Angiogenic Factors Correlate with T Cell Immune Reconstitution and Clinical Outcomes after Double-Unit Umbilical Cord Blood Transplantation in Adults

Abstract

Umbilical cord blood (UCB) is a valuable graft source for allogeneic hematopoietic stem cell transplantation (HSCT) in patients who lack adult donors. UCB transplantation (UCBT) in adults results in delayed immune reconstitution, leading to high infection-related morbidity and mortality. Angiogenic factors and markers of endothelial dysfunction have biologic and prognostic significance in conventional HSCT, but their role in UCBT has not been investigated. Furthermore, the interplay between angiogenesis and immune reconstitution has not been studied. Here we examined whether angiogenic cytokines, angiopoietin-1 (ANG-1) and vascular endothelial growth factor (VEGF), or markers of endothelial injury, thrombomodulin (TM) and angiopoietin-2 (ANG-2), associate with thymic regeneration as determined by T cell receptor excision circle (TREC) values and recovery of T cell subsets, as well as clinical outcomes in adult recipients of UCBT. We found that plasma levels of ANG-1 significantly correlated with the reconstitution of naive CD4⁺CD45RA⁺ and CD8⁺CD45RA⁺ T cell subsets, whereas plasma levels of VEGF displayed a positive correlation with CD4⁺CD45RO⁺ T cells and regulatory T cells and a weak correlation with TRECs. Assessment of TM and ANG-2 revealed a strong inverse correlation of both factors with naive T cells and TRECs. The angiogenic capacity of each patient's plasma, as determined by an in vitro angiogenesis assay, positively correlated with VEGF levels and with reconstitution of CD4⁺ T cell subsets. Higher VEGF levels were associated with worse progression-free survival and higher risk of relapse, whereas higher levels of TM were associated with chronic graft-versus-host disease and nonrelapse mortality. Thus, angiogenic factors may serve as valuable markers associated with T cell reconstitution and clinical outcomes after UCBT.

Introduction

The use of umbilical cord blood (UCB) as a graft source has expanded the application of allogeneic hematopoietic stem cell transplantation (HSCT) in patients who lack a suitable HLA-matched adult donor. The use of 2 UCB units has circumvented the stem cell dose barrier and has shortened the time to engraftment in adult recipients of UCB transplantation (UCBT) [1,2]. However, despite the improved myeloid engraftment, lymphoid reconstitution after UCBT remains delayed, even with the use of 2 UCB grafts [3,4].

Neovascularization is important in the hematopoietic and immunologic reconstitution process after HSCT [5]. Vascular endothelial growth factor (VEGF) has a regulatory role in reconstitution of hematopoiesis after bone marrow injury or transplantation, likely because hematopoietic and endothelial cells share a common progenitor and express membrane receptors (Flt-1, Flk-1, Tie 2) for VEGF and other angiogenic cytokines. Specifically, mice deficient in VEGFR2 display profound defects in both vasculogenesis and hema-topoiesis, whereas inhibition of VEGFR1 signaling prevents HSC differentiation and hematopoietic recovery after bone marrow suppression [6-8]. Consistent with the mandatory role of VEGF-mediated angiogenesis in hematopoiesis, VEGF blockade at the time of transplantation in lethally irradiated mice results in failure of hematopoietic reconstitution and early death [9]. Elevated levels of VEGF and/or its receptors have been reported in patients with hematologic malignancies, where they play a role in an autocrine/paracrine fashion [10-12], and have been shown to be of prognostic significance [13-16]. Levels of VEGF before allogeneic HSCT are associated with increased risk of relapse [17], but reports regarding its role on clinical outcomes such as graft-versus-host disease (GVHD), nonrelapse mortality (NRM), or survival are conflicting [17-20].

In addition to VEGF and its receptors, the angiopoietin/Tie-2 system plays a critical role in the regulation of vasculogenesis and hematopoiesis. Angiopoietin-1 (ANG-1) is a cognate ligand that serves as an agonist by inducing phosphorylation of the Tie-2 receptor and mediates vasculoprotective properties by promoting endothelial cell survival and migration [21]. In contrast, angiopoietin-2 (ANG-2) is a natural Tie-2 antagonist that renders blood vessels unstable [22]. The ANG–1–Tie-2 axis has been reported to play an important role in the repopulating activity of HSCs and recovery of hematopoiesis after myelosuppression [23,24]. Elevated levels of ANG-2 are associated with worse outcomes after HSCT [25,26]. However, the significance of ANG-1 and ANG-2 in the setting of UCBT is not known.

Thrombomodulin (TM) is a transmembrane glycoprotein expressed on the surface of all vascular endothelial cells where it promotes thrombin-mediated activation of protein C [27-29]. A soluble form of TM is also present in the plasma, produced either by secretion or by enzymatic cleavage of tissue TM after endothelial cell injury [30]. TM levels have been used as a marker of endothelial damage after HSCT [31,32], but the relationship between TM and clinical outcomes after UCBT is not well characterized.

