Keywords abdominal aortic aneurysm - D-dimer - diagnosis - myeloperoxidase - prognosis
Introduction
Abdominal aortic aneurysm (AAA) represents a major cause of cardiovascular mortality
despite advancements in surgical techniques. It is frequently asymptomatic, diagnosed
incidentally and associated with a high death rate at rupture.[1 ] Simple low-cost biomarkers for early disease detection would be beneficial in countries
where ultrasound screening for high-risk groups is not implemented. Importantly, prognostic
markers are of particular interest because rapid expansion of small aneurysms is difficult
to predict but associated with a high risk of rupture.[1 ] Blood-borne diagnostic parameters which have repeatedly shown association with AAA
are common biomarkers of cardiovascular disease such as D-dimer, fibrinogen[2 ]
[3 ] or C-reactive protein.[4 ] Among these, D-dimer was reported to be related to the occurrence of aortic aneurysm
independent of underlying atherosclerotic disease,[5 ] to correlate with AAA and intraluminal thrombus (ILT) size[6 ] and to predict aneurysm growth.[7 ]
[8 ]
Histologically, the AAA wall is characterized by destruction of the vascular media
associated with loss of elastic fibres and smooth muscle cells (SMCs).[9 ] Prominent inflammatory infiltrates in the adventitia as well as the ILT exert wall-weakening
activity which is primarily attributed to myeloid cells and their production of reactive
oxygen species (ROS) and proteolytic enzymes.[10 ] Several substances in the ILT and AAA wall which originate from activated neutrophils
and monocytes can be detected at increased levels in the peripheral blood of AAA patients
and have been investigated for their biomarker potential. These include: extracellular
matrix-degrading enzymes such as cathepsins,[11 ] matrix metalloproteinases, elastase,[10 ] neutrophil gelatinase-associated lipocalin[12 ] and inflammatory cytokines.[13 ] In contrast to these soluble factors, phenotypic alterations of the circulating
myeloid cell populations are largely unexplored in the context of AAA. Monocytes form
functionally distinct sub-sets that are recognized based on their differential surface
expression of CD14 and CD16.[14 ] There is a progressive differentiation from classical (CD14++ CD16−) via intermediate
(CD14++ CD16+) to non-classical (CD14+ CD16++) monocytes.[15 ] Classical monocytes were identified as potent phagocytic cells[16 ]
[17 ] with the capacity to down-regulate the pro-inflammatory reactions of the two CD16+
sub-sets.[18 ] The non-classical population exhibited patrolling behaviour maintaining vessel integrity,[19 ]
[20 ] while intermediate monocytes were proposed to have the highest capacity for antigen
presentation, angiogenesis and tissue remodelling.[16 ]
[17 ] Furthermore, intermediates showed the highest propensity to form monocyte–platelet
aggregates (MPAs), which are of relevance to thrombotic disease.[21 ] Increased levels of intermediate and non-classical monocytes were found to be associated
with a higher risk of cardiovascular events[22 ] as well as large AAAs.[23 ]
[24 ]
Comparably, the existence of neutrophil sub-sets with distinct activation status and
function has been proposed which are defined by the expression of CD62L and CD16.[25 ] Mature, quiescent circulating neutrophils are generally characterized by a high
level of both surface markers (CD16high CD62Lhigh ). The immature CD16low CD62Lhigh neutrophil sub-set which is newly released from bone marrow is marked by lower phagocytic
activity and decreased ROS-generating capability, while the activated CD16high CD62Llow sub-set is strongly pro-inflammatory and exerts main immunoregulatory functions.[25 ]
[26 ] Both, activated as well as newly released sub-sets are induced in endotoxaemia and
sepsis,[25 ] but relevant studies in other disease-specific contexts, including AAA, are lacking.
However, the interaction of neutrophils with platelets has been shown to play an important
role in AAA pathogenesis, enhancing neutrophil recruitment and activation.[10 ]
Based on the reports of neutrophil and monocyte involvement in AAA, we postulated
that the state of chronic inflammation is not restricted to the local AAA site but
is reflected in a systemic change in myeloid cell populations indicating on-going
leukocyte activation and recruitment. We thus assessed the marker potential of circulating
myeloid cell sub-sets and their secreted factors in comparison to previously reported
AAA parameters such as D-dimer. A diagnostic risk score was devised and challenged
in a validation cohort. Of particular interest, the risk score further proved effective
in predicting rapid disease progression.
Methods
Diagnostic Study Design
We conducted an observational case–control study, adhering to the principles of the
Declaration of Helsinki and the STROBE guidelines for reporting results. The study
was approved by the institutional ethics committee (license no. 1729/2014) and informed
consent was obtained from participating patients at the Vienna General Hospital (2014–2016).
Morphometric AAA analysis was performed using computed tomography angiography (CTA)
images at study inclusion. A control group was recruited from general surgery, urology
and ophthalmology patients (presenting for routine check-ups) with ultrasound-confirmed
absence of AAA. Subjects with recent (< 1 year) tumour and/or chemotherapy, systemic
autoimmune or haematological disease and organ transplant recipients were excluded.
In the present evaluation, data collected from 41 patients prior to immanent elective
surgical repair via open or endovascular approach were compared with the control group
(N = 38), which was matched in age, sex, body mass index and smoking habit.
For validation of the diagnostic score, citrated plasma samples collected at an independent
institution (Leeds General Infirmary as part of the Leeds Aneurysm Development Study,
approved by the Leeds East Ethics Committee, Ref 03/142[27 ]
[28 ]
[29 ]
[30 ]
[31 ]) were analysed which represented 63 AAA cases and 63 controls, recruited from other
medical and surgical outpatient departments at the same hospital. All patients received
abdominal aortic ultrasound for confirmation of diagnosis. For these samples, one-to-one
matching was performed for age (within 2 years), sex and the presence of clinical
cardiovascular disease (previous myocardial infarction, angina pectoris, peripheral
vascular disease or cerebrovascular disease).
