Neutropenia-associated ELANE mutations disrupting translation initiation produce novel neutrophil elastase isoforms

There are two main types of hereditary neutropenia: cyclic neutropenia and severe congenital neutropenia (SCN). In cyclic neutropenia, neutrophil counts oscillate with 21-day periodicity.¹  In SCN, neutrophil counts are statically low, promyelocytic maturation arrest occurs in the bone marrow, and disease often progresses to myelodysplasia or acute myeloid leukemia.¹  Heterozygous mutation of ELANE causes almost all cases of cyclic neutropenia²  and the majority of SCN.³  Because neutropenia is often lethal, germline mutations frequently arise de novo.¹  Additional genes causing SCN include HAX1,⁴ GFI1,⁵ G6PC3,⁶  and others.⁷ 

ELANE encodes the neutrophil granule serine protease, neutrophil elastase (NE).⁸  While it is unclear how ELANE mutations cause neutropenia, nearly all of its myriad mutations are either amino acid missense substitutions, small insertions or deletions preserving translational reading frame, or carboxyl-terminal chain-terminating mutations escaping nonsense-mediated decay.^1,9  The mutational spectrum would seem to exclude haploinsufficiency as a mechanism, because mutations predicting an absence of protein have not yet been reported.

Mutations distribute throughout NE, and effects on biochemical properties such as proteolytic activity, serpin inhibition, and glycosylation appear inconsistent.^9-11  Two theories on how mutations affects NE have been proposed. In one, mutant NE is mistrafficked, while, in the other, mutant NE misfolds, activating an unfolded protein response (UPR) in the endoplasmic reticulum (ER).

Relevant to the mistrafficking hypothesis, NE is stored in lysosome-like granules, but distributes to the plasma membrane and nucleus.⁸  Some ELANE mutations are reported to disturb NE trafficking, both in vitro^12,13  and in vivo¹⁰  (though other studies have not found mislocalization).⁹  Furthermore, mutations in the gene encoding the lysosomal transporter protein AP3B1, which is involved in trafficking NE,¹²  are responsible for the neutropenic disorders Hermansky-Pudlak syndrome type 2¹⁴  and canine cyclic neutropenia,¹²  and, at least in dogs, NE appears to be mislocalized.¹⁵  Chédiak-Higashi syndrome, caused by mutations in a different lysosomal trafficking protein, may also cause neutropenia,¹⁶  and in a mouse model of the disorder, NE appears to be mistrafficked.¹⁷  Finally, mutations in other genes involved in lysosomal trafficking, including VPS13B¹⁸  and VPS45,^19,20  also cause neutropenia.

With the misfolding hypothesis, when certain ELANE mutations are expressed in vitro, UPR markers including binding immunoglobulin protein, XBP1, and GRP78 are upregulated.^10,21  A supportive observation involves Wolcott-Rallison syndrome, which, in addition to other features, includes neutropenia and is caused by mutations in protein kinase RNA-like ER kinase (PERK) which functions as a sensor of ER stress.²²  Gene-targeted mice carrying a neutropenia-associated ELANE mutation develop neutropenia when ER degradation is blocked with the proteasome inhibitor bortezomib, resulting in high levels of ER stress.²³ 

Here, we describe a new category of ELANE mutation disrupting the translation initiation codon or the immediately adjacent Kozak sequence that does not easily fit with either the mislocalization or misfolding hypotheses. Because they might not produce a protein, these mutations would seem contrary to the concept that a mutant polypeptide causes disease. The aim of the present study is to determine the molecular effects of mutations involving ELANE’s initiation codon.

Sanger DNA sequencing of ELANE from polymerase chain reaction (PCR)-amplified peripheral blood genomic DNA was performed as described.²  Research was approved by the University of Washington Institutional Review Board, and participants gave written informed consent in accordance with the Declaration of Helsinki.

Peripheral blood mononuclear cells were transduced using lentiviral vectors containing Oct4, Sox2, Klf4, and c-Myc. See details in supplemental Materials and methods section on the Blood Web site.