A significant component of immune reconstitution after UCBT involves thymic-dependent generation of T cells [33]. HSCT compromises thymopoiesis by injury of the thymic microenvironment, particularly thymic epithelial cells. Human UCB is enriched in endothelial precursors that can sustain thymopoiesis in immunodeficient mice transplanted with human thymic grafts, where they engraft and promote neovascularization and wound healing [34]. Notably, the angiopoietins ANG-1 and ANG-2 have been implicated in the proliferation of endothelial cells from UCB CD34⁺ progenitors [35]. Furthermore, in murine studies, VEGF has been shown to play a key role in thymic reconstitution after experimental transplantation [36,37]. Therefore, we hypothesized that in addition to their role in hematopoietic recovery and clinical outcomes, angiogenic factors might also be involved in the thymic recovery and T cell reconstitution after HSCT.

In the present study we investigated the association of plasma angiogenic factors with thymic and T cell reconstitution and with clinical outcomes in a cohort of adult patients undergoing double-unit UCBT (dUCBT). We determined that plasma levels of the 2 proangiogenic factors, VEGF and ANG-1, positively correlate with T cell subsets and T cell receptor excision circles (TRECs). In contrast, patients with high TM and ANG-2, markers of endothelial injury, have inferior thymic reconstitution at 1 year after dUCBT. Further, high VEGF levels are associated with worse progression-free survival (PFS) and increased incidence of relapse, whereas TM levels are associated with chronic GVHD (cGVHD), higher NRM, and worse overall survival (OS). These findings suggest that circulating angiogenic factors may be involved in post-transplant immune reconstitution and might serve as prognostic markers for clinical outcomes after UCBT.

Methods

Patients and UCB Units

Patients were enrolled in a phase II study of dUCBT at the Dana Farber/Harvard Cancer Center. Eligibility criteria, detailed design, and clinical outcomes of the trial were previously reported [38,39]. In brief, patients were eligible for enrollment if they were adults, had a hematologic malignancy, and lacked a suitable related (6/6 or 5/6 HLA-A, -B, and -DRB1 matched) or unrelated (10/10 HLA-A, -B, -C, -DRB1, and -DQ matched) donor. UCB units were obtained from national and international cord blood banks. Each individual UCB unit was required to have a minimum of 1.5 × 10⁷ total nucleated cells/kg before cryopreservation, and the 2 UCB units selected for each subject were required to provide a minimum of combined pre-cryopreservation cell dose of 3.7 × 10⁷ total nucleated cells/kg. UCB units were required to be a 4/6 match or better at the allele level for HLA-A, -B, and -DRB1 with each other and with the recipient. The UCB units were hierarchically selected on the basis of higher cell dose, greater HLA match, and younger age of the unit.

Treatment Protocol

On a research study, all subjects were conditioned with fludarabine (30 mg/m²/day) for 6 consecutive days (days –8 through –3; total dose 180 mg/m²), melphalan (100 mg/m²) on day –2, and rabbit antithymocyte globulin (Thymoglobulin, Sangstat, Fremont, CA, 1.5 mg/kg/day) on days –7, –5, –3, and –1. Two UCB units were infused sequentially between 1 and 6 hours apart on day 0. After transplantation, patients received transfusion support, and filgastrim (5 μg/kg/day) was administered from day +5 until an absolute neutrophil count was higher than 2.0 × 10⁹ cells/L for 2 consecutive days. GVHD prophylaxis began on day –3 with tacrolimus (.05 mg/kg for a target serum level of 5 to 10 ng/mL) and an oral loading dose of sirolimus (12 mg). Sirolimus was subsequently dosed orally once a day for a goal serum trough level of 3 to 12 ng/mL. Both GVHD prophylaxis agents were tapered from day 100 through day 180 for patients with no evidence of GVHD.

The research protocol was approved by the Institutional Review Board of the Dana Farber/Harvard Cancer Center. Written informed consent was obtained from all patients before enrollment to the study. The trial was prospectively registered at http://www.clinicaltrials.gov (NCT00133367).

Immunophenotyping

Patient blood samples were collected before administration of conditioning chemotherapy and at 4 weeks, 8 weeks, 100 days, 6 months, and 1 year after transplantation. Peripheral blood mononuclear cells were isolated using Ficoll-Paque Plus (GE Healthcare, Chicago, IL) and stained with fluorescence-conjugated monoclonal antibodies for lineage-specific marker analysis, using a BD FACSCanto flow cytometer (BD Biosciences, San Jose, CA).

TREC Analysis

TREC analysis was performed according to a previously described protocol [40]. DNA was isolated from peripheral blood mononuclear cells with a QIAmp DNA Mini Kit (Qiagen, Germantown, MD). Quantification of signal-joint TREC DNA was performed by quantitative-competitive PCR, using a Rotor-Gene 6000 thermal cycler (Corbett Life Science). The standard curve was prepared with 10-fold dilutions of a plasmid containing the signal-joint TREC sequence (kindly provided by Daniel Douek, National Institute of Allergy and Infectious Diseases).