Prognostic Study Design
Based on the mentioned inclusion and exclusion criteria, AAA patients without surgical
indication were enrolled in a longitudinal observational study (2014–2018) at the
Vascular Surgery Department of the Medical University of Vienna, with serial blood
withdrawing and CTA analysis every 6 months. Thirty-three patients were assessed,
with a total of 68 six-month monitoring periods.
Flow Cytometry of Myeloid Cell Populations
Blood was collected into hirudin containing tubes (Roche, Basel, Switzerland) and
fixed immediately with ThromboFix solution (Beckman-Coulter, Indianapolis, Indiana,
United States) according to manufacturer's instructions. After 2 to 7 hours at room
temperature, leukocyte populations were stained in whole blood using the following
antibodies: CD66b-FITC, CD62L-AF647, CD16-PE/Cy7, CD41-PB (BioLegend, San Diego, California,
United States) and CD54-PE (Thermo Fisher Scientific, Waltham, Massachusetts, United
States) for neutrophil sub-sets; CD14-FITC, CD16-PE/Cy7, HLA-DR-PE and CD41-PB (BioLegend)
for monocyte sub-sets. 7-AAD dye (Beckman-Coulter) was used to exclude non-viable
cells. Following erythrocyte lysis with RBC Lysis Buffer (Thermo Fisher Scientific),
samples were diluted and immediately analysed with a Gallios flow cytometer (Beckman-Coulter).
Neutrophil sub-sets were evaluated as viable CD66b+ cells and distinguished from eosinophils
(CD16–). Sub-set discrimination was based on CD62L and CD16 expression levels ([Supplementary Figs. S1 ] and [S2 ], available in the online version), identifying mature, quiescent neutrophils as
CD16high CD62Lhigh , and the activated neutrophil population as CD16high CD62Llow as opposed to newly released neutrophils with CD16low CD62Lhigh surface expression. Monocyte sub-sets were quantified from viable HLA-DR+ CD14+ cells
using CD14 and CD16 surface levels ([Supplementary Fig. S3 ], available in the online version) for the distinction between classical (CD14++
CD16−), intermediate (CD14++ CD16+) and non-classical (CD14+ CD16++) monocytes. Both,
neutrophils and monocytes were additionally analysed for the proportion of platelet
(CD41+) aggregates. A blood count was performed using the same hirudin-treated blood
sample on a Sysmex XN-350 haemocytometer (Sysmex, Kobe, Japan) which was applied to
calculate absolute concentrations of leukocyte sub-sets.
Since standard methods were applied to measure plasma and tissue-released factors,
the respective methods are described in the [Supplementary Methods ] (available in the online version).
Statistical Evaluation
Data are generally given as median values and interquartile range (IQR). Non-parametric
tests were used for group comparisons (Mann–Whitney U, Wilcoxon signed-rank test)
or correlations (Spearman's r ). Contingency tables and chi-square test were applied for categorical variables.
Receiver operating characteristics (ROC) analysis served to evaluate the diagnostic
or prognostic marker potential. These analyses were performed with SPSS 24.0 software
(IBM, Armonk, New York, United States) and a significance level of p < 0.05 was applied. Sample size calculation was based on the previously reported
frequency of CD16+ monocytes in AAA patients and healthy controls[23 ] with a 1.4-fold difference in mean parameter levels at a high biological variation
(μ1 = 11.0%, μ2 = 15.1%, σ= 6.1%, α= 0.05, 1–β= 0.80) and yielded a minimal sample
size of 36 per group. Missing values were generally below 5% and hence not addressed
in statistical analysis.
Logistic regression models were developed with R software (R Core Team 2016, Vienna,
Austria) using stepwise backward elimination of insignificant variables and the Akaike
information criterion as the stopping criterion. The resulting diagnostic model was
internally validated, using fivefold cross-validated predicted probabilities. The
validation was repeated 100 times and results were averaged to avoid dependence on
a specific data split. The area under the ROC curve was deduced from ROC analyses
comparing the predicted probabilities and the observed event status. Finally, a coefficient
of discrimination[32 ] was computed as the difference in average (cross-validated) predicted probabilities
between patients diagnosed with AAA and individuals without AAA.
The developed diagnostic score was further evaluated in the original dataset by multivariable
binary logistic regression including co-morbidities and patient medication as categorical
variables: coronary heart disease (CHD), anti-platelet therapy, hypertension and/or
anti-hypertensive therapy, hyperlipidaemia and/or lipid-lowering therapy. (Due to
the inherent correlation of hypertension and hyperlipidaemia with their pertaining
therapies, the combined variables were applied.) The score was further characterized
in the above-mentioned, independent validation set by ROC curve analysis (SPSS 24.0).
Results
Patient and Control Group Differ Significantly in the Reported Haematological Biomarker
D-dimer, in AAA-Associated Co-Morbidities and Medication
Forty-one AAA patients and 38 control subjects were included in the analysis. Groups
were comparable in terms of: sex distribution, age, smoking status and body mass index
([Tables 1 ], [2 ], [3 ]). There were significant differences in the diagnosis of hypertension (p = 0.026) and prescription of anti-hypertensive therapy (p = 0.051), which were more frequent in the AAA cohort. More aneurysm patients than
controls were affected by hyperlipidaemia (p = 0.001) and under lipid-lowering treatment (p < 0.001). The AAA group had lower total cholesterol (p = 0.037) and high-density lipoprotein (p = 0.030). The groups further differed significantly in the occurrence of CHD (p = 0.012), with AAA patients having an almost threefold higher frequency of anti-platelet
drug use (p < 0.001). A higher cardiovascular burden in the AAA group was also evidenced by the
Framingham Risk Score in per cent 10-year risk of general cardiovascular disease (p = 0.026). Other co-morbidities such as chronic obstructive pulmonary disease and
nephropathy (mainly renal cysts) showed a trend but no significant elevation in the
disease collective. Nevertheless, kidney parameters indicated a poorer kidney function
in AAA patients: elevated blood urea nitrogen (p = 0.001) and creatinine (p = 0.059) and reduced estimated glomerular filtration rate (p = 0.045).
