS) expression via Rac1 inhibition [10]. Oxidative environments highly induce iNOS expression, which can lead to pathological nitric oxide (NO) production [14]. Reduction of iNOS through Rac1 inhibition can possibly contribute to suppression of oxidative stress, which is frequently implicated in atherosclerotic and cardiovascular disease [15,16]. Furthermore, in gut epithelial cells, thiopurines reduce the proinflammatory chemokines IL-8, via Rac1 inhibition, and CCL2 in a Rac1 independent manner. Overexpression of CCL2 and IL-8 are implicated in atherosclerotic disease development [17–19].

Given the pre-clinical link between Rac1 inhibition and atherosclerotic disease, we hypothesize that patients exposed to thiopurines, potent Rac1 inhibitors, will have a lower risk of atherosclerotic disease. We test this hypothesis using a national cohort of patients treated by the U.S. Department of Veteran’s Affairs.

Methods

Data

This drug disease association study was conducted using data from the Department of Veteran Affairs (VA). The study period was from January 2000 to 30 September 2020 and data were extracted from the VA Informatics and Computing Infrastructure (VINCI), which includes inpatient, outpatient data (coded with International classification of diseases (ICD) revision 9-CM, revision 10-CM), and pharmacy claims. The study was conducted in compliance with the Department of Veterans Affairs requirements, received Institutional Review Board, and Research & Development approval.

Cohort creation

This retrospective cohort study included patients diagnosed with inflammatory bowel disease (IBD). Patients with IBD were extracted from the VA outpatient and inpatient claims files using, ICD-9 CM 555.x, 556.x or ICD-10 CM K50.x, K51.x with their first diagnosis between January 2000 and October 2019. Patients were included in the study if they had no thiopurine, biologic or methotrexate use prior to their first IBD diagnosis. Patients were excluded if they had atherosclerosis prior to first IBD diagnosis or if they had less than 1 year of follow-up after medication start. If patients never initiated treatment with a study medication, they were required to have at least one year of follow-up from IBD diagnosis. Patients were followed from index until the first date of a) atherosclerosis diagnosis, b) treatment switch to another study treatment type (e.g. thiopurine to a biologic), c) death, or d) 30 September 2020. Patients were categorized into mutually exclusive cohorts by their first medication exposure (thiopurine, biologic, methotrexate, no medication exposure).

Exposure and outcome coding

To account for immortal time bias, treatment exposure was coded dynamically. Patients are considered unexposed until their first medication dispense after which they are considered exposed until study endpoint. Thiopurine medications include mercaptopurine and azathioprine. Biologic medications include adalimumab, certolizumab, infliximab, ustekinumab and vedolizumab. Patients without exposure to the above listed study medications were included in the control cohort. All medications including thiopurines, biologics and methotrexate were extracted from the VA outpatient pharmacy data. Atherosclerotic disease was indicated with three different outcome codings. First, we use the atherosclerotic cardiovascular disease coding (ASCVD) which includes acute coronary syndromes, angina diagnoses, coronary as well peripheral revascularization procedures, peripheral artery disease, ischaemic stroke, and transient ischaemic attack diagnoses [20,21]. All codes used for ASCVD are found in supplementary table S1. Second, we use an outcome with all diagnosis codes for atherosclerosis and peripheral vascular disease (PVD) (ICD-9-CM 440.x, 443.9, 414.0, 414.2, 414.3, 414.4; ICD-10-CM: I70.x, I73.9, I25.1, I25.7, I25.8). Third, we use diagnosis codes for select atherosclerosis diagnoses (e.g. aortic atherosclerosis sub code 440.0, and atherosclerosis of native arteries with claudication 440.21 opposed to all 440.x codes used above) and PVD (ICD-9 CM 440.0,440.21, 443.9, 414.0, 414.2, 414.3, 414.4 or ICD-10 CM I70.0, I70.2, I73.9, I25.1, I25.7, I25.8) extracted from the VA inpatient and outpatient claims data. The multiple outcome coding provides sensitivity analysis on the robustness of the associations found.