Cytokine Measurements

Patient plasma was collected after centrifugation of citrated blood samples collected before transplantation and at 4 weeks, 8 weeks, 100 days, 6 months, and 1 year after transplantation. For each time point the cytokines VEGF, ANG-1, ANG-2, and TM were measured with commercial Colorimetric Sandwich ELISA kits (R&D Systems), according to the manufacturer's instructions.

In Vitro Angiogenesis Assay

Human umbilical vein endothelial cells (HUVECs) were cultured in Vascular Cell Basal Medium (ATCC, Manassas, VA) supplemented with Endothelial Cell Growth Kit-BBE (ATCC) that does not contain VEGF. Fifteen-well plates (angiogenesis μ-slide; Ibidi, Madison, WI) were coated with Matrigel (BD Biosciences) at 4°C and incubated at 37°C to allow polymerization. Upon confluence, third or fourth passage HUVECs were released with Accutase cell detachment solution (Millipore, Billerica, MA), resuspended in serum-free growth medium, and plated onto the Matrigel-coated wells at 10,000 cells per well. Patient plasma was added at 2% concentration. After an 18-hour incubation, 3 visual fields per well were captured using an inverted phase microscope (Zeiss, Dublin, CA). Images were analyzed with image processing software (Angiogenesis Analyzer for ImageJ; National Institutes of Health, Bethesda, MD) for quantitative assessment of 3 different measurements of angiogenesis, specifically total segment length, number of segments, and number of meshes per visual field. The readout of each parameter was the median of 3 visual fields per well.

Statistical Analysis

Patient and cord blood unit characteristics are presented descriptively. The correlation between continuous variables was assessed using the Spearman rank test. Exact Wilcoxon-rank-sum test was used for group comparison of continuous variables. OS was defined as the time from transplantation to death from any cause, whereas PFS was defined as the time from transplantation to disease progression or death from any cause. OS and PFS were estimated using the Kaplan-Meier method. GVHD, relapse, and NRM were analyzed in the competing risks framework considering relapse/death without developing GVHD, NRM, and relapse as competing events, respectively. Multivariable Cox regression analysis was performed for OS, PFS, and relapse adjusting for age and univariable Cox regression analysis was performed for NRM and cGVHD. For all models, each of VEGF, ANG-1, ANG-2, and TM was treated as a time-dependent variable. We repeated the same multivariable models adjusting for either Disease Risk Index [41,42] or degree of HLA mismatch (4/6 + 4/6 versus other), and the results were consistent (data not shown). Because of the limited number of events, each model considered Firth's correction, and multivariable analysis was not performed for NRM and cGVHD.

Probability values were considered significant at the 2-sided .05 level without considering multiple comparisons. All calculations were done using SAS version 9.3 (SAS Institute Inc., Cary, NC) and R version 3.2.2 (the CRAN project, www.cran.r-project.org).

Results

Baseline Patient Characteristics, Outcomes, and Immune Reconstitution

Twenty-seven subjects who underwent transplantation and survived beyond 100 days from transplantation are included in this analysis (Table 1). Patients received dUCBT between October 2005 and September 2007. The median follow-up time among survivors was 74 months (range, 59 to 101). One patient developed grades II to IV acute GVHD and 6 patients developed cGVHD (22%). During the study follow-up period, 16 patients (59%) relapsed. The 5-year rates of OS and PFS were 52% and 22%, respectively. NRM accounted for 5 of 13 deaths observed. Detailed patient characteristics and clinical outcomes have been reported elsewhere [39].

Table 1. Baseline Characteristics of Patients (N = 27) and Cord Blood Units.

Characteristic	No. of Cases	Percent
Median age, yr (range)	48 (19-67)
Patient sex
Male	14	51.9
Female	13	48.1
Diagnosis
AML	7	25.9
CLL, SLL, PLL	1	3.7
CML	1	3.7
HL	5	18.5
ALL	1	3.7
MDS	2	7.4
MPD	1	3.7
NHL	8	29.6
Other acute leukemia	1	3.7
Disease Risk Index
Low	2	7.4
Intermediate	19	70.4
High	6	22.2
HLA match at A, B, DRB1
4/6 + 4/6	17	63.0
5/6 + 4/6	6	22.2
5/6 + 5/6	4	14.8

Median follow up among survivors 74 months (range 59-101).

AML indicates acute myelogenous leukemia; CLL, chronic lymphocytic leukemia; SLL, small lymphocytic lymphoma; PLL, prolymphocytic leukemia; CML, chronic myelogenous leukemia; ALL, acute lymphocytic leukemia; MDS, myelodysplastic syndrome; MPD, myeloproliferative neoplasm; HL, Hodgkin's lymphoma; NHL, non-Hodgkin lymphoma.