Human leukemic monocyte lymphoma (U937) and human promyelocytic leukemia (HL-60) cell lines were purchased from ATCC and cultured in RPMI 1610 (Life Technologies, Grand Island, NY). Henrietta Lacks (HeLa) and rat basophil leukemia 1 (RBL-1) cells were cultured in Dulbecco’s modified eagle medium (Life Technologies). Media was supplemented with 10% fetal bovine serum containing 100 units/mL penicillin and 100 μg/mL streptomycin.

ELANE segments were PCR-amplified from genomic DNA and inserted into a bicistronic (pRF) vector (gift from Dr. Vincent Mauro, The Scripps Research Institute, La Jolla, CA), containing firefly and Renilla luciferase. A construct containing the triplicated active region was made by separating 3 tandem repeats with a 9-bp oligonucleotide derived from a non–internal ribosome entry site (IRES) region from the mouse β-globin gene²⁴  (supplemental Table 1). Vectors were transfected into HeLa cells using Fugene HD (Promega, Madison,WI). Cells were harvested after 24 hours. Luciferase assays were performed using the Dual-Luciferase Reporter Assay System (Promega) on a Synergy 4 microplate reader (BioTek, Winooski, VT). Firefly activity was normalized to Renilla activity.

Chicken IgY raised against the NE carboxyl-terminus was used for western blots, as described.²⁵ 

Wild-type and mutant ELANE vectors were generated by inserting ELANE into pIRES2-ZsGreen (Clonetech, Mountain View, CA) obtained from previously described vectors.¹¹  U937 cells were transfected with the Amaxa Nucleofector Kit C (Lonza, Basel, Switzerland), 2 µg of DNA, and program W-01. After 24 hours posttransfection, single cells were sorted into 96 well plates using a FACSAria II (BD Biosciences, San Jose, CA), gated on ZsGreen fluorescent protein expression. Two weeks later, plates were scored for wells containing ≥ 20 cells.

HL-60 cells were transfected with the same vectors as the clonogenic capacity assay, with 1.25% dimethylsulfoxide.²⁶  ZsGreen-positive cells were sorted with a FACSAria II, and RNA was isolated using RNeasy Plus Mini Kits (Qiagen, Germantown, MD). RNA was reverse-transcribed with SuperScript II reverse transcriptase (Life Technologies). Heat-shock 70-kDa protein 5 (HSPA5) (Hs00607129_gH) expression, normalized to ACTB (4352935E) was quantified using TaqMan (Life Technologies) assays on an Applied Biosystems 7300 real-time PCR instrument. ER stress was induced in control cells with 2 μg/mL tunicamycin for 16 hours. For apoptosis assays, transfected HL-60 cells were incubated for 24 hours, and then serum-starved for 16 hours. Apoptosis was assessed with the Annexin V-PE Apoptosis Detection Kit 1 (BD Pharmingen) per manufacturer instructions on a FACSCanto II, gating ZsGreen-positive cells.

Proteolytic activity was assessed as described,¹¹  following 16-hour incubation.

Comparisons between groups employed Student two-tailed t test.

An SCN-associated ELANE mutation, with A>G substitution at the first position of the ATG methionine translation initiation codon (Table 1), was reported previously by the French Neutropenia Register.²⁷  Neither parent was affected or possessed the mutation, indicating it arose de novo. As an isolated case, it remained uncertain whether the mutation was causative. Since then, the proband has had an affected child inheriting the mutation, and we have observed another de novo occurrence in an unrelated patient in the French Register. The same mutation, again de novo, has subsequently been described in two additional, unrelated patients,^9,28  including one unresponsive to recombinant human granulocyte colony stimulating factor therapy.²⁸  Recently, additional SCN patients with mutations of the second (T>G) and third (G>C) positions of the ATG methionine initiation codon have been described,⁹  and, in the case of the former, was also seen in the French Register (Table 1).

We report here for the first time, a different mutation (G>A) at the third position of the ATG initiation codon arising de novo in an SCN patient, whose affected child also inherited the mutation (Table 1 and supplemental Figure 1A). Also for the first time, we describe a cyclic neutropenia patient with mutation (A>T) of the conserved −3 position of the Kozak translation initiation sequence, where a purine, as found in wild-type ELANE, is strongly preferred²⁹  (Table 1 and supplemental Figure 1B).

