Introduction

Double unit cord blood (dCB) transplantation (dCBT) is efficacious for adults with high-risk hematologic malignancies and has been associated with comparable progression-free survival to that of unrelated donor transplantation in multiple series [1,2,3]. While we and others have demonstrated high rates of sustained donor engraftment after dCBT [4,5,6,7], delayed count recovery is common. For example, myeloablated dCBT recipients at our center engraft at a median of 24 days [5]. Slow engraftment can increase morbidity, prolong hospitalization, and increase costs. A novel approach to abrogate prolonged cytopenia pioneered by Fernandez et al. [8,9,10,11] and others [12,13,14], and an alternative to ex vivo expanded CB, is the combination of a CB graft with peripheral blood-derived haplo-identical or third-party donor CD34+ cells. This strategy aims to facilitate early haplo-identical (or third party) donor-derived neutrophil recovery (myeloid bridging) until CB engraftment is achieved. This platform, however, is not standardized and the efficacy of myeloid bridging in the absence of antithymocyte globulin (ATG) is unknown.

To address this question, we have investigated adding haplo-identical CD34+ cells (haplo-CD34+) to dCB grafts (haplo-dCB) in patients transplanted with myeloablative conditioning and no ATG. Herein, we report the kinetics of engraftment after these serotherapy-free haplo-dCB transplants (haplo-dCBT). Adult patients received standard immunoprophylaxis with cyclosporine-A (CSA) and mycophenolate mofetil (MMF). ATG was not used due to its adverse impact on immune reconstitution [15,16,17,18,19,20,21] and the substantial evidence of increased mortality in ATG-based CBT [19, 20, 22,23,24,25,26,27]. Double unit CB grafts were used to enhance safety given transplantation of CB combined with haplo-CD34+ cells has not previously been investigated in an ATG-free setting. In addition, the use of dCB grafts permits comparison of engraftment with historical dCBT controls transplanted with identical conditioning and immunosuppression but without haplo-CD34+ cells. Our primary aim was to determine the speed and success of sustained neutrophil recovery after haplo-dCBT. Our hypothesis was that the addition of a haplo-CD34+ graft would provide a haplo-identical donor-derived myeloid bridge prior to sustained CB-derived engraftment.

Methods

Patients

Patients were treated on a phase II trial (clinicaltrials.gov NCT01682226) between September 2012 and December 2017. The trial was conducted in accordance with the Declaration of Helsinki and was approved by the Memorial Sloan Kettering Cancer Center Institutional Review/Privacy Board. This trial enrolled pediatric and adult patients with high-risk hematologic malignancies without a suitable human leukocyte antigen (HLA)-matched related or unrelated donor, who had a suitable CB graft and a suitable haplo-identical donor. For the purposes of this analysis, only adult haplo-dCBT recipients were included to permit comparison with the engraftment kinetics of historic adult dCBT controls. In addition, two patients who underwent identical haplo-dCBT under Single Patient Use were included (one severe aplastic anemia, one whose insurance denied clinical trial participation). All patients were assayed for HLA antibodies as previously described [28]. Antibody titers with mean fluorescence intensity > 1000 were considered positive.

CB graft selection

Unit selection was based on unit quality/bank of origin, total nucleated cell (TNC) dose, and donor-recipient HLA-match. Units contained a minimum cryopreserved TNC dose of 1.5 × 107/kg and were ≥4/6 HLA-A, -B antigen, -DRB1 allele matched to the recipient. Cryopreserved CD34+ cell dose and 8-allele HLA-match were also considered in CB graft selection [5, 29, 30]. The presence of donor-specific HLA antibodies (DSA) against one or both CB units was not a contraindication to unit selection [28]. The HLA-match of the units to each other or the haplo-identical donor was not considered.

Haplo-identical donor selection and collection

Haplo-identical grafts were derived from mobilized peripheral blood; bone marrow harvests were not permitted even in the setting of poor mobilization. Younger adult donors were given priority with emphasis upon availability, compliance, avoidance of a large donor-recipient weight discrepancy, and adequacy of peripheral access. Donors against whom the recipient had DSA were avoided in the latter phase of the trial.