Detailed assessment of quantitative T cell and thymic reconstitution in this patient cohort has been previously reported [38]. Briefly, median values of CD4⁺ and CD8⁺ T cells and subsets remained substantially depressed through 6 months after dUCBT, when a slight increase in CD4⁺ T cells was first noted. This was largely accounted for by an increase in the CD4⁺CD45RO⁺ T cell subset, whereas the paucity of CD4⁺CD45RA⁺ naive T cells persisted. By 1 year after UCBT, subsets of CD8⁺ T cells reached normal median values, whereas CD4⁺ subsets remained below normal range. Thymic reconstitution was assessed by quantification of TRECs [33,40]. Before transplantation, median TREC values were slightly below the lowest level of normal range in this heavily pretreated patient population. After transplantation, median TREC values fell below pretransplant levels and remained undetectable through 100 days. In contrast, a marked increase in TRECs was observed at 6 months and levels reached low normal levels by 1 year.

Plasma Levels of VEGF and ANG-1 Positively Correlate with T Cell Subsets after dUCBT

First, we examined the kinetics of plasma levels of VEGF and ANG-1 in the first year after dUCBT and their association with T cell subsets. The 2 cytokines followed parallel kinetics with a marked rapid drop immediately after transplantation and a nadir at 4 weeks, and their levels remained low through the first 2 months after dUCBT. Substantial recovery was first observed at day 100, and values remained relatively stable onward until 1 year after dUCBT, although ANG-1 remained persistently lower than pretransplant levels (Figure 1A,B). VEGF and ANG-1 levels were positively correlated with each other at all time points during follow-up (data not shown).

Figure 1.

Kinetics of plasma VEGF, ANG-1, TM, and ANG-2 levels after dUCBT. (A-D) Median levels of VEGF (A), ANG-1 (B), TM (C) and ANG-2 (D) from baseline through 12 months post-transplantation are shown. Error bars denote 25th and 75th percentiles.

At 6 months after dUCBT, VEGF levels positively correlated with CD4⁺ cells (r = .53, P= .02), CD4⁺CD45RO⁺ (r= .53, P= .02), CD4⁺CD25⁺ regulatory T cells (Tregs; r = .74, P= .0005), and naive CD4⁺CD45RA⁺T cells (r = .39,P=.096), although the correlation with naive CD4⁺CD45RA⁺ T cells was weak (Table 2). In contrast, there was no correlation between plasma VEGF levels and total CD8⁺ T cells or CD8⁺ T cell subsets. When VEGF levels were analyzed in conjunction with TRECs, we observed a positive correlation between VEGF and TREC values at 6 months (r = .50, P = .059). Similarly to VEGF, at 6 months after dUCBT, ANG-1 also displayed largely positive correlation with CD4⁺ T cells (r = .46, P = .049), naive CD4⁺CD45RA⁺ T cells (r = .56, P = .01), CD4⁺CD25⁺ Tregs (r = .48, P = .04), and CD4⁺CD45RO⁺ T cells (r= .40,P= .08) (Table 2). However, unlike in VEGF, ANG-1 also positively correlated with CD8⁺ (r = .55, P = .01) and CD8⁺CD45RA⁺ (r = .69, P = .001) T cells.

Table 2. Plasma Levels of VEGF and ANG-1 Positively Correlate with T Cell Subsets at 6 Months after dUCBT.

	n	Correlation coefficient (r)	P
VEGF
CD4+	19	.53	.02
CD4+CD45RA+	19	.39	.096
CD4+CD45RO+	19	.53	.02
CD4+CD25+	18	.74	.0005
CD8+	19	.08	.75
CD8+CD45RA+	19	.14	.56
CD8+CD45RO+	19	.13	.59
TRECs	15	.50	.059
ANG-1
CD4+	19	.46	.049
CD4+CD45RA+	19	.56	.01
CD4+CD45RO+	19	.40	.08
CD4+CD25+	18	.48	.04
CD8+	19	.55	.01
CD8+CD45RA+	19	.69	.001
CD8+CD45RO+	19	.19	.43
TRECs	15	.32	.24

We also investigated whether correlation of angiogenic factors and immune reconstitution might differ between age groups by dichotomizing our cohort in patients ≥ 50 and < 50 years. In general, VEGF correlations with T cell subsets were stronger in the younger age group, whereas ANG-1 correlations with T cell subsets were stronger in the older age group (see Supplementary Table 1). These findings should be interpreted with caution, given the smaller sample size after dichotomization.