Table 1
Patient and control demographics: categorical variables
Characteristic
Healthy (N = 38)
AAA (N = 41)
p -Value
N (%)
N (%)
Sex
0.390
Female
5 (13.2)
3 (7.3)
Male
33 (86.8)
38 (92.7)
Smoker status
Never
8 (21.1)
4 (9.8)
0.160
Past
13 (34.2)
22 (53.7)
Current
16 (42.1)
14 (34.1)
Unknown
1 (2.6)
1 (2.4)
Hypertension
23 (60.5)
34 (82.9)
0.026
Hyperlipidaemia
11 (28.9)
27 (65.9)
0.001
Peripheral artery disease
No
34 (89.5)
32 (78.0)
0.250
Yes
4 (10.5)
7 (17.1)
Unknown
0 (0.0)
2 (4.9)
Coronary heart disease
4 (10.5)
14 (34.1)
0.012
Myocardial infarction
3 (7.9)
9 (22.0)
0.082
Stroke
0 (0.0)
3 (7.3)
0.236
Diabetes mellitus
4 (10.5)
9 (22.0)
0.171
COPD
5 (13.2)
13 (31.7)
0.050
Nephropathy/renal cysts
No
31 (81.6)
27 (65.9)
0.227
Yes
7 (18.4)
13 (31.7)
Unknown
0 (0.0)
1 (2.4)
AAA family history
No
34 (89.5)
34 (82.9)
0.236
Yes
4 (10.5)
4 (9.8)
Unknown
0 (0.0)
3 (7.3)
Abbreviations: AAA, abdominal aortic aneurysm; COPD, chronic obstructive pulmonary
disease.
Table 2
Patient and control demographics: metric variables
Characteristic
Healthy
AAA
p -Value
Median (Range)
Median (Range)
Age [y]
67 (39–83)
71 (52–83)
0.705
Body mass index [kg/m2 ]
26.2 (20.7–40.1)
28.4 (17.8–34.1)
0.370
Median (IQR)
Median (IQR)
White blood cells [×106 /L]
5650 (2100)
6400 (2190)
0.114
Red blood cells [×1012 /L]
4.70 (0.68)
4.52 (0.73)
0.441
Haemoglobin [g/dL]
14.2 (2.7)
14.6 (2.4)
0.543
Haematocrit [%]
41.6 (6.1)
41.5 (6.7)
0.669
Platelets [×109 /L]
178.5 (91)
157 (80)
0.101
Lymphocytes [×106 /L]
1641 (818)
1611 (989)
0.322
Monocytes [×106 /L]
518 (292)
497 (356)
0.928
Neutrophils [×106 /L]
3375 (1535)
3742 (1350)
0.217
C-reactive protein [mg/dL]
0.26 (0.46)
0.40 (0.43)
0.075
Fibrinogen - Clauss [mg/dL]
375 (90)
396 (143)
0.258
D-dimer [µg/mL]
0.47 (0.49)
1.30 (1.68)
< 0.001
Lipoprotein (a) [nmol/L]
20.0 (39.0)
19.5 (115.0)
0.482
Triglycerides [mg/dL]
121 (73)
141 (59)
0.070
Total cholesterol [mg/dL]
198 (23)
170 (60)
0.037
LDL cholesterol [mg/dL]
119 (47)
93 (71)
0.171
HDL cholesterol [mg/dL]
54 (25)
50 (17)
0.030
Total cholesterol/HDL ratio
3.6 (2.3)
3.6 (1.4)
0.697
LDL/HDL ratio
2.2 (1.7)
2.0 (1.4)
0.996
FRS [risk factors]
0.86 (1.37)
1.13 (0.95)
0.064
FRS [%]
22.9 (23.8)
30.4 (25.7)
0.026
Creatinine [μg/dL]
0.94 (0.22)
1.02 (0.45)
0.059
Blood urea nitrogen [mg/dL]
15.1 (4.1)
17.8 (6.4)
0.001
eGFR [mL/min/1.73 m2 ]
84.7 (20.0)
76.0 (38.7)
0.045
Abbreviations: AAA, abdominal aortic aneurysm; eGFR, estimated glomerular filtration
rate; FRS, Framingham Risk Score; HDL, high-density lipoprotein; IQR, interquartile
range; LDL, low-density lipoprotein.
Table 3
Patient and control medication
Characteristic
Healthy (N = 38)
AAA (N = 41)
p -Value
N (%)
N (%)
Anti-platelet therapy
12 (31.6)
36 (87.8)
< 0.001
Acetylsalicylic acid
11 (28.9)
36 (87.8)
< 0.001
Clopidogrel
1 (2.6)
3 (7.3)
0.343
Prasugrel or Ticagrelor
0 (0.0)
2 (4.9)
0.386
Anticoagulation therapy
6 (15.8)
9 (22.0)
0.485
Anti-hypertensive therapy
23 (60.5)
33 (80.5)
0.051
ACE inhibitor
5 (13.2)
12 (29.3)
0.082
Angiotensin receptor blocker
10 (26.3)
16 (39.0)
0.230
Beta blocker
11 (28.9)
21 (51.2)
0.044
Calcium channel blocker
7 (18.4)
11 (26.8)
0.373
Nitrate
0 (0.0)
1 (2.4)
0.333
Diuretic
8 (21.1)
16 (39.0)
0.083
Lipid-lowering therapy
6 (15.8)
34 (82.9)
< 0.001
Statins
6 (15.8)
33 (80.5)
< 0.001
Diabetic medication
4 (10.5)
8 (19.5)
0.266
Insulin
1 (2.6)
1 (2.4)
0.957
Metformin
3 (7.9)
7 (17.1)
0.220
Hormones
3 (7.9)
6 (14.6)
0.346
Gout medication
4 (10.5)
6 (14.6)
0.583
Abbreviations: AAA, abdominal aortic aneurysm; ACE, angiotensin-converting enzyme.