In summary, there were 9 unrelated probands with neutropenia comprising 5 different translation start mutations (Table 1). No other ELANE-coding sequence variants were observed among these patients. These variants have not been described in 6503 individuals in the National Heart, Lung, and Blood Institute’s Exome Sequencing Project (http://evs.gs.washington.edu), cataloged in 1000 Genomes,³⁰  dbSNP³¹  (version 135), or among hundreds of controls.⁹  Since these variants were exclusively found in patients with neutropenia who lacked other causative mutations, and often arose de novo, but can be autosomally dominantly transmitted, we conclude that they are causal. Among a total of 191 different known ELANE mutations,⁹  all, except for these, were predicted to result in translation of correspondingly mutated NE protein. A closer look at this class of mutations is warranted.

We hypothesized that instead of preventing translation of NE, mutations of ELANE’s translational start site may, instead, lead to translation commencing from internal methionine ATG codons. There are three downstream, in-frame ATG codons that could potentially be used as start sites when the canonical initiation codon is mutated (Figure 1A). (There are also two out-of-frame initiation codons. One is followed immediately afterward by a stop codon, and the other is terminated after 9 codons.)

To test this possibility, we transfected ELANE complementary DNA (cDNA) containing mutations involving the initiation codon and the Kozak sequence in RBL-1 cells (which are conventionally employed for the study of NE because although they lack endogenous NE, they correctly target it to granules),³²  and performed a western blot with an antibody detecting a carboxyl-terminal epitope. When the initiation codon was mutated from ATG to either ATA or GTG, shortened forms of NE were evident (Figure 2, c.3G>A, c.1A>G). When the highly conserved nucleotide in the Kozak sequence 3 nucleotides upstream from the initiation codon was mutated from A to T, shorter proteins were produced along with the wild-type protein (Figure 2, c.-3A>T). To verify that shortened forms of NE represent translation initiation from alternate sites and are not due to posttranslational processing, expression vectors were generated containing just the open reading frames (ORFs) representing all in-frame methionine codons (wild-type compared with ATG2, ATG3, and ATG4). Correspondingly shorter isoforms were also produced when the region upstream from each internal in-frame ATG was deleted, with the exception that there was no detectable expression from ATG4 ORF (Figure 2).

We tested for the presence of shorter NE isoforms in patient-derived samples. Bone marrow was unavailable in patients with start-site mutations. Western blot performed on peripheral blood could not be interpreted due to neutropenia (not shown). We, therefore, generated iPSC from a patient with the c.1A>G mutation (supplemental Figure 2), which showed normal karyotype, pluripotent stem-cell marker expression, and the ability to differentiate into all three germ layers. Compared with control, upon hematopoietic differentiation, there was no significant difference in the level of CD34⁺ cells, but subsequent myeloid differentiation demonstrated the following: fewer bands and neutrophils; increased promyelocytes, myelocytes, and metamyelocytes; more monocytes; and increased apoptosis of CD34⁻/CD33⁺ cells (supplemental Figure 3A-D). (The extent of apoptosis, however, may fall short of explaining cytopenias and could be an epiphenomenon.) Western blot of CD34⁻/CD13⁺/CD11b^low/CD15⁺⁺ myeloid cells derived from iPSC (supplemental Figure 3E) confirmed reduced abundance of full-length NE (as expected with heterozygosity for the mutation, where wild-type is also present) and uniquely increased abundance of at least one shorter isoform. Additional higher molecular weight products, as seen in RBL-1 cells, were detected. Isoform sizes in iPSC appeared to be slightly shifted upward, compared with RBL-1 cells, suggesting differences in posttranslational modification. Experiments with iPSC, therefore, corroborate phenotype and downstream initiation of translation.

Figure 2 shows that translation normally initiating from the canonical ATG1 start site in wild-type ELANE does not lead to translation of polypeptides from internal ORFs. However, the ATG2 ORF expression vector yields a range of shorter polypeptides corresponding to translation initiation from ATG2, ATG3, and ATG4. We reasoned that there could be an IRES situated between ATG2 and ATG3. IRES sequences are regions of messenger RNA (mRNA) that recruit the ribosome to sites adjacent to internal ATG codons and permit internal initiation of translation in a cap-independent manner.³³ 

To confirm the presence of an IRES and exclude the possibility that the cap site is still used for ribosome entry when the canonical initiation codon is mutated, we introduced a termination codon (c.83C>A = p.S-2X) between ATG1 and the first internal methionine codon, ATG2, in cis with the Kozak sequence mutation (c.-3A>T). The shorter isoforms were present in greater abundance, while polypeptides initiating from ATG1 terminated prior to the carboxyl terminus and were consequently no longer detected by antibody to the carboxyl-terminus (supplemental Figure 4).

