Dear Editor,

Acute myeloid leukemia (AML) patients carrying complex karyotype or aneuploidies have a very poor prognosis, with a 5-year overall survival (OS) <20%1. We and others have shown that these patients are characterized by high genomic instability, along with defects of DNA damage response (DDR) genes2,3.

Partner and localizer of BRCA2 (PALB2) is a key player in the homology recombination (HR) pathway, since it functions as a scaffold for the HR complex4. PALB2 monoallelic germline mutations are associated with an increased risk of developing breast, pancreatic, and ovarian cancers, whereas biallelic mutations lead to Fanconi anemia5. Genomic lesions, including sequence mutations and copy-number alterations (CNAs) of HR and DDR genes, are rare events in sporadic cancers and no recurrent genomic alterations in these genes have been reported so far in AML.

Here, we studied the genomic alterations of PALB2 in AML, including its molecular interactions and prognostic relevance.

Single-nucleotide polymorphism (SNP) arrays of 119 adult AML cases and whole-exome sequencing (WES) data of 68 adult AML cases were downloaded from the NGS-PTL project repository. Mutational data from the Beat-AML study (n = 531) and acute erythroid leukemia (AEL) cohort (n = 159) were retrieved from https://www.cbioportal.org/ and https://pecan.stjude.cloud/, respectively. The GSE23452 SNP array (n = 144), the GSE14468 (n = 486), and The Cancer Genome Atlas (TCGA, n = 200) AML transcriptomic datasets were obtained from Gene Expression Omnibus collection (https://www.ncbi.nlm.nih.gov/gds) and https://cancergenome.nih.gov/, respectively. SNP array data files for the NGS-PTL cohort (n = 119) are available through the Gene Expression Omnibus public database (accession no. GSE160982). Detailed dataset information and analysis are described in the Supplemental Material section.

We first looked for PALB2 sequence variations in publicly available genomic datasets (TCGA, Beat-AML, NGS-PTL, and AEL): one somatic frameshift mutation was identified in one AML patient (NM_024675:exon3:T51Qfs*2, TCGA cohort) and two missense mutations (NM_024675:exon 4:Q460R, NM_024675:exon 4:G165C) were identified in two AEL cases. We then asked whether PALB2 was affected by CNAs. PALB2 CN loss was detected in 12/233 patients (5.2%), with a minimal common deleted region of 6.6 kb at the C terminus. This region included exon 12, which encodes for the domain responsible for the interaction with the HR proteins RAD51 and BRCA2 and the DNA repair-related polymerase POLH (Fig. 1A and Supplementary Table 1). PALB2 deletion was associated with CN loss of genomic regions frequently altered in poor-prognosis AML, including chromosome 5q and 17p13 (p < 0.05 and false discovery rate <0.25; Fig. 1B and Supplementary Table 2). AML patients carrying PALB2 genomic loss were older than wild-type cases (median age of 65 years vs. 59 in the wild-type cohort, p = 0.044; Table 1) and were characterized by a poor prognosis (75.0% vs. 37.1% of PALB2 wild-type cases, p = 0.016; Table 1).

Fig. 1: PALB2 CN loss in AML and its prognostic relevance.
figure 1

A Graphical representation of PALB2 mRNA and protein functional domains. The coiled-coil domain (red) is involved in the interaction with BRCA1; the chromatin-association motif (ChAM, blue) is required for chromatin association and it mediates nucleosome association; the WD40 domain (green) is responsible for the interaction with RAD51, BRCA2, and POLH. Gray bars represent CN loss detected in AML patients. B Frequency (%) and the significance of genomic CN changes between PALB2 loss and wild-type patients. Blue indicates CN gains and red indicates CN loss. Results are summarized for a group of patients (PALB2 loss vs. PALB2 wild-type). C Kaplan–Meier curve of PALB2 loss (blue) vs. PALB2 wild-type (green) AML. D Forest plot for the multivariable Cox regression analysis. yrs years, HR hazard ratio, CI confidence interval.

Full size image
Table 1 Comparison of biological and molecular features between PALB2 loss and PALB2 wild-type AML cases.
Full size table

To better understand the phenotype of AML carrying PALB2 CN loss, we performed pathway analysis of genes significantly affected by CNAs in those patients (compared with PALB2 wild-type cases, p < 0.05). WNT and ERBB2 signaling pathways, cell adhesion, cytokines, and inflammatory response, metabolism, necroptotic process, senescence, and autophagy in cancer were the most relevant pathways enriched of genes affected by CN loss (p < 0.01, adj-p < 0.25; Supplementary Table 3). By integrating CN and mutational data, we observed an association between PALB2 deletions and TP53 genomic alterations: 8/12 PALB2 loss patients had TP53 somatic mutations (66.7%), in comparison to 19.6% of PALB2 wild-type cases (p < 0.001; Table 1). Six patients with PALB2 loss and TP53 mutations had biallelic inactivation of TP53 (loss of heterozygosity: n = 1; CN loss: n = 5), indicating that TP53 functionality was potentially lost in these cases. To identify additional mutations that cooperated with PALB2 loss, we integrated CN results with WES data (n = 68) of available AML cases from the NGS-PTL cohort3 that was enriched for patients having complex karyotype (Supplementary Table 4). In this cohort, 6 out of 68 patients (8.8%) carried PALB2 CN loss. In addition to TP53 (n = 3, 50.0%), PALB2 loss patients had mutations in DNMT3A (n = 2, 33.3%), BCOR (n = 1), CDKN2A (n = 1), FLT3 (n = 1), NPM1 (n = 1), NRAS (n = 1), and WT1 (n = 1, 16.6%; Supplementary Fig. 1).