In the AAA patient group, the median maximum AAA diameter reached 57 mm (range: 44–114
mm) with a median aneurysm length of 92 mm (38–147 mm). Regarding the ILT, a median
maximum thickness of 22 mm (5–45 mm) and volume of 75 mm3 (2–347 mm3 ) was recorded ([Table 4 ]). Only 5 of 41 patients showed a concomitant thoracic aortic aneurysm; the AAA form
was predominantly fusiform (63%).
Table 4
Aneurysm morphology
Characteristic
Median
Range
AAA maximum diameter [mm]
56.7
44.0–114.3
AAA length [mm]
92.3
37.5–147.0
ILT maximum diameter [mm]
21.6
4.8–45.3
ILT volume [mm3 ]
75
2–347
Characteristic
N
%
Aneurysm type
AAA only
36
87.8
AAA and TAA
5
12.2
AAA form
Saccular
9
22.0
Fusiform
26
63.4
Unknown
6
14.6
Circumference [%]
< 25
2
4.9
50
3
7.3
75
13
31.7
100
15
36.6
Unknown
8
19.5
ILT presence
Yes
33
80.5
Unknown
8
19.5
ILT coverage
Circular
15
36.6
Partial
17
41.5
Unknown
9
22.0
Abbreviations: AAA, abdominal aortic aneurysm; ILT, intraluminal thrombus; TAA, thoracic
aortic aneurysm.
In line with published studies,[3 ] the plasma D-dimer concentration ([Table 2 ]) was highly increased in AAA patients compared with controls (median: 1.30 μg/mL
[IQR = 1.68] vs. 0.47 μg/mL [IQR = 0.49], p < 0.001) and correlated with AAA maximum diameter (r = 0.537, p = 0.001) and ILT maximum diameter (r = 0.500, p = 0.009).
AAA Patients Show a Higher Frequency of Activated and Newly Released Neutrophils as
well as CD16+ Monocytes
The study focused on evaluating possible alterations in the distribution of circulating
neutrophil and monocyte sub-sets in AAA patients compared with controls. Activated
neutrophils, marked by the loss of CD62L fluorescent signal intensity ([Supplementary Figs. S1 ] and [S2 ], available in the online version), were significantly more prevalent in the AAA
patient group ([Table 5 ]). The median number of activated neutrophils in AAA patients ([Fig. 1C ]) was 31 × 106 cells/L (IQR = 24), while it was 21 × 106 cells/L (IQR = 16) in the healthy group, p = 0.005. Regarding newly released neutrophils ([Fig. 1B ]), AAA patients had 38 × 106 cells/L (IQR = 31) versus 31 × 106 cells/L (IQR = 23) in the healthy control group, p = 0.020. A significant correlation was found between the absolute concentration of
circulating activated and newly released neutrophils (Spearman's r = 0.584, p < 0.001, [Fig. 1F ]). The concentration of total neutrophils ([Fig. 1A ]), the mature, quiescent sub-set ([Fig. 1D ]) and neutrophil–platelet aggregates ([Fig. 1E ]) did not differ significantly between the two groups.
Fig. 1 Distribution of neutrophil populations in peripheral blood of abdominal aortic aneurysm
(AAA) patients and healthy controls. Boxplots illustrate the absolute concentration
of (A ) total neutrophils, (B ) the newly released, (C ) activated or (D ) mature, quiescent neutrophil sub-set as well as (E ) neutrophil–platelet aggregates. Group comparisons were performed by Mann–Whitney
U test. (F ) The linear correlation between activated and newly released neutrophils is evaluated
by scattergram and Spearman's coefficient of correlation.
Table 5
Neutrophil and monocyte sub-set frequencies in peripheral blood of AAA patients and
healthy control individuals
Parameter
Healthy
AAA
p -Value
Neutrophil populations
Median (IQR)
Median (IQR)
Total neutrophils [×106 /L]
3,375 (1,535)
3,742 (1,351)
0.217
Quiescent neutrophils [×106 /L]
3,063 (1,400)
3,627 (1,344)
0.188
Newly released neutrophils [×106 /L]
31 (23)
38 (31)
0.020
Activated neutrophils [×106 /L]
21 (16)
31 (24)
0.005
Neutrophil-platelet aggregates [×106 /L]
344 (410)
401 (562)
0.617
Quiescent neutrophils [%]
98.37 (0.99)
98.13 (1.20)
0.030
Newly released neutrophils [%]
0.92 (0.68)
1.09 (0.64)
0.079
Activated neutrophils [%]
0.62 (0.46)
0.87 (0.59)
0.030
Neutrophil-platelet aggregates [%]
11.10 (7.30)
12.40 (14.62)
0.737
Monocyte populations
Median (IQR)
Median (IQR)
Total monocytes [×106 /L]
518 (292)
497 (356)
0.928
Classical monocytes [×106 /L]
441 (277)
366 (256)
0.506
Intermediate monocytes [×106 /L]
31 (41)
50 (48)
0.259
Non-classical monocytes [×106 /L]
38 (44)
49 (51)
0.315
CD16+ monocytes [×106 /L]
73 (70)
109 (91)
0.219
Monocyte-platelet aggregates [×106 /L]
74 (73)
82 (79)
0.970
CD16-/CD16+ monocyte ratio
5.24 (4.80)
4.33 (2.73)
0.039
Classical monocytes [%]
83.97 (10.65)
81.22 (10.02)
0.039
Intermediate monocytes [%]
7.12 (4.72)
7.86 (3.79)
0.116
Non-classical monocytes [%]
7.21 (6.76)
9.99 (8.50)
0.060
CD16+ monocytes [%]
16.03 (10.65)
18.78 (10.02)
0.039
Monocyte–platelet aggregates [%]
14.34 (6.94)
16.09 (16.10)
0.888
Abbreviations: AAA, abdominal aortic aneurysm; IQR, interquartile range.