One IRES, found in mouse Nkx6-2, exhibits reverse complementarity to 18S ribosomal RNA (rRNA), and mutations disturbing complementarity between Nkx6-2 mRNA and 18S rRNA disrupt IRES activity.³⁴  We scrutinized ELANE mRNA and confirmed it contains a region of reverse complementarity to 18S rRNA just downstream from ATG2, spanning the region of complementarity to 18S rRNA defined in Nkx6-2 (Figure 1B). In computer simulations where random sequences the same length as the putative IRES were generated from ELANE mRNA and compared with 18S rRNA, sequence complementarity was determined to be highly nonrandom (P = 1.3 × 10⁶) (supplemental Figure 5).

We, therefore, evaluated whether the region of 18S rRNA complementarity possesses IRES activity. A test vector containing distinguishable tandem luciferase (Renilla and firefly) reporters, was used. The potential IRES was inserted in-between the two luciferase genes. The upstream Renilla luciferase contains a Kozak consensus sequence needed for translation initiation; however, translation of the downstream firefly luciferase requires that the inserted sequences possess IRES activity. Several overlapping segments from within the region of ELANE exhibiting complementarity to 18S rRNA for IRES activity in transfected HeLa cells were tested. Region 1 showed a 2.5-fold increase in IRES activity, but regions 2 and 3 did not exhibit IRES activity compared with vector alone (Figure 3A), possibly due to the presence of additional start codons within those regions. Additionally, a 5′-UTR of similar length from an arbitrary gene (KLHDC8B) lacked IRES activity (not shown). To further verify that region 1 contained an IRES, we evaluated its activity when it was triplicated in a head-to-tail sequence. IRES activity was increased by ∼sixfold, compared with the vector alone (Figure 3B), suggesting that IRES activity is not adventitious.

It should be noted that ELANE mutations frequently locate to the putative IRES.^1,9  We tested 3 of these mutations (p.V16M, p.S17F, and p.L18P) for effect upon IRES activity, using the dual luciferase reporter containing triplicated ELANE IRES (Figure 3B). Two of the mutations (p.V16M and p.S17F) markedly increased activity, compared with the empty vector. When transfected into RBL-1 cells (supplemental Figure 4), p.V16M produced one of the shorter isoforms also seen with the c.-3A>T mutation. Results for S17F were inconclusive because there was limited expression of even the wild-type isoform, which would also be consistent with the possibility that S17F disrupts translation from the canonical start site. Further study is needed to determine the nature of polypeptides produced by these mutations, optimal sequence for IRES function, and how mutations within ELANE mRNA may modulate IRES activity.

To determine if the expression of NE polypeptides initiating from internal translational start sites are harmful to cells, we expressed cDNAs containing internal ORFs in U937 promonocytes, performing a previously described “clonogenic capacity” assay, developed to study neutropenia-associated ELANE mutation induction of the UPR.²¹  Three different vectors were used in order to isolate the different ORFs and determine their effects; in addition to expressing the ATG2 ORF and the ATG3 ORF, we also evaluated the ATG>GTG mutation, which produces all 3 internal ORFs. (We did not test ORF4 individually, because its production seems to require upstream sequences, and, in isolation, does not produce detectable protein.) All of the vectors expressing shortened forms of NE reduced clonogenic capacity (Figure 4). When comparing the shorter isoforms of NE to each other, the GTG mutation, which leads to production of all 3 internal ORFs, had a greater effect on inhibiting clonogenic capacity than did the ATG2 ORF by itself.

To determine if amino-terminally truncated forms of NE retain enzymatic activity, we tested their ability to cleave the NE-specific substrate MeO-Suc-Ala-Ala-Pro-Val-pNA following expression in transfected RBL-1 cells. We found that the c.1A>G mutation retained minimal, yet statistically significant residual activity, but that isolated ORFs corresponding to the second and third ATGs lacked activity (Figure 5A and supplemental Figure 6).