Baseline data

We account for important baseline factors such as demographic (age, race, sex), comorbid factors that may affect development of AD such as the Charlson comorbidity score, body mass index (BMI), smoking, hypertension, hyperlipidaemia, chronic kidney disease and diabetes. All comorbid factors except BMI were pulled via ICD-9 and ICD-10 CM codes. BMI was calculated from height and weight data extracted from the VA vital signs data file.

Statistical analysis

To analyse the association between thiopurine medication use and atherosclerosis, we generated summaries of the baseline demographic, comorbid, and clinical characteristics for each cohort. To evaluate differences across cohorts, we use p-values calculated from the chi-square test or ANOVA F-test, where appropriate, as well as the standardized difference. We calculated the standardized difference as the differences between the treatment means divided by the pooled standard deviation. We estimated a time-dependent cox proportional hazards model to quantify the hazard of AD for each cohort. Because treatment assignment is not randomized, we utilized inverse probability treatment weights to minimize potential bias. We used a generalized boosted model (GBM) to estimate the propensity weights [22,23]. Generalized boosted models consist of multiple regression trees, which are then aggregated into one final model. We used the standardized mean difference as the model stopping rule. All covariates in Table 2 were included in the propensity score model. We then estimated a weighted time-dependent cox proportional hazards model including all variables to provide doubly robust estimates.Table 1

Table 1. Sample attrition

Inclusion criteria	N
Patients with a IBD diagnosis	119,281
Patients without treatment prior to Dx	117,175
Patients no PAD prior to	100,159
Patients with at least 1 year of follow-up	91,938

This table displays the inclusion criteria and the subsequent sample size resulting in the final sample size of 91,938.

Table 2. Baseline sample characteristics

		Biologics	Methotrexate	Control	Thiopurines		Maximum Absolute standardized difference
	Variable	N = 2,862	N = 820	N = 79,600	N = 8,116	P-value†
Age		44.93 (16.08)	51.42 (15.31)	53.03 (15.83)	50.29 (16.36)	<0.001	0.507
Race	Black	390 (13.63%)	129 (15.73%)	17,592 (22.1%)	925 (11.4%)	<0.001	0.263
	Other/unknown	153 (5.35%)	61 (7.44%)	8269 (10.39%)	738 (9.09%)		0.166
	White	2319 (81.03%)	630 (76.83%)	53,739 (67.51%)	6453 (79.51%)		0.291
Sex	Female	341 (11.91%)	106 (12.93%)	7458 (9.37%)	671 (8.27%)	<0.001	0.16
	Male	2521 (88.09%)	714 (87.07%)	72,142 (90.63%)	7445 (91.73%)		0.16
BMI	<18.5	43 (1.5%)	9 (1.1%)	1068 (1.34%)	159 (1.96%)	<0.001	0.073
	18.5–24.9	854 (29.84%)	206 (25.12%)	21,176 (26.6%)	2410 (29.69%)		0.106
	25–29.9	1076 (37.6%)	344 (41.95%)	31,123 (39.1%)	3144 (38.74%)		0.089
	30+	876 (30.61%)	259 (31.59%)	25,659 (32.23%)	2364 (29.13%)		0.067
	Missing	13 (0.45%)	2 (0.24%)	574 (0.72%)	39 (0.48%)		0.057
Charlson comorbidity		0.24 (0.75)	0.36 (0.95)	0.62 (1.44)	0.22 (0.7)	<0.001	0.29
Smoking		976 (34.1%)	340 (41.46%)	33,454 (42.03%)	3035 (37.4%)	<0.001	0.161
Hypertension		316 (11.04%)	139 (16.95%)	17,833 (22.4%)	1015 (12.51%)	<0.001	0.277
Chronic kidney disease		21 (0.73%)	6 (0.73%)	1018 (1.28%)	37 (0.46%)	<0.001	0.076
Hyperlipidaemia		319 (11.15%)	122 (14.88%)	15,317 (19.24%)	865 (10.66%)	<0.001	0.221
Diabetes		153 (5.35%)	55 (6.71%)	7466 (9.38%)	403 (4.97%)	<0.001	0.155
Year		2011.67 (5.2)	2007.16 (5.75)	2007.55 (5.76)	2006.64 (5.4)	<0.001	0.872

† P-values are from either the chi-square test(for categorical variables) or anova model (continuous variables).