Donors were mobilized with 10 mcg/kg of granulocyte colony stimulating factor (G-CSF) rounded to vial size subcutaneously daily for 5 days. Initially only one collection was performed. The study was later amended to allow a second leukapheresis if the first yielded <3 × 106/kg CD34+ cells (before CD34+ selection). Grafts were CD34+ cell selected using the CliniMACS CD34 Reagent System (Miltenyi Biotech, Gladbach, Germany) under an Investigational New Device from the US Food and Drug Administration. To guard against permanent haplo-identical donor engraftment, the goal for the maximum haplo-identical graft CD3+ cell dose was 8 × 103/kg. Initially the haplo-CD34+ cell dose was capped at 3 × 106/kg. Subsequently, the target CD34+ cell dose was increased to ~5 × 106/kg without an upper limit.

Conditioning regimens, immunoprophylaxis, and growth factor support

Patients received myeloablative conditioning [4, 31]. The intensity was based on diagnosis, disease status, age, and hematopoietic cell transplant co-morbidity index (HCT-CI) [32] score. High dose conditioning (cyclophosphamide (Cy) 120 mg/kg, fludarabine (Flu) 75 mg/m2, and total body irradiation (TBI) 1375 cGy (Cy 120/Flu 75/TBI 1375)) was considered for fit patients < 30 years with hematologic malignancies. Remaining patients received intermediate intensity conditioning (Cy 50 mg/kg, Flu 150 mg/m2, thiotepa (Thio) 10 mg/kg, TBI 400 cGy (Cy 50/Flu 150/Thio 10/TBI 400)) with a reduced Thio dose (5 mg/kg) in patients 60–70 years or those with HCT-CI score ≥ 5.

CSA and MMF (15 mg/kg every 8 h) were started intravenously on day −3 for graft-versus-host disease prophylaxis. No patient received ATG. All patients received G-CSF 5 mcg/kg/day from day 7 posttransplant until neutrophil recovery. In patients with a second neutrophil nadir, G-CSF was resumed until sustained engraftment was achieved.

Engraftment monitoring and definitions

A white cell count (WCC) differential was obtained once the WCC was >0.5 × 109/L. Neutrophil recovery was defined as the first of three consecutive days of neutrophils ≥0.5 × 109/L. Platelet recovery was the first day of ≥20 × 109/L platelets without transfusion for 7 consecutive days. Graft failure was defined as requirement for a second stem cell infusion or death without neutrophil recovery on day 28 or later.

Haplo-identical and CB donor chimerism were monitored using PCR amplification of informative recipient and donor short tandem repeats. Whole blood assays were done on days 14, 28, 60, 100, 180, and 365 posttransplant. White cell subset chimerism analyses were performed in sorted myeloid, T-, B- and NK-cell subsets (purity > 95%) on days 28, 100, and 365 posttransplant. Analysis of lineage-specific chimerism was foregone if the purity threshold was not achieved or if the specific cell subset count was too low. Of the two CB units infused, the dominant (or engrafting) CB unit was the only one detected or the one with sustained > 50% contribution to CB-derived chimerism.

Statistical methods

Our objective was to determine the speed and success of sustained myeloid recovery after haplo-dCBT. Success was arbitrarily defined as neutrophil recovery by 2 weeks posttransplant (prior to or on day 14). Cumulative incidences of neutrophil and platelet recovery were estimated considering early death as a competing risk. Chimerism was analyzed using summary statistics and box-and-whisker diagrams. Correlation of haplo-CD34+ and CB cell doses with days to neutrophil recovery was evaluated using Spearman’s rank correlation coefficient. Cell doses of dominant and nondominant CB units were compared using the Wilcoxon signed-rank test. Univariate and multivariate logistic regression analyses were performed in patients who achieved sustained CB engraftment to evaluate factors associated with higher odds of successful haplo-CD34+ myeloid bridging. All variables with p < 0.10 in univariate analysis were included in the multivariate model. Transplant-related mortality (TRM) was compared across WCC recovery groups using Gray’s test in a day 28 landmark analysis considering relapse as a competing risk. Results with two-tailed p values < 0.05 were considered significant. All analyses were conducted using R statistical software, version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Patient and graft characteristics