It has been previously reported that heterozygous PALB2 genomic alterations, including deletion and truncating or single-nucleotide variants induced a reduction of protein expression6,7. Here, we analyzed PALB2 expression in leukemic cells and its potential association with patients’ biological or molecular features. Data from two independent cohorts revealed that elderly patients expressed lower levels of PALB2 than younger ones (TCGA: PALB2 median expression 21.3 vs. 23.3 counts per million, respectively, p = 0.007; GSE14468: PALB2 median expression 213.8 vs. 232.3, respectively, p = 0.055; Supplementary Table 5). This result is in line with the above data on AML patients carrying PALB2 genomic loss. Notably, we observed significant differences in PALB2 levels among cytogenetic risk classes in the two cohorts (GSE14468, p = 0.020 and TCGA, p = 0.006). Specifically, patients with favorable prognosis (t(8;21) or inv(16) AML) expressed higher levels of PALB2, while intermediate and adverse risk cases had lower expression (normal karyotype, complex karyotypes, KMT2A-rearranged AML, and other abnormalities; Supplementary Table 5). This observation also recapitulates the association between PALB2 genomic loss and high cytogenetic risk. The association between PALB2 expression and karyotype was not confirmed in the GSE14468 dataset, potentially due to the unbalanced proportion of cases belonging to the different cytogenetic subgroups between the two cohorts.

We then investigated the prognostic role of genetic alterations of PALB2 in AML. Despite the low number of altered cases, PALB2 loss patients had a worse prognosis compared to PALB2 wild-type ones (median OS PALB2 loss: 2.0 months, 95% confidence intervals (CIs): 1.2–not reached; PALB2 wild-type: 16.2 months, 95% CIs: 12.9–19.1; p < 0.0001; Fig. 1C). Moreover, in a multivariate Cox model PALB2 loss negatively affected the prognosis of AML patients with a hazard ratio of 2.74 (95% CIs: 1.42–5.31, p = 0.003; Fig. 1D) and its effect was not dependent on age, cytogenetic risk, and, notably, on TP53 mutational status (which were significant predictors in univariate analysis and were in most cases associated with PALB2 loss). Therefore, PALB2 might be a novel biomarker of poor prognosis in AML.

PALB2 is a tumor suppressor affected by mutations and CNAs in patients with either a familiar history of breast, pancreatic, and ovarian cancers5, or, rarely, in sporadic solid and hematological tumors (https://www.cbioportal.org/). Notably, PALB2 participates also in the Fanconi anemia pathway and biallelic germline mutations were associated with the development of AML, among other childhood cancers4.

In this study, we report that PALB2 is rarely mutated at the somatic level in adult AML. However, in a cohort of 233 patients, PALB2 CN loss was detected in 5.2% of cases and it was associated with TP53 genomic alterations. Large PALB2 genomic deletions were also reported in one AML, one myelodysplastic syndrome8, and four AEL cases9. Remarkably, the co-association of PALB2 and TP53 alterations has been previously described in PALB2-related breast cancer10. Moreover, in our cohort, we detected seven patients with a focal deletion involving exons 11 and 12, which encode for the WD40 domain. Mutations in this domain were reported to disrupt DNA repair activity6.

Enrichment analysis of genes within chromosomal regions affected by CN loss and significantly associated with PALB2 loss pointed to the ERBB2 (HER2) signaling pathway (5/37 enriched pathways, p ≤ 0.0114). These data suggest that the phenotype of PALB2 loss patients might resemble that of HER2-negative breast cancers. Of note, breast cancer patients carrying PALB2 germline mutations were associated with the HER2-negative phenotype10. No familial history of cancer was reported in patients from our AML cohort, suggesting that the detected PALB2 CN loss might be somatic alterations acquired by the leukemic cells. However, we cannot exclude the germline origin of the deletion, since paired-germline samples were not available.

Few studies described an association between PALB2 genomic lesions and patients’ outcome so far11. We here show a negative impact of PALB2 genomic alterations on AML patients’ survival with cases carrying PALB2 loss having shorter OS. This predictive role was independent of other negative prognostic factors (e.g., age, cytogenetic risk, and, notably, TP53 alterations). Nevertheless, a limitation of our study is the heterogeneity of treatments and patients’ age, pointing out the need for a larger and homogeneous AML cohort to validate the potential of PALB2 as a biomarker for poor-prognosis AML.

At the therapeutic level, high PALB2 mRNA levels predicted response to cisplatin–docetaxel in non-small cell lung cancer12, while PALB2-mutated pancreatic cancers were described to benefit from mitomycin C or PARP inhibition (PARPi) treatment8,13. Despite the potential relevance, these studies were characterized by a limited number of patients and the role of PALB2 alterations in therapy response deserves further investigation. This is particularly important in AML, where functional studies elucidating PALB2 role in AML cells are needed. PARPi is currently being explored as therapeutic options alone or in combination with chemotherapy, alkylating, or demethylating agents in AML2. Different studies reported a deregulated expression of HR genes in AML and preclinical studies showed that t(8;21) and t(15;17) AML are extremely sensitive to PARPi2,14,15. Taken together, our data indicate that therapies targeting the HR pathway may become a valuable therapeutic option for the subgroup of AML patients characterized by PALB2 CN loss and dismal outcome.