In this issue of Blood, Borrow et al reanalyze published DNA sequencing data and report, in back-to-back papers,1,2  that the lymphoid enzyme terminal deoxynucleotide transferase (TdT) appears to be involved in the causation of the 2 most common types of gene mutation in human acute myeloid leukemia (AML),3  namely internal tandem duplications of FLT3 (FLT3-ITD) and short insertions/duplications within the final exon of NPM1 (NPM1c). The normal function of TdT is to increase the diversity of the immunoglobulin and T-cell receptor (TCR) loci, by adding nontemplated nucleotides to their variable regions. Other enzymes involved in generating this diversity, such as RAG and AID, can act illegitimately to cause oncogenic mutations in acute lymphoblastic leukemia (ALL).4  Here, the work by Borrow et al proposes TdT as a major mutagen in AML, an unexpected finding that forces a reevaluation of our understanding of myeloid leukemogenesis.

Role of TdT in the causation of FLT3-ITD and NPM1c mutations in AML. (A) In normal DNA replication, the daughter strand (gray) is a perfect complement of the template strand (black). (B) FLT3-ITDs can arise as a result of replication slippage during which the DNA polymerase resumes replication from an upstream position microhomologous with the end of the elongating daughter strand (red). (C) In approximately one-third of FLT3-ITDs, there is no visible microhomology. Instead, TdT is predicted to add nucleotides (blue) to the end of the daughter strand in order to provide the occult (or “missing”) microhomology (of at least 1 basepair) between the daughter and template strands. In this specific instance, a run of 5 Gs was added, only the final one of which was used for microhomology. The first 4 Gs then appear as a “filler” sequence in the final mutant DNA and also serve to keep the FLT3 mRNA reading frame open (the total number of additional nucleotides [21] is divisible by 3). (D) Similarly, a single T nucleotide is predicted to be added to the daughter strand during the formation of the common NPM1 mutation, a 4-bp tandem duplication. Borrow et al provide evidence that the extra nucleotides facilitating the formation of approximately one-third of FLT3-ITDs and almost all NPM1c mutations2  follow a pattern that strongly suggests illegitimate TdT activity. Panels B to D are derived from real examples of mutations depicted by Borrow et al. bp, basepair; nt, nucleotide. Professional illustration by Patrick Lane, ScEYEnce Studios.

Role of TdT in the causation of FLT3-ITD and NPM1c mutations in AML. (A) In normal DNA replication, the daughter strand (gray) is a perfect complement of the template strand (black). (B) FLT3-ITDs can arise as a result of replication slippage during which the DNA polymerase resumes replication from an upstream position microhomologous with the end of the elongating daughter strand (red). (C) In approximately one-third of FLT3-ITDs, there is no visible microhomology. Instead, TdT is predicted to add nucleotides (blue) to the end of the daughter strand in order to provide the occult (or “missing”) microhomology (of at least 1 basepair) between the daughter and template strands. In this specific instance, a run of 5 Gs was added, only the final one of which was used for microhomology. The first 4 Gs then appear as a “filler” sequence in the final mutant DNA and also serve to keep the FLT3 mRNA reading frame open (the total number of additional nucleotides [21] is divisible by 3). (D) Similarly, a single T nucleotide is predicted to be added to the daughter strand during the formation of the common NPM1 mutation, a 4-bp tandem duplication. Borrow et al provide evidence that the extra nucleotides facilitating the formation of approximately one-third of FLT3-ITDs and almost all NPM1c mutations2  follow a pattern that strongly suggests illegitimate TdT activity. Panels B to D are derived from real examples of mutations depicted by Borrow et al. bp, basepair; nt, nucleotide. Professional illustration by Patrick Lane, ScEYEnce Studios.