We assessed whether amino-terminally truncated forms of NE induce ER stress by measuring expression of HSPA5, which codes for the binding immunoglobulin protein. We found no increase in HSPA5 in HL-60 cells transfected with ELANE mutations altering the translational start site (Figure 5B), whereas control cells transfected with wild-type ELANE and stimulated with tunicamycin, which induces ER stress,²¹  showed a 12-fold increase in HSPA5 expression (supplemental Figure 7).

We also tested whether amino-terminally truncated forms of NE induce apoptosis in HL-60 cells transfected with vectors expressing either wild-type or amino-terminally truncated forms of NE. There was no increase in apoptosis as observed by Annexin V binding (Figure 5C and supplemental Figure 8). Additionally, reverse transcription (transcriptase)-PCR assays revealed no increase in expression across a panel of apoptosis-related genes (supplemental Figure 9). The fact that apoptosis was evident in iPSC exposes the limitations of HL-60 cells.

To determine if amino-terminally truncated forms of NE subcellularly mislocalize, we transfected RBL-1 cells with the c.1A>G and c.3G>A mutations. Compared with granular cytoplasmic distribution of the wild-type, the shorter polypeptides appeared to localize to nuclei (supplemental Figure 10).

Somatic and germline mutations of translational start sites are reported for several human disease-associated genes, where it is speculated that, instead of abrogating translation, they may force utilization of downstream initiation codons.^35,36  However, the demonstration of this phenomenon has been rare. One example involved acquired mutations of the hematopoietic transcription factor GATA1 associated with transient myeloproliferative disorder and megakaryoblastic leukemia of Down syndrome.³⁷  Start codon mutations or, more often, proximal chain-terminating mutations just downstream from the initiation codon, lead to the production of a shorter GATA1 variant commencing from a second initiation codon, which lacks the activation domain yet retains the ability to bind DNA and cofactors.^38,39  Similar mutations producing similar effects occur in the transcription factor C/EBPα in acute myeloid leukemia.⁴⁰  Another example involved germline mutation of TPO, encoding thrombopoietin, in hereditary thrombocytopenia, and disrupting a splice-site, which leads to deletion of the canonical start site with initiation shifting downstream to the next in-frame ATG codon.⁴¹  Here, we show that neutropenia-associated germline mutations disrupt the initial ATG codon, and adjacent noncoding Kozak translation initiation sequence of ELANE lead to production of amino-terminally truncated isoforms of NE-initiating translation from downstream, in-frame methionine ATG codons.

Potential alternate translation start sites are predicted for as many as 12% of human genes.⁴²  Furthermore, potential alternate start sites are evolutionarily conserved and are used in many cases.^43-45 

One of the hypothesized roles of alternate translation start sites is to alter amino terminal localization signals, and ultimately, protein location; 30% of proteins derived from potential alternate start sites are predicted to have different subcellular localization.⁴²  One example is neuropeptide Y, which has two translation start sites. The protein produced from the first initiation codon is targeted to secretory vesicles whereas the protein initiating downstream locates to mitochondria.⁴⁶  We have shown that, at least in vitro, pathogenic isoforms of NE resulting from start-site mutations similarly become mislocalized when truncated at the amino terminus, probably because the shorter isoforms lack the ER-localizing signal sequence.

We demonstrated IRES activity for a region of ELANE between the second and third internal ATG codons. Based on prior studies of Nkx6-2, we suspected this activity was due to direct ribosomal binding mediated by ELANE mRNA sequence complementarity to 18S rRNA.⁴⁷  This IRES permits the production of both the ATG2 and ATG3 ORF NE polypeptides from a single mRNA when the initial ATG codon is mutated. It is less clear if this IRES normally initiates translation when the initial ATG codon or flanking Kozak sequence is not altered. Another interesting observation was that mutations of this region enhanced activity despite reducing complementarity to 18S rRNA compared with wild-type. In previous studies, it was found that the optimum length of complementarity to 18S rRNA was 7 nucleotides, and any sequence longer or shorter than that, decreased IRES activity.⁴⁸ 