Our analysis further revealed a higher median frequency of CD16+ circulating monocytes
of 18.8% (IQR = 10.0) in the AAA patient group compared with 16.0% (IQR = 10.7) in
the control collective, p = 0.039 ([Table 5 ], [Fig. 2F ]). The ratio of CD16-/CD16+ monocytes was calculated as a measure of monocyte activation
and was significantly changed in AAA patients with a median value of 4.33 (IQR = 2.73)
versus 5.24 (IQR = 4.80) in control individuals (p = 0.039). The absolute number of monocytes as well as the three respective sub-sets
did not differ significantly between groups ([Fig. 2A ]–[D ]). Furthermore, the blood level of MPAs was not significantly different between patients
and controls ([Fig. 2E ]).
Fig. 2 Distribution of monocyte populations in peripheral blood of abdominal aortic aneurysm
(AAA) patients and healthy controls. Boxplots illustrate the absolute concentration
of (A ) total monocytes, (B ) the classical, (C ) intermediate or (D ) non-classical monocyte sub-set as well as (E ) monocyte–platelet aggregates. Intermediate and non-classical monocytes were further
combined to the group of CD16+ monocytes, and their relative frequency is shown in
(F ). Group comparisons were performed by Mann–Whitney U test.
Plasma Factors Related to Myeloid Cell Activation and Recruitment are Elevated in
AAA Patients and Plasma Myeloperoxidase Correlates with Tissue-Released Protein from
Aneurysm Wall and ILT
We screened plasma samples from 10 AAA patients and 10 control subjects using a multi-cytokine
panel (myeloperoxidase [MPO], eotaxin, macrophage inflammatory protein [MIP]-1α, growth-regulated
oncogene-α, interferon gamma-inducible protein-10, interleukin-8, C-C chemokine ligand
2 [CCL2], MIP-1β, RANTES and stromal cell-derived factor-1α). Only plasma levels of
MPO and CCL2 (also termed monocyte chemoattractant protein 1) were significantly increased
in the AAA patients in the pilot analysis, so MPO and CCL2 were prioritized for assessment
in the entire cohort. This difference was confirmed using enzyme-linked immunosorbent
assay (ELISA) for the total sample set. The median CCL2 chemokine concentration ([Fig. 3A ]) in blood of AAA patients was 183 pg/mL (IQR = 45) as compared with 170 pg/mL (IQR = 62)
in the healthy controls, p = 0.048. Of note, the soluble MPO level ([Fig. 3B ]) was almost twice as high among AAA patients compared with control subjects (median:
13.3 ng/mL [IQR = 13.1] vs. 7.7 ng/mL [IQR = 3.6], p < 0.001).
Fig. 3 Increased plasma and tissue levels of soluble factors associated with myeloid cell
activation. The boxplot comparison illustrates the distribution of (A ) plasma C-C chemokine ligand 2 (CCL2), (B ) plasma myeloperoxidase (MPO) and (C ) tissue MPO concentrations as measured by enzyme-linked immunosorbent assay (ELISA)
in samples of abdominal aortic aneurysm (AAA) patients and healthy controls. For tissue
analysis, sections from healthy and AAA aortic walls were incubated in medium overnight
to release tissue-contained MPO. Group comparisons were performed by Mann–Whitney
U test. Correlations between (D ) plasma and tissue-released MPO, as well as maximum AAA diameter and (E ) tissue MPO or (F ) plasma MPO levels are shown in scattergrams (Spearman's coefficient of correlation).
Based on the robust association of plasma MPO level with AAA status, we investigated
conditioned media from the AAA wall (N = 13) to see if the potential source of MPO was indeed the AAA site. Compared with
healthy aortic tissue (N = 13), more MPO was released from AAA wall samples (median: 639 ng/mL [IQR = 401]
vs. 56 ng/mL [IQR = 277], p < 0.001). Of note, there was a linear correlation between plasma and tissue-released
MPO of patients whose aortic wall had been analysed (Spearman's coefficient r = 0.620, p < 0.001; [Fig. 3C ] and [D ]). Furthermore, tissue-released and plasma MPO levels correlated significantly with
AAA maximum diameter: r = 0.566, p = 0.044 and r = 0.391, p = 0.013, respectively ([Fig. 3E ] and [F ]).
Plasma Myeloperoxidase and D-dimer Levels are Independently Associated with AAA Status
and may be Combined to Yield an Advanced Diagnostic Score
Next, a ROC analysis was performed using explorative variables which have shown significant
differences between the subject collectives. Of these, the highest diagnostic marker
potential was seen with plasma MPO levels showing an area under the curve (AUC) of
0.816, p < 0.001, as compared with activated and newly released neutrophils (AUC = 0.684,
p = 0.005 and AUC = 0.654, p = 0.020, respectively), %CD16+ monocytes (AUC = 0.646, p = 0.039) as well as CCL2 (AUC = 0.631, p = 0.048).