Seventy-eight patients (median age 48 years (range 21–68), median weight 82 kg (range 48–138)) underwent haplo-dCBT. Thirty-seven patients (47%) were male and 44 (56%) were CMV seropositive. Diagnoses included 54 (69%) acute leukemias, 10 (13%) myelodysplasia/myeloproliferative diseases, 13 (17%) lymphomas, and 1 aplastic anemia. Three patients were second allograft recipients. Conditioning was high dose (Cy 120/Flu 75/TBI 1375, n = 1) or intermediate intensity (Cy 50/Flu 150/Thio 10/TBI 400 (n = 64), Cy 50/Flu 150/Thio 5/TBI 400 (n = 13)).

Infused CB unit (n = 156) and haplo-CD34+ (n = 78) graft characteristics are shown in Table 1. CB units had a median infused TNC dose of 2.3 (range 1.0–5.7) × 107/kg/unit and a median infused viable CD34+ cell dose of 1.1 (range 0.1–3.1) × 105/kg/unit. The median infused viable CD3+ cell dose was 2.9 (range 0.3–8.0) × 106/unit. The majority of units were 4/6 HLA-A, -B antigen, -DRB1 allele matched to the patient, and the median CB unit-recipient HLA-allele match was 5/8 (range 2–7/8). Seven patients (9%) had DSA against their CB graft.

Table 1 Infused CB and haplo-identical graft characteristics.
Full size table

Haplo-CD34+ grafts were most commonly procured from children (46%) or siblings (31%). Haplo-identical donors had a median age of 33 years (range 15–71). The median infused CD34+ dose was 5.2 (range 1.1–16.8) × 106/kg. The median infused CD3+ cell dose was 1.6 (range 0.3–13.7) × 103/kg and approximately three logs lower than that of the CB units. The majority of haplo-CD34+ grafts (n = 61, 78%) were 4/8 HLA-allele matched to the patient. Eleven patients (14%) had DSA against their haplo-identical graft.

Overall hematopoietic engraftment

Of the 78 analyzed patients, 75 engrafted, 2 had graft failure, and 1 heavily pretreated patient died on day 14 from veno-occlusive disease and multi-organ failure. The cumulative incidence of sustained neutrophil recovery for the entire cohort was 96% (95% CI: 87–99). The day 100 cumulative incidence of platelet recovery was 87% (95% CI: 77–93).

In the 75 engrafting patients, sustained engraftment was mediated by a dominant CB unit. The dominant units had a median infused viable CD34+ cell dose of 1.23 (range 0.24−2.95) × 105/kg and a median infused viable CD3+ cell dose of 1.02 (range 0.14−3.4) × 106/kg. The median HLA-match of the dominant CB unit to the recipient and to the haplo-identical graft were 5/8 (range 3–7/8) and 3/8 (range 1–7/8), respectively. The pattern of whole blood chimerism is shown in Fig. 1. Overall, haplo-identical grafts predominated early posttransplant (median day 14 chimerism 88% (range 0–100)). However, no haplo-identical graft was detected in 51% of evaluable patients at day 28, in 61% at day 60, and in 78% at day 100 posttransplant. In addition, the haplo-identical graft comprised only a minor component of donor chimerism in nearly all remaining patients. Concurrently, the dominant CB unit whole blood chimerism increased from a median of 10% (range 0–100) at day 14 to 91% (range 0–100) at day 28, 100% (range 0–100) at day 60, and 100% (range 12–100) at day 100 and beyond. White cell subset chimerism analysis revealed that loss of the haplo-identical graft was associated with early dominant CB unit chimerism in the T-cell fraction.

Fig. 1: Pattern of whole blood chimerism in engrafting patients in the first year after haplo-dCBT (n = 75).
figure 1

The contribution of the haplo-identical donor (red), the dominant (engrafting) CB unit (blue), and the nondominant (non-engrafting) unit (green) to whole blood chimerism at each time point posttransplant is shown. While the haplo-identical donors contributed to initial hematopoiesis, one CB unit predominated by day 28 in the majority of patients with the median donor chimerism being 100% the dominant CB unit by day 100 and beyond.