Both FLT3-ITD and NPM1c mutations are thought to arise during cell division as a result of DNA replication slippage (also known as slipped strand mispairing). This is a molecular error in which the DNA polymerase and the attached nascent daughter DNA strand temporarily dissociate from the template strand and move to an upstream location with microhomology to the 3′ end of the daughter strand. The DNA polymerase then “resumes” replication from this position and, in so doing, reincorporates the same nucleotides into the elongating daughter strand. Although this error can be detected by excision repair proteins, it is not always correctly repaired, leading to a permanent duplication mutation. Replication slippage is more likely to occur at repetitive regions, where repeats can form hairpin loops and a “slipped” polymerase can readily find a complementary sequence on the template strand at which to relocate. This error has been put forward as an explanation for the presence of repetitive DNA sequences in genomes and for the intergenerational expansion of trinucleotide repeats in disorders such as Huntington disease.5 

The presence of repetitive DNA sequences at exons 14 and 15 of FLT3 has been widely thought to promote the occurrence of FLT3-ITD mutations. However, by studying the sequences of 300 different FLT3-ITDs from published datasets, Borrow et al observed that the mutations were not always perfect duplications, but frequently contained additional filler nucleotides, something reported before, but not closely scrutinized. In addition, they noted that, in approximately one-third of cases, the ITD could not have been generated by simple replication slippage, because there were no homologous nucleotides at the “slipped” position from which DNA polymerase would “resume” replication, a phenomenon they refer to as missing microhomology (see figure). The authors then set out to investigate the origin of the filler nucleotides and also to find explanation for the missing microhomology. First, they hypothesized that microhomology of at least 1 basepair was generated by addition of nucleotides to the end of the daughter strand and inferred what the sequence of these nucleotides had to be. Then they examined the sequence composition of (i) filler nucleotides and (ii) nucleotides added to generate microhomology and found that both closely matched the pattern of nucleotides added by TdT (in terms of GC content, length, and incidence of homodimers). In their second paper,2  the authors turned their attention to NPM1 exon 12 mutations and studied 2430 individual cases. They again observed that the most common form of NPM1 mutation (type A = TCTG duplication) represents a case of replication slippage with missing microhomology (see figure). By examining other less common types of NPM1 mutation, the authors discovered that almost all of these were consistent with instances of missing microhomology. As with FLT3-ITD, the patterns of added nucleotides closely matched those expected by the action of TdT.

Although the evidence for the action of TdT is expectedly indirect, TdT is the most template-independent of the Pol X family of DNA polymerases6  and is expressed by a significant proportion of AMLs, whereas other AMLs have rearranged immunoglobulin or TCR loci (indicating TdT activity in an ancestral cell).7  Furthermore, the authors allude to unpublished evidence that one of the few other types of recurrent ITDs in cancer, namely BCOR-ITDs in solid cancers (where TdT is not active), does not harbor additional nucleotides compatible with TdT activity.2  Also, in mice, where TdT activity is lower than in humans,8 Flt3-ITDs do not occur spontaneously in Npm1-mutant animals.9  In this light, the 2 papers make a strong case that, through its role in causing NPM1c and FLT3-ITD mutations, TdT is involved in the development of almost half of all cases of human AML, a remarkable finding. AML genomes harbor only small numbers of somatic mutations, mostly nucleotide substitutions, that arise stochastically through rare mistakes in genome maintenance and accumulate slowly with age.10 NPM1c and FLT3-ITD mutations never fitted this pattern, and the work by Borrow et al provides a plausible explanation for their causation.

Incriminating TdT in the causation of the 2 most common mutations in AML may have important mechanistic and by extension clinical implications. For example, mutations in either NPM1 or FLT3 have not been described in clonal hematopoiesis (CH) and are usually acquired in cells harboring preexisting mutations in genes such as DNMT3A, leading to AML. It has been hypothesized that such CH mutations “synergize” with NPM1 to drive leukemogenesis. Could they instead, or in addition, be involved in epigenetic activation of TdT activity in stem/progenitor cells? Alternatively, could TdT activation be transient and related to environmental exposures or interim illnesses, as has been proposed for pediatric ALL?4,11  Furthermore, could the shared causation by TdT underlie the preferred cooccurrence of NPM1c with FLT3-ITD (compared to other synergistic mutations)?