Plasma Levels of TM and ANG-2 Inversely Correlate with T Cell Subsets after dUCBT

Next, we examined the kinetics of TM and ANG-2 levels, which have been proposed as markers of endothelial injury. In contrast to the marked drop of the 2 proangiogenic cytokines, VEGF and ANG-1, in the early post-transplant period, plasma levels of TM and ANG-2 were relatively stable or slightly increased after conditioning, although ANG-2 levels showed greater variability compared with TM (Figure 1C,D). TM levels displayed an overall inverse correlation with CD4⁺ and CD8⁺ T cells and subsets. Specifically, TM levels showed an inverse correlation with CD4⁺ (r =−.64, P = .02) and CD8⁺ T cells (r =−.49, P = .09) at 1 year after dUCBT, although the latter correlation was not significant at the .05 level (Table 3). When CD4⁺ and CD8⁺ subsets were examined, there was a strong inverse correlation between TM and the naive CD4⁺CD45RA⁺ (r =−.64, P = .02) and CD8⁺CD45RA⁺ (r =−.65, P = .02) T cell subsets. In contrast, TM did not correlate with memory CD4⁺CD45RO⁺ (r =−.48, P = .09) and CD8⁺CD45RO⁺ (r =−.32, P = .29) T cell subsets. We did not detect any association between TM and Tregs at any time point after dUCBT. Importantly, we observed a striking inverse correlation between plasma TM and TREC values (r =−.76, P = .003) at 1 year after dUCBT (Table 3). ANG-2 levels also displayed an overall inverse correlation with CD4⁺ and CD8⁺ T cells, although the correlation was not as strong as TM. Specifically, ANG-2 levels showed a statistically significant inverse correlation with CD4⁺CD45RA⁺ T cells at 6 months (r =−.46, P = .046) and TRECs (r =−.62, P = .03) at 1 year after transplantation (Table 3). No clear age effect was observed on the correlations between ANG-1 and TM with T cell subsets by patient age group, but the number of assessable patients was small (Supplementary Table 1).

Table 3. Plasma levels of TM and of ANG-2 Inversely Correlate with T Cell Subsets at 1 Year after dUCBT.

	n	Correlation coefficient (r)	P
TM
CD4+	13	–.64	.02
CD4+CD45RA+	13	–.64	.02
CD4+CD45RO+	13	–.48	.09
CD4+CD25+	12	–.19	.58
CD8+	13	–.49	.09
CD8+CD45RA+	13	–.65	.02
CD8+CD45RO+	13	–.32	.29
TRECs	13	–.76	.003
ANG-2
CD4+	12	–.39	.2
CD4+CD45RA+	12	–.43	.1 6
	19*	–.46*	.046*
CD4+CD45RO+	12	–.38	.22
CD4+CD25+	11	–.52	.07
CD8+	12	–.42	.18
CD8+CD45RA+	12	–.45	.14
CD8+CD45RO+	12	–.38	.23
TRECs	12	–.62	.03

^*6 months.

Angiogenic Activity of Patients' Plasma after dUCBT

To determine the biologic relevance of our findings regarding the relationship between proangiogenic cytokines and quantitative T cell recovery, we used a modified in vitro angiogenesis assay to determine the angiogenic activity of patients' plasma, as previously described [43]. In this assay, HUVEC cells were resuspended in serum-free culture medium and either FCS or patient plasma added to assess angiogenic activity. After incubation, HUVEC cells underwent differentiation into mesh-like tubular structures, and their quantification provided an objective assessment of angiogenesis. Images were captured and analyzed as described in Methods. We measured 3 established parameters of angiogenesis, specifically total segment length, number of segments, and number of meshes.

First, we optimized the assay by assessing kinetics of in vitro angiogenesis and optimal FCS concentrations (Figure 2, rows 1 to 3). We determined that peak tubule formation was reached with 2% FCS at 18 hours, and slow disintegration was observed at later time points. The same kinetics were observed when 2% human plasma was used instead of FCS (Figure 2, row 4). We then used this assay using plasma from each 1 of 21 individual patients assessable at 6 months and observed a clear difference in the plasma angiogenic capacity among different patients (Figure 3A). At 6 months after dUCBT, we detected a positive correlation between total segment length of in vitro formed vascular structures and CD4⁺ (r = .45, P = .04), CD4⁺CD45RO⁺ (r = .49, P = .03), and CD4⁺CD25⁺ (r = .49, P = .03) T cell subsets (Figure 3B). We also observed a significant positive correlation between total segment length and VEGF levels (r = .56, P = .008). In contrast, there was no correlation between in vitro angiogenic capacity of patients' plasma and CD8⁺ T cells (data not shown). These results provide evidence that VEGF in patients' plasma is biologically functional and are consistent with our findings that plasma VEGF levels significantly correlated with CD4⁺, CD4⁺CD45RO⁺, and Treg cells but not with CD8⁺ T cells.

Figure 2.

In vitro angiogenesis assay. (A) Representative images of HUVEC cells cultured with no serum (top row) or with two different concentrations of FCS (second and third row). FCS in the culture medium replaced by human plasma (bottom row).

Figure 3.