Results

A total of 91,938 patients met the inclusion criteria (Table1) consisting of 2,862 patients treated with biologics, 820 patients treated with methotrexate, 8,116 patients treated with thiopurines, and 79,600 patients with no study medication (control). Table 2 displays the baseline sample characteristics. In each cohort, patients were, on average, middle-aged white males with a BMI between 25 and 29.9. The no medication cohort had, on average, a higher Charlson comorbidity index, a higher frequency of smokers, hypertension, chronic kidney disease, hyperlipidaemia, and diabetes compared to the other cohorts. The sample characteristics for the weighted samples appear in Table 3. Overall, the propensity score methods resulted in well-balanced samples where all standardized differences are below 0.15.

Table 3. Weighted Sample Characteristics

Variable		Biologics	Methotrexate	No medication	Thiopurines	Maximum absolute standard difference
Age		50.66	52.13	52.12	52.11	0.092
Race	Black	19%	19%	20%	19%	0.040
	Other/unknown	8%	9%	10%	10%	0.054
	White	73%	72%	70%	71%	0.067
Sex	Female	9%	9%	9%	9%	0.008
	Male	91%	91%	91%	91%	0.008
BMI	<18.5	1%	1%	1%	1%	0.026
	18.5–24.9	27%	26%	27%	27%	0.038
	25–29.9	38%	40%	39%	39%	0.031
	30+	33%	33%	32%	32%	0.029
	Missing	0%	0%	1%	1%	0.083
Charlson comorbidity		0.34	0.42	0.53	0.46	0.143
Smoking		41%	41%	41%	40%	0.014
Hypertension		19%	18%	20%	19%	0.050
Chronic kidney disease		1%	1%	1%	1%	0.027
Hyperlipidaemia		16%	16%	17%	17%	0.045
Diabetes		7%	8%	8%	8%	0.035
Year		2008.04	2007.49	2007.63	2007.7	0.095

The time-dependent Cox proportional hazards model estimates, both original and IPTW weighted, are shown in Table 4. Thiopurine exposure is consistently associated with a statistically significantly lower risk of atherosclerosis across all outcome definitions. In terms of the ASCVD coding, thiopurine exposure is associated with a 7.5% lower risk, unweighted model (HR = 0.925 95% CI = (0.87–0.984)) and a 6.6% reduction, IPTW weighted model (HR = 0.934; 95% CI = (0.896–0.975)), compared to controls. Similarly, using all atherosclerosis diagnosis codes, we found that thiopurine exposure is associated with a 6.7% lower risk, unweighted model (HR = 0.933 95% CI = (0.874–0.997)) and a 8% reduction, IPTW weighted model (HR = 0.92; 95% CI = (0.852–0.993)), compared to controls. In the outcome with coronary atherosclerosis or PVD diagnosis, we found that thiopurine exposure reduces by the risk by 12% (HR = 0.88; 95% CI = (0.83–0.94)) compared to controls. The IPTW weighted analysis reveals thiopurine exposure reduces the risk by 13% (HR = 0.87; 95% CI = (0.81–0.94)) compared to controls. Other covariates associated with a lower risk of AD include patients of other/unknown race (compared to blacks), patients with BMI 18.5–24.9 (compared to BMI <18.5), and more recent index years. However, some covariates were associated with a higher risk of AD which consists of whites (compared to blacks), males (compared to females), patients with BMI 30+ (compared to BMI <18.5), and those patients with increasing Charlson comorbidity index. Also, patients who smoke or have hypertension, chronic kidney disease, hyperlipidaemia, or diabetes have a higher risk.