Full size image

At a median survivor follow-up of 3 years and 9 months (range 1–6 years) all evaluable patients maintain engraftment with a dominant CB unit.

Patterns of hematopoietic recovery

While 75 of 77 evaluable patients had sustained CB engraftment that was mediated by a dominant CB unit, the success of obtaining an early haplo-identical donor-derived myeloid bridge was variable between patients. Three distinct engraftment patterns were observed (Table 2, Fig. 2a–d) and were characterized by distinct chimerism patterns associated with the speed of rejection of the haplo-CD34+ graft by the dominant CB unit (Figs. 3–5).

Table 2 Engraftment patterns in haplo-dCBT recipients (n = 77 evaluable patients)a.
Full size table
Fig. 2: Probability of neutrophil recovery after haplo-dCBT by engraftment group (Groups 1–3, n = 75).
figure 2

a Group 1 patients (n = 34) had early sustained myeloid recovery by day 14 posttransplant (median 12 days). b Group 2 patients (n = 20) had transient myeloid recovery (median neutrophil recovery 12 days) followed by a second nadir preceding sustained engraftment (median second neutrophil recovery 26.5 days). c Percentage of Group 2 patients with a neutrophil count ≥ 500/dL per posttransplant day. d Group 3 patients (n = 21) had delayed myeloid recovery (median 25 days).

Full size image
Fig. 3: Chimerism patterns after haplo-dCBT in patients with an early myeloid bridge (Group 1, n = 34).
figure 3

Whole blood chimerism analysis (a) revealed predominant engraftment of the haplo-identical donor early posttransplant (median day 14 chimerism 95%, range 24–100) with subsequent increasing dominant CB unit chimerism. As of day 28 posttransplant, the majority of myeloid cells (b) were derived from the haplo-identical donor, whereas T cells (c) were primarily derived from the dominant CB unit, with progressive increase in dominant CB unit-derived chimerism in all lineages thereafter. Red: haplo-identical donor; blue: dominant (engrafting) CB unit; green: nondominant (non-engrafting) CB unit.

Full size image
Fig. 4: Chimerism patterns after haplo-dCBT in patients with a transient myeloid bridge (Group 2, n = 20).
figure 4

Whole blood chimerism analysis (a) revealed short-lived haplo-identical donor engraftment (median day 14 haplo-identical donor chimerism 82%, range 0–100%) followed by sustained CB-derived hematopoiesis. Both myeloid (b) and T cell (c) lineages were primarily derived from the dominant CB unit as early as day 28. Red: haplo-identical donor; blue: dominant (engrafting) CB unit; green: nondominant (non-engrafting) CB unit.

Full size image
Fig. 5: Chimerism patterns after haplo-dCBT in patients with no myeloid bridge and sustained CB engraftment (Group 3, n = 21).
figure 5

Whole blood chimerism analysis (a) revealed that the majority of Group 3 patients had either no or minimal haplo-identical donor engraftment (day 14 median chimerism 10%, range 0–94%) and the dominant CB unit predominated thereafter. The dominant CB unit was either the only or greatest contributor to myeloid (b) and T-cell (c) lineages at day 28 and beyond. Red: haplo-identical donor; blue: dominant (engrafting) CB unit; green: nondominant (non-engrafting) CB unit.

Full size image

Group 1 patients (34/77, 44%) had early sustained myeloid recovery at a median of 12 days (range 10–14) posttransplant (Fig. 2a). This rapid recovery was almost always mediated by the haplo-identical graft with subsequent transition to CB-derived hematopoiesis. These patients had a high median day 14 haplo-identical graft whole blood chimerism of 95% (range 24–100) (Fig. 3a). Subsequently, they had increasing dominant CB unit chimerism. By day 180, the haplo-identical graft was detected in only a minority (5/28, 18%) of patients (median haplo-identical donor chimerism 7%, range 2–68). By 1 year, haplo-identical cells were only detected in three patients (contributions 3%, 12% and 52%) with two having since converted to 100% dominant CB unit. In white cell subset analyses, the majority of myeloid cells were derived from the haplo-identical donor at day 28 (Fig. 3b). In contrast, T cells were derived from the dominant CB unit (Fig. 3c) with a progressive increase in dominant CB unit-derived chimerism in all lineages thereafter. In Group 1 patients, a higher infused haplo-CD34+ dose was associated with faster neutrophil recovery (r = −0.74, p < 0.001). Platelet recovery was also enhanced in these patients (Table 2).