In vitro angiogenesis assay and correlations with T cell subsets. (A) In vitro angiogenesis assay was performed using 2% plasma from each 1 of 21 dUCBT recipients and angiogenic activity after 18-hour incubation was quantified as described in Methods. Representative images of patients with high (pt1) and low (pt2) plasma angiogenic activity are shown. (B) Total segment length positively correlates with CD4⁺ (r = .45, P = .04), CD4⁺CD45R0⁺ (r = .49, P = .03), and CD4⁺CD25⁺ T cells (r = .49, P = .03) after dUCBT. Number of segments and meshes also positively correlated with T cell subsets and TRECs but fell short of statistical significance (data not shown).

Prognostic Value of Angiogenic Cytokines for Clinical Outcomes of dUCBT

Previous studies have shown that thymic recovery is a key component of immune reconstitution after UCBT and is associated with improved clinical outcomes [33,38]. Our present results showed that angiogenic factors correlate with thymic recovery and quantitative T cell reconstitution. Previous studies also showed that angiogenic factors can be used as biomarkers for GVHD, NRM, or relapse after HSCT [18,19,26,44,45], but data in the setting of UCBT are sparse. For these reasons we examined the association of VEGF, ANG-1, ANG-2, and TM with long-term clinical outcomes of dUCBT (Table 4).

Table 4. Multivariable Cox Model with Time-Dependent Variable Adjusting for Age (N = 27)^*.

		Hazard Ratio	95% Confident Interval	P
VEGF	OS	1.11	.94-1.26	.17
	PFS	1.78	1.13-2.76	.01
	NRM	1.04	.21-4.07	.96
	Relapse	1.95	1.21-3.13	.007
	cGVHD	1.73	.72-4.04	.22
ANG-1	OS	1.01	.86-1.15	.95
	PFS	.98	.84-1.11	.74
	NRM	.76	.44-1.07	.20
	Relapse	1.05	.89-1.21	.52
	cGVHD	.96	.68-1.20	.78
TM	OS	1.57	1.15-2.18	.007
	PFS	1.02	.78-1.30	.86
	NRM	1.97	1.22-3.9	.02
	Relapse	.80	.54-1.11	.21
	cGVHD	1.41	1.01-1.93	.043
ANG-2	OS	1.07	.97-1.15	.10
	PFS	1.04	.99-1.09	.14
	NRM	1.06	.97-1.13	.15
	Relapse	1.01	.96-1.07	.62
	cGVHD	1.01	.80-1.08	.90

^*Because of a small numbers of events, Cox univariable analysis was performed for NRM and cGVHD. Hazard ratio is per 100 pg/mL for VEGF and ANG-2 and per 1000 pg/mL for ANG-1 and TM.

In Cox regression analysis, high TM levels over time were associated with worse OS (P = .007) and a higher incidence of NRM (P = .02) and cGVHD (P = .043) (Table 4). Notably, patients who developed cGVHD had significantly higher levels of TM both before transplant and during the first 100 days after dUCBT (Figure 4). Similarly, patients who died of NRM had approximately 50% to 100% higher TM levels at baseline and throughout the first year after dUCBT, which is reflected in the Cox model of NRM (Supplementary Table 2). The causes of death for these 5 patients were infections in 2 (Epstein Barr virus (EBV)/Post-transplant lymphoproliferative disease (PTLD)), organ failure in 1 (pulmonary/idiopathic pneumonia syndrome), secondary malignancy in 1 (donor-derived leukemia), and trauma in 1 (subdural hemorrhage). Four of these 5 patients had developed cGVHD before death. In Cox regression analysis, we also found that high levels of VEGF over time were associated with a higher risk of relapse (P = .007) and worse PFS (P = .01) (Table 4). However, despite the increased hazard of relapse over time with higher VEGF, the difference between VEGF levels in relapsed versus relapse-free patients at individual time points before dUCBT and during the follow-up period (excluding samples collected after relapse) was not significant (Supplementary Table 3). VEGF was not associated with cGVHD or NRM. Furthermore, neither ANG-1 nor ANG-2 were associated with any long-term outcomes of dUCBT (Table 4).

Figure 4.

Median TM levels in patients with and without cGVHD. Patients who developed GVHD had higher TM levels at baseline (P < .001), at 4 weeks(P = .0045), 8 weeks (P = .01), and 100 days (P = .056) after dUCBT compared with patients without cGVHD.

It should be noted that the present results included analysis of patients who survived at least 100 days after transplantation, so that at least 3 post-transplant follow-up samples had been collected. For this reason, 4 patients who died before day 100 and 1 patient with missing samples who died on day 122 were excluded from the study because of inadequate follow up samples. All 5 excluded patients died of NRM (2 died of EBV/PTLD and 3 of sepsis/multisystem organ failure).