A diagnostic logistic regression model was developed including the following parameters:
blood counts of total neutrophils, quiescent neutrophils, activated neutrophils, newly
released neutrophils, total monocytes, the percentage of CD16+ monocytes as well as
plasma levels of D-dimer, MPO and CCL2. After backward elimination, only D-dimer (odds
ratio: 3.96 and 95% confidence interval [95% CI]: 1.65–12.81) and MPO (odds ratio:
1.23, 95% CI: 1.08–1.43) had significant independent importance for AAA diagnosis
and were retained in the model. From the logistic model, a linear score was derived,
according to the estimated regression coefficients, as ‘score’ = –3.442 + 1.375 *
D-dimer + 0.205 * MPO. When compared with ROC analysis for MPO (AUC = 0.816) and D-dimer
(AUC = 0.830) alone, the combined score performed better (AUC = 0.877) in discriminating
patients from healthy individuals ([Fig. 4A ]). Evaluation of the dichotomized score (grouped into positive and negative values)
revealed a diagnostic sensitivity of 73% and specificity of 80%. The score also showed
a higher coefficient of correlation (r = 0.664, p < 0.001) than MPO (r = 0.391, p = 0.013) and D-dimer (r = 0.537, p = 0.001) for maximal AAA diameter ([Table 6 ]), while D-dimer (r = 0.687, p < 0.001) was more closely correlated with ILT volume than the calculated diagnostic
score (r = 0.450, p = 0.031).
Table 6
Correlation of explorative parameters with dimensions of the aneurysm and intraluminal
thrombus
Clinical parameter
R
p -Value
AAA maximum diameter [mm]
D-dimer [µg/mL]
0.537
0.001
MPO [ng/mL]
0.391
0.013
Score
0.664
< 0.001
ILT maximum diameter [mm]
Newly released neutrophils [×106 /L]
0.528
0.002
D-dimer [µg/mL]
0.500
0.009
ILT volume [mm3 ]
D-dimer [µg/mL]
0.687
< 0.001
Score
0.450
0.031
Abbreviations: AAA, abdominal aortic aneurysm; ILT, intraluminal thrombus; MPO, myeloperoxidase.
Fig. 4 Diagnostic marker potential of the D-dimer and myeloperoxidase (MPO) based score.
Receiver operating characteristics (ROC) curves are given for (A ) the initial cohort of abdominal aortic aneurysm (AAA) patients and healthy controls,
and (B ) the validation cohort. (C ) The score distribution is further illustrated by boxplot for the validation set,
with a significant difference between AAA patients and controls matched for age, sex
and previous cardiovascular disease (Mann–Whitney U test). (D ) Score values of the validation cohort are shown in relation to AAA maximum diameter
(Spearman's coefficient of correlation).
Myeloid Cell Parameters are Sensitive to Confounders but the Diagnostic Score Identifies
AAA Irrespective of Co-Morbidities and Patient Medication
It was further of interest to evaluate whether the significant difference in explorative
parameters between AAA patients and healthy controls was lost when the collectives
were matched for the presence or absence (retaining the larger sample size) of a particular
co-morbidity or medication ([Supplementary Tables S1–S5 ], available in the online version). The blood populations of activated and newly
released neutrophils were influenced by hyperlipidaemia and/or statin therapy. Similarly,
when the collectives were matched for aspirin intake, the difference in neutrophil
sub-set frequency was lost. Hypertension and anti-hypertensive drugs seemed to have
a minor impact on the distribution of neutrophil populations.
Despite the fact that the CD16+ monocyte population had only a weak diagnostic marker
value for the original collective (p = 0.039), reductions in sample size due to confounder matching mostly resulted in
p -values of ≤ 0.1 after adjusting for co-morbidities or medication. Yet, matching for
aspirin intake, hyperlipidaemia and/or statin therapy as well as hypertension and
in particular angiotensin receptor blocker treatment rendered the group difference
in CD16+ monocyte frequency insignificant. Comparably, the diagnostic power of the
monocyte chemotactic factor CCL2 was lost when the collectives were matched for CHD,
hyperlipidaemia, statin or anti-hypertensive therapy.
In contrast, the significant difference in plasma MPO, D-dimer and calculated score
was not lost when the AAA and patient collectives were matched for any co-morbidity
or medication, even though the MPO marker value was lowered when aspirin or statin
therapy were taken into consideration. Of note, MPO, D-dimer and score also correlated
with kidney function parameters ([Supplementary Table S6 ], available in the online version), but none of the investigated parameters showed
a significant association with the Framingham Risk Score. Finally, when the novel
AAA score was evaluated in multivariable analysis by binary logistic regression ([Table 7 ]), the score as well as hyperlipidaemia and/or lipid-lowering therapy were revealed
as significant and independent diagnostic variables, while hypertension and/or anti-hypertensive
therapy, CHD and anti-platelet therapy were not.
Table 7
Multivariable analysis (binary logistic regression) based on 41 AAA cases and 38 controls
of the original dataset
Parameter
Exp(B)
95% CI lower value
95% CI upper value
p -Value
Score
1.873
1.085
3.234
0.024
Hypertension and/or anti-hypertensive therapy
1.186
0.118
11.879
0.885
Hyperlipidaemia and/or lipid-lowering therapy
13.178
2.291
75.814
0.004
Coronary heart disease
1.162
0.148
9.150
0.887
Anti-platelet therapy
3.410
0.580
20.035
0.175
Constant
0.085
0.027
Abbreviation: CI, confidence interval.
Note: Exp(B), odds ratio.