Group 2 patients (20/77, 26%) had initial myeloid recovery (defined as a neutrophil count ≥0.5 k/mcL for ≥1 day) followed by a second nadir (<0.5 k/mcL for ≥2 consecutive days) preceding sustained engraftment (Fig. 2b, c). The median times to first and second neutrophil recoveries were 12 days (range 11–18) and 26.5 days (range 20–46), respectively. Short-lived haplo-CD34+ engraftment accounted for the transient myeloid bridge (median day 14 haplo-identical donor chimerism 82% (range 0–100)), although in two patients the haplo-identical donor was undetectable by 2 weeks posttransplant (Fig. 4a). This was followed by sustained CB-derived hematopoiesis (median whole blood chimerism 100% dominant CB unit by day 60 and beyond). In addition, myeloid and T-cell lineages were primarily derived from the dominant CB unit as early as day 28 (Fig. 4b, c). Notably, a higher dominant CB unit infused viable CD3+ dose correlated with a slower time to the first (haplo-identical donor-derived) neutrophil recovery (r = 0.43, p = 0.058), whereas a higher dominant CB unit infused viable CD34+ cell dose correlated with a faster time to the second (CB-derived) neutrophil recovery (r = −0.44, p = 0.052), although statistical significance at the 0.05 level was not reached.

Group 3 patients (21/77, 27%) had delayed neutrophil recovery (median 25 days, range 15–33) (Fig. 2d). At day 14 posttransplant, the majority had either no or minimal haplo-identical donor chimerism (median 10% (range 0–94)) in whole blood (Fig. 5a), and no haplo-identical donor was ever detected in nine patients. In addition, seven patients had no myeloid bridge despite a day 14 haplo-identical donor contribution >50%. By day 60, all but two patients were 100% donor with the dominant CB unit, and this unit accounted for hematopoiesis in all patients subsequently. Moreover, in subset analysis, the dominant CB unit was either the only or greatest contributor to all lineages as of day 28 and beyond (Fig. 5b, c). In Group 3 patients, a higher infused viable CD34+ cell dose of the dominant CB unit was associated with faster neutrophil recovery, although this correlation was not significant at the 0.05 level (r = −0.42, p = 0.06).

The two remaining evaluable patients (Group 4, 3%) had graft failure (Table 2). One, a prior allograft recipient, with DSA against the haplo-identical donor and both CB units had failure of haplo-CD34+ and CB engraftment. The second had transient haplo-identical donor-derived neutrophil recovery followed by failure of CB engraftment. CB graft failure was likely due to the very low infused viable CD34+ cell doses of each unit (0.22 × 105/kg and 0.35 × 105/kg, respectively). Both patients were successfully re-transplanted with single CB units.

Factors associated with an optimal myeloid bridge

Next, we investigated the association between graft characteristics and the likelihood of achieving an optimal myeloid bridge (Table 3). This was defined as early, haplo-identical donor-derived, sustained neutrophil recovery by 2 weeks posttransplant (i.e., without a second nadir) prior to CB engraftment. In multivariate analysis, a higher haplo-CD34+ dose (OR: 1.2 (95% CI: 1.01–1.47), p = 0.047) significantly improved the odds of achieving an optimal myeloid bridge. While there was no haplo-identical CD34+ dose threshold that could guarantee optimal bridging, none of the eight patients who received a haplo-CD34+ cell dose <3 × 106/kg had a bridge. Furthermore, a ≥3/8 HLA-match of the dominant CB unit to the haplo-identical donor was also associated with higher odds of optimal bridging (OR: 3.49 (95% CI: 1.27–10.42), p = 0.019).