Discussion

The role of neovascularization in the hematopoietic recovery and outcomes after HSCT is being increasingly recognized [5], but such studies in the context of UCBT are lacking. Here we analyzed the biologic impact and clinical correlatives of angiogenic factors in adult recipients of dUCBT. We observed a marked universal decrease of VEGF at 4 weeks after UCBT, with substantial recovery approaching pretransplant levels at 100 days. ANG-1, the second proangiogenic cytokine, followed parallel kinetics. This observation is in contrast to a previous report in patients undergoing HSCT from adult donors who displayed recovery of VEGF to pretransplant levels by day 15 [20]. In that study, VEGF levels correlated with CD34⁺ cell dose and neutrophil recovery, suggesting that donor-derived HSCs and their progeny are the main source of circulating VEGF after transplantation [20]. Thus, it is likely that the delayed recovery of VEGF levels in recipients of UCBT is due to the lower dose of CD34⁺ cells and the delayed reconstitution of neutrophils and platelets. However, other factors contributing to the levels of VEGF, such as disease type and status, conditioning, or GVHD prophylaxis, cannot be excluded.

Studies examining the prognostic value of VEGF levels for clinical outcomes of HSCT are contradictory. Some studies have reported that high VEGF levels are associated with the development or severity of GVHD [18,46], whereas others have found that VEGF either has a protective effect against severe acute GVHD [19,20,47] or no correlation with GVHD [17]. These discrepancies might be due to assessment of VEGF at different time points after HSCT in each of these studies [47,48]. In addition, the use of serum versus plasma might be a contributing factor, because VEGF is released by activated platelets leading to higher VEGF levels when serum is used [47]. Thus, values in serum samples may not reliably reflect circulating VEGF levels. The protective role of VEGF against GVHD is supported by a murine model of HSCT where VEGF blockade led to aggravation of experimental GVHD [49], as well as by analysis of VEGF gene polymorphisms in humans undergoing HSCT [50]. In our study, it was not feasible to assess association between VEGF and acute GVHD because of the low incidence. No association was observed between VEGF levels and cGVHD, but the sample size might have precluded the detection of such association. Notably, patients with persistently elevated plasma levels of VEGF post-transplant had higher relapse rate and worse PFS. However, we did not observe significantly worse OS, suggesting that some of the relapsed patients might have responded to salvage therapy. Our findings are consistent with previous observations that elevated VEGF is associated with worse outcomes in patients with hematologic malignancies after chemotherapy [14-16] and that increased pretransplant VEGF levels are associated with higher rate of relapse and worse PFS after allogeneic HSCT from adult donors [17].

It has been previously observed that VEGF has immunomodulatory effects. In murine models, pathologically high systemic levels of VEGF led to inhibition of T cell development and profound thymic atrophy, similarly to what has been observed in cancer patients with elevated circulating VEGF [51-53]. These inhibitory effects are mediated via VEGFR-2 signaling, whereas VEGFR-1 mediates the opposite outcome [51]. Importantly, bone marrow progenitors from VEGF-treated mice have dramatically enhanced ability to colonize the thymus of irradiated hosts [52], suggesting that VEGF exposure results in an increase of thymus-committed progenitors. VEGF also plays an important role in the development of thymic vasculature [54]. In neonatal mice, VEGF is highly expressed in thymic epithelial cells, and to a lesser extent in specific thymocyte subsets, and mediates development of dense immature vasculature, whereas inhibition of VEGF leads to profoundly reduced thymopoiesis [55]. Furthermore, thymic reconstitution after HSCT in neonatal mice is VEGF-dependent [36]. VEGF is also implicated in the reparative angiogenesis that accompanies thymic recovery after cyclophosphamide-induced acute thymic involution in adult rats [37]. The presence of VEGF in normal thymus has been documented in humans [56]. In our cohort, we observed a positive correlation between plasma levels of VEGF and CD4⁺ lymphocytes and between VEGF and TREC levels at 6 months after UCBT. By use of an in vitro angiogenesis assay, we evaluated the angiogenic activity of patient plasma and were able to determine that VEGF levels assessed by ELISA in patients' plasma were biologically active, as they could drive HUVEC differentiation in a concentration-dependent manner. In turn, patients with higher plasma angiogenic activity had enhanced recovery of CD4⁺ cells. ANG-1 levels also positively correlated with the thymus-derived naive CD4+CD45RA+and CD8+CD45RA+ subsets.

Our findings strongly suggest that angiogenic factors might be involved in thymic regeneration and T cell reconstitution after UCBT. It should be noted that our studies cannot determine whether proangiogenic factors might promote thymic reconstitution after UCBT by mediating the differentiation and recruitment of thymus-committed progenitors from HSCs to the thymus or by mediating the development of thymic vasculature, but based on studies in animal models [37,52,54] both these VEGF-mediated functions could have active roles in thymic regeneration after UCBT. Because VEGF levels also correlated with memory CD4⁺ T cells and Tregs, VEGF may additionally be involved in thymus-independent mechanisms of T cell reconstitution, such as T cell differentiation in secondary lymphoid organs or T cell trafficking. Further research is required to investigate how angiogenic factors may impact immune reconstitution post-transplant and the differential effects of VEGF and ANG-1 on CD8⁺ T cell recovery. Finally, the observation that high VEGF levels correlate both with T cell reconstitution and relapse risk might seem contradictory. It is possible that similar VEGF-dependent proangiogenic mechanisms might be involved in tumorigenesis as well as in T cell reconstitution after HSCT. VEGF also displayed the strongest correlation with Treg counts, suggesting that VEGF might preferentially promote expansion of Tregs that subsequently mediate tumor immune escape.