To validate the score, additional plasma samples of 63 AAA cases and 63 controls were
analysed which were independently collected in Leeds (United Kingdom) and one-to-one
matched for age, sex and diagnosis of previous cardiovascular disease (composite of:
myocardial infarction, angina pectoris, peripheral vascular disease or cerebrovascular
disease). In this validation cohort, the AAA and control group differed significantly
(p < 0.001) in score values ([Fig. 4C ]) with an AUC of 0.693 in ROC analysis ([Fig. 4B ]). In addition, the score correlated positively with maximum aortic diameter ([Fig. 4D ]) in the AAA collective (r = 0.433, p < 0.001) but not in the control group.
The Score Further Predicts Disease Progression
In a prognostic study design, 33 patients with small aneurysms (without indication
for surgical repair; [Supplementary Table S7 ], available in the online version) were followed at 6-month intervals for AAA expansion.
In 68 datasets, the score was calculated at the beginning of the monitoring period
and found to correlate significantly with aneurysm growth over the next 6 months (r = 0.437, p < 0.001; [Fig. 5A ]). When the events were grouped into slow (< 2 mm) and fast (≥ 2 mm) progression
over 6 months, the score ([Fig. 5B ] and [C ]) outperformed D-dimer ([Fig. 5E ] and [F ]) in identifying patients with rapid aneurysm progression with an AUC value of 0.701
(p = 0.006) as compared with 0.612 (p = 0.124). A score cut-off at 0.935 was calculated by Youden's index to yield a prognostic
sensitivity and specificity of 72 and 67%.
Fig. 5 Prognostic marker potential of the D-dimer and myeloperoxidase (MPO)-based score.
The score (A–C ) was compared with D-dimer plasma values (D–F ) in predicting abdominal aortic aneurysm (AAA) expansion over the next 6 months.
(A , D ) A direct correlation was assessed by Spearman's test. (B , E ) Events were grouped into slow (< 2 mm) and fast (≥ 2 mm) aneurysm growth in 6 months
and evaluated by Mann–Whitney U test as well as (C , F ) receiver operating characteristics (ROC) analysis. Of note, two samples with exceedingly
high values for D-dimer (15.3 and 24.3 µg/mL) and the calculated score (21.0 and 37.9)
were omitted from the graphs A , B , D and E to improve resolution.
Discussion
In the present study, we focused on the analysis of circulating monocyte and neutrophil
sub-sets and associated plasma factors CCL2 and MPO, evaluating their potential as
biomarkers in the context of AAA. We prioritized matching the control group according
to sex, age, body mass index and smoking status, with age and smoking being the most
important risk factors for AAA.[33 ] A strong association of plasma D-dimer concentration and AAA has been established
for some time and several studies have shown its potential as a diagnostic and prognostic
marker.[6 ]
[7 ]
[8 ] In agreement with these reports, the D-dimer levels were 2.8-fold increased in our
patient group and showed a significant correlation with AAA maximum diameter and markedly
with ILT volume.
The chronic inflammation and in particular the infiltration by myeloid cell populations
in ILT and AAA wall are considered central elements of AAA pathogenesis.[34 ]
[35 ] We hypothesized that changes in the activation state of monocytes and neutrophils
might also be observed in the systemic circulation of AAA patients. Both chronic and
acute inflammatory states are often accompanied by a selective increase in the CD16+
monocyte sub-sets.[17 ]
[36 ] A rise in the frequency of CD16 expressing monocytes was also found in our AAA patients
(18.8%) versus controls (16.0%) and is in agreement with previously published results
by Ghigliotti et al[23 ] who reported 15.1 and 11.0%, respectively. It should be noted that the difference
in recorded values may relate to distinct flow cytometric approaches to eliminate
‘contaminating’ lymphocytes and granulocytes in monocyte detection. However, we found
the diagnostic power of CD16+ monocytes for AAA (AUC = 0.646, p = 0.039) to be rather limited when compared with other myeloid cell activation markers.
Considering the neutrophil sub-sets, we based our methodological approach on the previously
published marker selection by Pillay et al.[25 ] To exclude interference by transient inflammatory states, blood collection was postponed
if patients had a known infection. A significantly increased number of activated and
newly released neutrophils was recorded in the AAA patient versus control group. This
likely reflects the AAA site of chronic inflammation and immune cell recruitment.
In the further investigation, we followed the notion that systemically observed cellular
activation and recruitment of myeloid cells might be accompanied by a measurable change
in plasma cytokine or chemokine levels. Thus, an initial screening of 10 patient versus
10 control samples was performed using a multiplex bead array for several factors
known to influence or reflect myeloid cell activation, differentiation and migration.