Finally, beyond its scientific impact, the authors’ investigation stands out as an elegant demonstration of how a subtle observation can give unexpected insights with far-reaching implications. In this manner, their scrutiny of the previously inconspicuous nucleotides and their conclusions about their significance are reminiscent of the fabled adventure alluded to in this piece’s title: the fact that the dog did not bark in the nighttime was easy to ignore, but to Sherlock Holmes it was the key evidence leading him to the identity of the intruder.12 

Conflict-of-interest disclosure: G.S.V. is a consultant to Kymab and Oxstem.

REFERENCES

REFERENCES
1.
Borrow
J
,
Dyer
SA
,
Akiki
S
,
Griffiths
MJ
.
Terminal deoxynucleotidyl transferase promotes acute myeloid leukemia by priming FLT3-ITD replication slippage
.
Blood
.
2019
;
2281
-
2290
.
2.
Borrow
J
,
Dyer
SA
,
Akiki
S
,
Griffiths
MJ
.
Molecular roulette: nucleophosmin mutations in AML are orchestrated through N-nucleotide addition by TdT
.
Blood
.
2019
;
2291
-
2303
.
3.
Papaemmanuil
E
,
Gerstung
M
,
Bullinger
L
, et al
.
Genomic classification and prognosis in acute myeloid leukemia
.
N Engl J Med
.
2016
;
374
(
23
):
2209
-
2221
.
4.
Greaves
M
.
A causal mechanism for childhood acute lymphoblastic leukaemia [published correction appears in Nat Rev Cancer. 2018;18(8):426]
.
Nat Rev Cancer
.
2018
;
18
(
8
):
471
-
484
.
5.
McMurray
CT
.
Mechanisms of trinucleotide repeat instability during human development [published correction appears in Nat Rev Genet. 2010;11(12):886]
.
Nat Rev Genet
.
2010
;
11
(
11
):
786
-
799
.
6.
Nick McElhinny
SA
,
Havener
JM
,
Garcia-Diaz
M
, et al
.
A gradient of template dependence defines distinct biological roles for family X polymerases in nonhomologous end joining
.
Mol Cell
.
2005
;
19
(
3
):
357
-
366
.
7.
Drexler
HG
,
Sperling
C
,
Ludwig
WD
.
Terminal deoxynucleotidyl transferase (TdT) expression in acute myeloid leukemia
.
Leukemia
.
1993
;
7
(
8
):
1142
-
1150
.
8.
Bentolila
LA
,
Fanton d’Andon
M
,
Nguyen
QT
,
Martinez
O
,
Rougeon
F
,
Doyen
N
.
The two isoforms of mouse terminal deoxynucleotidyl transferase differ in both the ability to add N regions and subcellular localization
.
EMBO J
.
1995
;
14
(
17
):
4221
-
4229
.
9.
Dovey
OM
,
Cooper
JL
,
Mupo
A
, et al
.
Molecular synergy underlies the co-occurrence patterns and phenotype of NPM1-mutant acute myeloid leukemia
.
Blood
.
2017
;
130
(
17
):
1911
-
1922
.
10.
Welch
JS
,
Ley
TJ
,
Link
DC
, et al
.
The origin and evolution of mutations in acute myeloid leukemia
.
Cell
.
2012
;
150
(
2
):
264
-
278
.
11.
Papaemmanuil
E
,
Rapado
I
,
Li
Y
, et al
.
RAG-mediated recombination is the predominant driver of oncogenic rearrangement in ETV6-RUNX1 acute lymphoblastic leukemia
.
Nat Genet
.
2014
;
46
(
2
):
116
-
125
.
12.
Doyle
AC
.
Adventure 1: Silver Blaze. The Memoirs of Sherlock Holmes
.
London, United Kingdom
:
George Newes
;
1894
.