Previous studies have shown no significant increase in plasma TM levels after autologous transplantation [32,45,57]. In contrast, TM increases after allogeneic transplantation [45,57], likely as a result of endothelial injury induced by the conditioning, attack by alloreactive donor T cells or microangiopathy induced by calcineurin inhibitors and rapamycin used for GVHD prophylaxis [58,59]. Elevated levels of TM have been associated with several post-transplant complications characterized by endothelial injury or dysfunction, such as veno-occlusive disease, sepsis, thrombotic microangiopathy, or neurologic complications [32,45,60-63], as well as with the occurrence and severity of GVHD [44,48,57]. Furthermore, high levels before or after HSCT are associated with increased NRM [44,48] and lower OS [64]. Data in UCBT are limited, but 1 study including 8 patients reported relative stability of TM in UCBT recipients compared with recipients of peripherally mobilized HSCs, treated with similar conditioning [65]. The investigators speculated that this difference might be because cord blood T cells potentially provoke less endothelial damage due to decreased alloreactivity. In our study we observed that TM levels did not increase significantly from baseline during the first year after dUBCT. We found that patients who developed cGVHD had significantly higher levels of TM before conditioning and in the early post-transplant period, corroborating previous observations in recipients of HSCT from adult donors [65]. In addition, higher TM levels before dUCBT and in the post-transplant period were associated with increased incidence of NRM and worse OS, despite the fact that no death was attributable to GVHD in this cohort. Hence, TM may serve as a prognostic marker for both cGVHD and NRM after dUCBT.

We found no prognostic significance of ANG-2 on longterm outcomes after UCBT, in contrast to TM. Previous studies have shown that elevated pretransplant ANG-2 levels predict worse PFS after HSCT for high-risk myeloid malignancies [26]. Also, high ANG-2 levels after HSCT have been associated with NRM, GVHD severity, and complications related to endothelial damage [44]. Several factors might account for the lack of prognostic significance of ANG-2 levels on clinical outcomes in our dUCBT cohort. First, our study population included a large fraction of patients with lymphoid malignancies, where an effect of ANG-2 has not been identified. Second, very few patients had ANG-2 levels > 2000 pg/mL at any point, a previously determined level of prognostic significance [44]. Third, it has been suggested that the prognostic effect of ANG-2 on NRM is mainly due to GVHD mortality [44], but we observed no GVHD-related NRM in our cohort. Fourth, 1 study found correlation between ANG-2 and post-transplant outcomes only when examined in association with VEGF [18]. Because of the limited sample size, we were not able to perform such analysis.

In contrast to the proangiogenic cytokines VEGF and ANG-1, there was an inverse correlation between TM and reconstitution of T cell populations, especially the naive subsets, as well as TRECs. ANG-2 levels showed similar pattern. These findings likely reflect the negative impact of GVHD and its treatment—and possibly other complications associated with endothelial damage—on thymopoiesis and immune reconstitution. Because recipient age is a major determinant of immune reconstitution post-transplant, we examined whether age might also impact the correlation of the above cytokines with T cell subsets (Supplementary Table 1). Although our findings suggest a tentative age effect for VEGF and ANG-1, a larger study is warranted for definitive conclusions.

Our observations support the significance of angiogenic factors as prognostic markers for clinical outcomes after dUCBT but also reveal notable differences compared with previous findings in conventional HSCT. However, it is not possible to conclude whether these differences are related to the properties of the UCB grafts or to other confounding factors, including the use of antithymocyte globulin. Because we excluded patients with early deaths to ensure adequate number of follow-up samples per patient for our correlative studies, our observations are only applicable to patients who survive at least 100 days. Thus, the prognostic value of angiogenic cytokines for early post-dUCBT NRM requires further confirmation.

In summary, we determined that plasma levels of the 2 proangiogenic factors, VEGF and ANG-1, positively correlate with T cell subsets and TRECs after dUCBT, whereas 2 markers of endothelial damage, TM and ANG-2, show an inverse correlation with naive T cells and TRECs, suggesting that angiogenesis and intact endothelium might facilitate thymic recovery and immune reconstitution after dUCBT. Further, high VEGF levels over time are associated with worse PFS and increased relapse rates post-transplant, whereas TM levels are associated with cGVHD and NRM. Thus, circulating angiogenic factors might serve as prognostic markers for clinical outcomes after UCBT.