Table 3 Factors associated with an optimal haplo-identical donor-derived myeloid bridge after haplo-dCBT (n = 75)a.
Full size table

A ≥5/8 HLA-match of the haplo-identical donor to the recipient was associated with optimal myeloid bridging in the univariate but not the multivariate model. Presence of DSA against the haplo-identical graft had no impact. Cell dose and HLA-match of the nondominant CB unit were also not associated with the likelihood of optimal bridging.

Association of optimal myeloid bridging with duration of hospitalization and day 100 TRM

Of 70 engrafted patients discharged from their initial hospitalization, Group 1 patients with sustained myeloid bridge were discharged earlier (median 28 days (range 20–60)) than the Group 2–3 patients with transient or no bridging (median 36 days (range 28–98)). However, the day 100 TRM in Group 1 patients was not different than that of Group 2–3 patients (9% (95% CI: 2–21) versus 15% (95% CI: 6–27), p = 0.388). In addition, optimal bridging in Group 1 patients was not associated with improved immune recovery (Supplementary Fig. 1).

Determinants of CB unit dominance

CB unit dominance was associated with a higher infused viable CD3+ cell dose (median dose 3.3 (range: 0.9–8.0) × 106/kg versus median 2.6 (range: 0.3–6.0) × 106/kg, p < 0.001). In contrast, CB infused TNC dose, infused viable CD34+ cell dose and 8-allele CB unit-recipient HLA-match were not significant (data not shown).

Discussion

The addition of haplo-identical CD34+ cells to CB grafts to enhance myeloid recovery by providing an early myeloid bridge has been investigated as a potential alternative to ex vivo CB expansion. This approach has the advantages that most patients have a suitable haplo-identical donor, CD34+ cell selection is feasible in many centers, and complex CB graft manipulation is not required. Multiple groups have demonstrated that neutrophil recovery is enhanced in most recipients of ATG-based haplo-CBT [8, 12, 14]. Many unanswered questions remain, however, as the influence of the conditioning, immunosuppression, and ATG omission on the likelihood of haplo-identical myeloid bridging has not been investigated.

We report for the first time the efficacy of the addition of haplo-identical CD34+ cells to CB grafts as a strategy to enhance myeloid recovery in a serotherapy-free platform. Investigation of this approach in the absence of ATG is critical given it is well established that use of ATG in adult CBT is associated with increased mortality [15, 23,24,25,26,27]. We acknowledge that the cost of a dCB graft supplemented with haplo-identical CD34+ cells would be prohibitive for most centers. However, in this single center trial, a dCB graft was used to ensure patient safety in case the haplo-identical graft was rejected, but a single unit was inadequate to facilitate sustained hematopoiesis. In addition, given only a single variable was changed (the addition of haplo-CD34+ cells), this approach enables direct comparison with our dCBT only controls who have had neutrophil recovery at a median of 24 days posttransplant [5].

We demonstrate that haplo-dCBT recipients who achieved an optimal myeloid bridge had faster neutrophil and platelet recovery compared to either haplo-dCBT patients who did not bridge or dCBT only historic controls [5]. Optimal myeloid bridging was also associated with earlier hospital discharge. However, bridging was observed in less than half of the ATG-free haplo-dCBT recipients and it was not associated with an improvement in early TRM or immune recovery (Supplementary Fig. 1).

Notably, we found that long-term hematopoiesis was provided by a single dominant CB unit in all engrafted patients (including those with minimal early CB-derived hematopoiesis), and CB graft failure was very rare. Furthermore, dominant CB unit T cells were observed as early as day 28 in almost all patients. High day 28 dominant unit T-cell chimerism was associated with loss of the haplo-identical graft and conversion to dominant CB unit-derived hematopoiesis. This finding suggests that in haplo-dCBT, dominant CB unit T cells are able to reject the non-engrafting unit (as with dCBT alone) and also the haplo-identical graft even despite its higher CD34+ cell dose. Moreover, haplo-identical donor-derived myeloid bridging was more likely with a higher degree of HLA-match between the haplo-identical graft and the dominant CB unit. This novel finding suggests that better HLA-match may slow the speed of haplo-identical graft rejection and is consistent with the observation that, in dCBT, closer unit–unit HLA-match is associated with initial co-engraftment of both units [33].