Table 4. Time-dependent cox proportional hazards model: Hazard ratio (HR) and 95% confidence intervals

	ASCVD coding		All atherosclerosis/PVD diagnosis		Coronary atherosclerosis/PVD diagnosis
	Original	IPTW weighted	Original	IPTW weighted	Original	IPTW weighted
Biologics	0.972 (0.839–1.125)	0.901 (0.858–0.946)	0.943 (0.8–1.11)	0.95 (0.779–1.16)	0.95 (0.81–1.11)	0.93 (0.77–1.12)
Methotrexate	1.037 (0.854–1.26)	0.987 (0.944–1.032)	1.02 (0.824–1.263)	0.961 (0.769–1.201)	0.99 (0.81–1.22)	0.95 (0.77–1.17)
Thiopurines	0.925 (0.87–0.984)	0.934 (0.896–0.975)	0.933 (0.874–0.997)	0.92 (0.852–0.993)	0.88 (0.83–0.94)	0.87 (0.81–0.94)
Age	1.039 (1.038–1.041)	1.039 (1.037–1.04)	1.043 (1.042–1.045)	1.043 (1.039–1.048)	1.04 (1.04–1.04)	1.04 (1.04–1.05)
Race other/unknown vs. black	0.697 (0.654–0.744)	0.73 (0.679–0.784)	0.77 (0.715–0.829)	0.743 (0.547–1.009)	0.74 (0.70–0.80)	0.80 (0.60–1.06)
white vs. black	0.973 (0.937–1.01)	0.981 (0.939–1.023)	1.081 (1.035–1.13)	1.039 (0.854–1.263)	1.06 (1.02–1.10)	1.07 (0.90–1.27)
Male	1.322 (1.235–1.414)	1.131 (1.056–1.211)	1.697 (1.557–1.85)	1.782 (1.359–2.337)	1.65 (1.52–1.78)	1.73 (1.34–2.23)
BMI 18.5–24.9 vs <18.5	0.893 (0.778–1.025)	0.977 (0.832–1.147)	0.883 (0.753–1.034)	0.912 (0.544–1.531)	0.86 (0.74–0.99)	1.05 (0.72–1.53)
25–29.9 vs <18.5	0.99 (0.864–1.136)	0.985 (0.84–1.155)	1.002 (0.856–1.173)	0.987 (0.59–1.651)	0.96 (0.83–1.10)	1.22 (0.84–1.78)
30+ vs. <18.5	1.17 (1.019–1.342)	1.085 (0.924–1.274)	1.269 (1.084–1.486)	1.265 (0.758–2.114)	1.20 (1.04–1.38)	1.48 (1.02–2.15)
Missing vs. <18.5	0.711 (0.539–0.937)	0.745 (0.535–1.039)	0.686 (0.495–0.95)	1.058 (0.449–2.494)	0.66 (0.49–0.89)	1.09 (0.51–2.34)
Charlson comorbidity	1.042 (1.03–1.053)	1.052 (1.037–1.067)	1.037 (1.025–1.05)	1.042 (0.979–1.108)	1.05 (1.03–1.06)	1.05 (0.99–1.11)
Smoking	1.497 (1.452–1.544)	1.52 (1.47–1.572)	1.506 (1.455–1.559)	1.632 (1.433–1.859)	1.52 (1.48–1.57)	1.70 (1.51–1.92)
Hypertension	1.351 (1.303–1.4)	1.147 (1.1–1.195)	1.329 (1.277–1.383)	1.004 (0.848–1.188)	1.33 (1.28–1.38)	1.06 (0.91–1.24)
Chronic kidney disease	1.328 (1.171–1.506)	1.443 (1.24–1.679)	1.352 (1.188–1.539)	1.166 (0.691–1.969)	1.33 (1.18–1.50)	1.16 (0.71–1.88)
Hyperlipidaemia	1.061 (1.019–1.104)	1.067 (1.02–1.117)	1.062 (1.016–1.109)	1.145 (0.965–1.358)	1.05 (1.01–1.10)	1.17 (0.99–1.38)
Diabetes	1.445 (1.375–1.518)	1.455 (1.375–1.541)	1.477 (1.4–1.559)	1.435 (1.143–1.801)	1.56 (1.49–1.64)	1.38 (1.11–1.71)
Year	0.984 (0.98–0.987)	1.004 (1–1.008)	0.998 (0.994–1.002)	1.014 (0.999–1.03)	0.98 (0.97–0.98)	1.00 (0.98–1.01)

PVD: peripheral vascular disease.