It is likely that other coding sequences within ELANE mRNA contribute to translational regulation directly either by influencing binding to the ribosome or other translation factors, or indirectly by affecting mRNA folding. If so, some coding sequence ELANE mutations may favor production of internally translated isoforms, in which case, protein sequence changes could prove incidental and noncontributory to pathogenesis. Conversely, ELANE variants involving synonymous codon substitutions that do not alter coding sequence may nevertheless prove pathogenic, because they disturb the translational start site, IRES, or other mRNA sequences regulating translation from internal start sites in ELANE. In our own genetic testing of neutropenic patients, we have detected novel synonymous ELANE variants, but have not previously designated them as deleterious.¹ 

ELANE mutations inactivating the ATG initiation codon are found in SCN patients and lead to the absence of full-length protein with expression of amino-terminally truncated isoforms not seen with the wild-type allele. However, the Kozak sequence mutation occurs in patients with cyclic neutropenia and results in the production of both full-length protein, and compared with SCN patients, reduced levels of the amino-terminally truncated isoforms, suggesting that less expression of the shortened isoforms correlates with what might be considered, milder disease.

The shortened isoforms reduced clonogenic capacity in an assay of myeloid proliferation. Cells transfected with the shorter forms of NE did not exhibit increases in ER stress when compared with wild-type. It is not surprising that the mutations did not induce ER stress since they were predicted to lack the signal sequence necessary to translocate to the ER. However, the presence of a “smear” of high molecular weight products evident in western blots in Figure 2 and supplemental Figure 3E, suggested that the shortened forms of NE may still misfold and aggregate, albeit not in the ER. It is possible that aggregated NE may sequester transcription factors. ORFs 2 and 3 are additionally predicted to lack the pro-zymogen activation sequence whose removal is required for catalytic function; yet they retained the catalytic triad of histidine, aspartate, and serine necessary for proteolytic activity, raising the possibility that they were mature and possessed enzymatic activity immediately following translational synthesis. We found that mutation disrupting the canonical translation start site retained only minimal proteolytic activity; yet it is possible that even residual activity could be damaging if not properly compartmentalized.

In evaluating the relevance of this new category of mutations disrupting the translational start site, it is unlikely that all of the numerous ELANE mutations would disrupt translational regulation of mRNA and lead to production of internally translated ORFs. Nevertheless, we speculate that there may be a mechanism through which conventional mutations alter the primary sequence of NE, which may lead to the production of internally-translated isoforms via protein misfolding.

The presence of shorter isoforms of ELANE generated by an IRES, actually fits with the two current models on how mutations may cause disease (Figure 6). If the misfolding hypothesis holds true, the shorter isoforms could be produced by the IRES through a trans mechanism. In times of stress, protein misfolding within the ER activates PERK, which phosphorylates eukaryotic translation initiation factor 2,⁴⁹  resulting in a preference for IRES-mediated translation over normal cap-dependent translation.⁵⁰  Our observations also fit well with the mislocalization hypothesis. Since the shortened isoforms do not contain a propeptide signal sequence, they are unlikely to correctly transit through the ER, where they would ordinarily become glycosylated. Previously, we demonstrated that when sites of asparagine-linked glycosylation in NE are mutated, NE accumulates in the nucleus.²⁵  In fact, the shorter isoforms appear to localize to the nucleus in RBL-1 cells, congruent with the theory that they do not enter the ER and are not glycosylated. Aberrant localization may contribute to pathogenesis, but further research is needed to establish if this is also true in human neutrophils and what link it may have to disease.

This work was supported by research grants from the National Institutes of Health (R01DK58161) and the US Department of Defense (BM120152).

Contribution: T.T., J.W., and M.S.H. conceived and designed the study; H.L.G., C.L., and J.A.C designed iPSC studies; T.T., J.W., S.J.C., H.L.G., C.L., J.A.C., and M.S.H. analyzed and interpreted the data; T.T. and M.S.H. wrote the paper; J.W. performed isoform expression and western blot experiments (except for iPSC); R.C.N. and L.T. performed iPSC studies; S.J.S. performed immunolocalization studies; T.T. performed all other studies with the assistance of J.W.; and S.J.C., T.G., J.C.C., and J.D. ascertained and clinically evaluated patients described here.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Marshall S. Horwitz, Department of Pathology, University of Washington School of Medicine, Box 358056, Seattle, WA 98195; e-mail: horwitz@uw.edu.