There were multiple additional findings of interest. As expected [34], early neutrophil recovery was derived from the haplo-CD34+ graft. However, a high haplo-identical donor chimerism early posttransplant did not guarantee myeloid bridging. Also, a higher dominant CB unit infused viable CD3+ cell dose was associated with delayed time to first (haplo derived) neutrophil recovery in patients with a transient bridge. These observations could be explained by dominant CB unit T-cell mediated inhibition of haplo-identical hematopoiesis. It was also notable that, like dCBT alone [4, 33, 35, 36], a higher infused CD3+ cell dose was associated with CB unit dominance. Moreover, in patients with either a transient or no bridge a higher dominant unit CD34+ cell dose was associated with faster sustained neutrophil recovery [5].

Our findings are in marked contrast to ATG-based haplo-CBT series that have been characterized by accelerated haplo-identical donor-derived neutrophil recovery in the majority of patients [11, 12], slower transition to CB-derived hematopoiesis [12], and higher rates of CB graft failure [34, 37, 38]. In addition, unlike our findings, low CB chimerism early posttransplant predicted for CB graft failure in those series [34, 37]. Also, a higher haplo-identical cell dose [37, 39] and a better haplo-identical donor-recipient HLA-match [39] have been associated with failure of CB engraftment in ATG-based haplo-CBT. Conversely, these graft characteristics were associated with an increased likelihood of bridging and did not lead to CB graft failure in our study. Finally, as with ATG-free CBT [19, 20, 40] and in contrast to ATG-based haplo-CBT [18], T-cell recovery after haplo-dCBT was prompt regardless of bridging [40].

These differences have important implications, considering that ATG omission in adult CBT is associated with reduced mortality [23,24,25,26,27] and improved immune reconstitution [15,16,17, 21, 40]. Most critical is that myeloid bridging is not guaranteed after serotherapy-free haplo-dCBT and the engrafting CB unit will mediate sustained engraftment. Consequently, the CB graft characteristics are paramount and it should not be assumed that such a strategy could be used to safely transplant small CB units that would otherwise be inadequate for transplantation [11, 12, 14, 39]. It is likely that these differences would also apply in ATG-free haplo-CBT using single units, although it is possible that the higher T-cell dose and unit–unit interactions of a double unit graft could augment the speed of haplo-identical graft rejection.

While we have not performed formal cost and resource utilization analyses in comparison to dCBT historic controls, our results suggest that haplo-dCBT is not cost-effective considering the high cost of graft acquisition and the limited efficacy. They also support the pursuit of alternative strategies to improve CB myeloid recovery including enrichment of the CB inventory with high dose units [41, 42], optimization of unit selection incorporating unit quality and CD34+ cell dose [30], and investigation of ex vivo expansion or augmentation of stem cell homing [43,44,45,46,47,48,49,50]. Ongoing investigation of such strategies is warranted given that multiple series have demonstrated high survival after CBT [2, 3] and that some patients, especially those of African ancestry, do not have suitable haplo-identical donors [51].

Our findings also have important implications for strategies that combine unmanipulated CB with any third party or ex vivo expanded T-cell depleted product, as well as the design of future clinical trials aiming to enhance engraftment post-CBT. They suggest that a better HLA-match of a third-party product to the unmanipulated CB unit, and a higher dose of third-party CD34+ cells, could improve the likelihood of bridging. Notably, however, CB graft supplementation with a high dose CB-derived myeloid product that was not HLA-matched did not enhance engraftment in the trial of Milano et al. [52], likely due to early product rejection by the unmanipulated CB graft. Therefore, the requirement for at least partial HLA matching cannot be obviated. This greatly limits haplo-identical cell supplementation, as purposeful HLA matching of the haplo-identical donor to a CB graft is not feasible. Therefore, these observations are now forming the basis of a novel clinical trial at our center in which we will investigate engraftment after transplantation of a single unmanipulated CB unit supplemented by a second, CD34+ selected, ex vivo expanded unit specifically chosen to be well matched to the first. The aim of this trial will be to enhance the potential for a myeloid bridge by close HLA matching of expanded T-cell depleted CB cells to a T-cell replete single unit CB graft.