Discussion

Pre-clinical data suggest that Rac1 as a therapeutic target for atherosclerosis, which is considered an inflammatory condition. Inhibiting Rac1 with thiopurines could suppress the inflammation associated with atherosclerosis development. In vivo studies show that Rac1 deficient mice had lower macrophage IL-6 and TNF-α levels, both of which are effective targets in cardiovascular disease. Thiopurines could reduce atherosclerosis development via inhibiting inflammatory processes in a Rac1 dependent and independent manner. By inhibiting Rac1, thiopurines are able to block CD4⁺ T cell activation and subsequently minimize inflammation [13], reduce inducible nitric oxide synthase (iNOS) expression [10], and reduce proinflammatory cytokines [17–19]. Given the preclinical evidence showing thiopurines inhibit Rac1, a possible therapeutic target in atherosclerosis, we hypothesized that thiopurine exposure is associated with a lower risk of atherosclerotic disease development. Our analysis, utilizing appropriate statistical methods, identified an association between exposure to thiopurines and lower risk of incident atherosclerosis among a cohort of US veterans with inflammatory bowel disease. Our result is consistent with the pre-clinical data suggesting Rac1 inhibition should lower the risk of atherosclerosis.

The primary limitation of this study is non-random treatment assignment, common to all retrospective observational studies. However, our analysis attempted to mitigate confounding due to non-randomization of treatment via propensity score weighting. Propensity score analysis is a common method used to create causal estimates from observational data [24–27]. Our propensity score weighted results are consistent with the unweighted model suggesting a protective effect of thiopurine exposure. Furthermore, we utilized robust statistical techniques such as time-dependent cox models to avoid immortal time bias, which biases estimates towards being more protective [28,29]. Importantly, this study cannot definitively link the inhibition of Rac1 with the lower risk of atherosclerosis. Rather, the scientific premise of this research rests on the prior findings that thiopurines inhibit Rac1 [9–12] and that Rac1 could be a novel target in the treatment of cardiovascular disease [2–8]. As such, this study represents an exploration of the role of Rac1 on incident atherosclerosis. This study is in no way recommending the utilization of thiopurines to prevent atherosclerotic diseases. This study was conducted to compliment pre-clinical work on the role of Rac1 in AD using existing administrative claims data.

The relative strengths of our study include the large sample size of patients studied within the veteran’s healthcare system, one of the largest integrated systems in the United States and the statistical methods used to minimize potential bias in the results. Further, our analysis is consistent with the pre-clinical data linking Rac1 to atherosclerosis and provides evidence warranting additional research into this potential novel target for the treatment or prevention of atherosclerosis.

Conclusion

Thiopurine exposure is associated with a lower risk of incident atherosclerotic disease. These results suggest that targeting inflammatory and immune processes through Rac1 inhibition could be a promising therapeutic strategy for the prevention of atherosclerosis and warrants further investigation.

Acknowledgments

No funding agency had a role in study design or conduct, data collection, analysis, interpretation, or manuscript writing. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the US Department of Veterans Affairs, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government. This paper represents original research conducted using data from the Department of Veterans Affairs and is, in part, the result of work supported with resources and the use of facilities at the Dorn Research Institute, Columbia VA Health Care System, Columbia, South Carolina.

Disclosure statement

S.S.S. has received research grants from Boehringer Ingelheim, Gilead Sciences, Portola Pharmaceuticals, and United Therapeutics unrelated to this work. Other authors declare no competing interests.

Data availability statement

Analyses of the Veterans Health Administration Database were performed using data within the US Department of Veterans Affairs secure research environment, the VA Informatics and Computing Infrastructure (VINCI) and cannot be extracted. All relevant data outputs are within the paper.

Supplementary material

Supplemental data for this article can be accessed here