While most of the investigated cytokines and chemokines had previously been detected
in AAA tissue,[37 ] limited information was available on their blood values in AAA patients.[38 ] In our reduced collective, only MPO and CCL2 differed significantly between the
two study groups which was subsequently confirmed by ELISA. Elevations of other factors
might be detected with a larger sample size. However, the increase of plasma MPO and
CCL2 in AAA patients (compared with healthy controls) proved to be most pronounced
and hence exhibit the best biomarker potential. CCL2 is a chemoattractant for monocytes,[39 ] a differentiation stimulus enhancing CD16 expression on monocytes[40 ] and drives a vicious circle of inflammation and SMC apoptosis.[41 ]
[42 ] MPO has received little research attention in the AAA context with differential
results regarding its marker potential.[8 ]
[43 ] It is of central importance for the innate defence catalyzing the formation of highly
bacteriotoxic hypochlorite radical species, but is also known to contribute to media
destruction in AAA.[34 ] In our study, we found moderately elevated plasma levels of CCL2 but highly increased
(in median doubled) MPO blood concentrations in the patient group. Moreover, plasma
MPO correlated with tissue-released MPO and with maximum AAA diameter, identifying
the aneurysm as the primary source for circulating MPO. Of note, several patients
with an aortic diameter beyond 80 mm had moderate MPO blood concentrations. This might
indicate that in very large, non-ruptured aneurysms the balance tilts towards fibrotic
rather than inflammatory processes. The superior marker potential of plasma MPO was
further confirmed by ROC analysis and may relate to the circumstance that MPO is released
by both, activated neutrophils as well as CD16+ monocytes.[44 ]
[45 ]
In multivariable logistic regression models including the investigated myeloid cell
parameters, MPO prevailed as an independent marker for AAA diagnosis along with the
established AAA marker, D-dimer. The two parameters may be combined in a diagnostic
score reaching a sensitivity of 73% and specificity of 80% with the cut-off set to
0, and independent of age, sex and smoker status. To address the issue of potential
confounders not considered when matching patient and control collectives, the devised
diagnostic score and all explorative parameters, which had been found to differ significantly
between the groups, were re-evaluated after adjustment for co-morbidities or stable
medication. The two collectives differed most prominently in hyperlipidaemia, statin
use and aspirin therapy, but also in the frequency of CHD, hypertension and in kidney
function. Overall, the neutrophil and monocyte sub-sets as well as CCL2, which had
moderate diagnostic marker value, were more sensitive to confounders. In contrast,
plasma MPO, D-dimer and the deduced score proved significantly different between the
collectives even after adjustment for confounding conditions, and the score prevailed
in multivariable analysis as an independent diagnostic variable. A further indication
of the robustness of the diagnostic score was provided by a validation set of 126
independently collected plasma samples from AAA patients and controls matched for
sex, age and previous cardiovascular disease.
Importantly, regarding the monitoring of patients with a known AAA diagnosis, the
score also showed prognostic marker value which exceeded the potency of D-dimer alone.
Based on a cut-off set to 0.935, the score could identify patients with rapid progression
(≥ 2 mm over the next 6 months) with 72% sensitivity and 67% specificity. Thus, the
score holds an important potential for patient stratification.
In conclusion, our data provide proof-of-concept that myeloid inflammation markers
constitute a circulating footprint of atherosclerotic aortic disease. We have provided
first evidence that a combined score of plasma MPO and D-dimer may prove beneficial
in AAA diagnosis and prognosis. While D-dimer is a confirmed and robust, albeit non-specific
AAA indicator, it is primarily associated with the presence and size of the ILT. By
including MPO as a second parameter related to the central pathophysiological component
of inflammation, the deduced score is more closely associated with AAA diameter than
D-dimer and holds the potential to also detect and monitor aneurysms with no or little
thrombus. The diagnostic application of the score may be limited to countries without
ultrasound screening program and requires further large-scale validation with unmatched
cohorts. The prognostic value of the score to predict disease progression seems to
be of particular interest as it may serve clinicians to better guide the timing of
therapeutic interventions and would thus merit evaluation in prospective multi-centre
studies.
Limitations of the Study
The sample size of our study has certainly limited the detection of potential biomarkers
to parameters which show a substantial difference in mean values between groups and/or
small biological variance. While this may prove beneficial in clinical practice, it
does not reflect the parameter importance in AAA pathogenesis and may also explain
discordances between our data and published literature.
As the study was explorative in design, no correction for multiple testing was applied
which increases the risk of committing a type I error. However, the deduced diagnostic
score was subsequently challenged with a validation cohort and proved significant
irrespective of cardiovascular disease.
D-dimer and MPO assessment can be included in routine blood analysis at a cost of
€83 for both tests (€24 for D-dimer, €59 for MPO) currently charged by the local routine
laboratory. D-dimer measurements are generally robust, based on citrated plasma and
widely offered by routine laboratories. In contrast, MPO assays are less frequent
in clinical use, but standardized tests are available for validated general chemistry
analysers (e.g. Diazyme Latex Enhanced Immunoturbidimetric MPO Assay). While these
assays are offered for standard ethylenediaminetetraacetic acid or heparin plasma,
a more careful pre-analytical sample handling (such as blood storage on ice, centrifugation
within the hour, as conducted in our study) may be required to prevent artificial
myeloid cell activation, enable accurate plasma MPO assessment and score calculation
and will thus have to be further evaluated for clinical routine application.
With respect to leukocyte sub-set measurement, a single platform approach with absolute
counting beads would have provided greater precision with a potentially better diagnostic
value for neutrophil and monocyte sub-sets than detected in the current study. Furthermore,
the apoptotic rate of myeloid cells was not addressed in our study (since dead cells
were excluded from the analysis by 7-AAD staining). However, the increased numbers
of activated neutrophils which are reportedly exhibiting delayed apoptosis[46 ] and of newly released neutrophils recruited upon neutrophil loss may indirectly
reflect changes in the balance of the neutrophil apoptotic rate.
What is known about this topic?
The abdominal aortic aneurysm (AAA) is mostly asymptomatic and hence difficult to
diagnose unless routine ultrasound screening is implemented in the national health
plan. Furthermore, the prediction of rapid AAA progression is vital, as it is associated
with a high risk of rupture and patient death.
The best characterized diagnostic and prognostic blood parameter to date is the fibrin
degradation product D-dimer which is related to the frequently occurring intraluminal
thrombus but is also increased in other cardiovascular disorders.
Pre-clinical models have shown a central role for neutrophils and monocytes in AAA
pathogenesis which points to a potential of myeloid factors as AAA biomarkers.
What does this paper add?
Based on a comprehensive comparison of neutrophil and monocyte activation parameters,
myeloperoxidase was identified as the most elevated myeloid marker in AAA patient
blood.
D-dimer and myeloperoxidase represent two sensitive biomarkers of AAA which reflect
distinct components of the AAA pathomechanism and when combined yield an improved
score for AAA diagnosis and, more importantly, for prediction of rapid progression.
