mRNA—From COVID-19 Treatment to Cancer Immunotherapy

This review provides an overview covering mRNA from its use in the COVID-19 pandemic to cancer immunotherapy, starting from the selection of appropriate antigens, tumor-associated and tumor-specific antigens, neoantigens, the basics of optimizing the mRNA molecule in terms of stability, efficacy, and tolerability, choosing the best formulation and the optimal route of administration, to summarizing current clinical trials of mRNA vaccines in tumor therapy.


Introduction
mRNA-messenger ribonucleic acid-is a single-strand copy of a selected part of the genetic sequence of a gene, negatively charged with secondary and tertiary structure formations. It is produced from DNA and-after appropriate modification-leads to the synthesis of a specific protein by a ribosome. It was first described in 1961 [1]. The feasibility of in vivo expression following the direct injection of in vitro-transcribed mRNA was demonstrated in 1990 [2]. The first commercially available mRNA-based vaccines were Comirnaty (BNT162b2, tozinameran) and Spikevax (mRNA-1273, elasomeran) [3,4]. The major hurdles that had to be overcome before the successful in vivo use of mRNA as a vaccine were instability of the molecule, innate immunogenicity, and delivery. The general structure of mRNA is illustrated in Figure 1.  Table 1. Building blocks of mRNA and their functions. The Kozak sequence is a group of nucleotides that initiates the protein translation in the ribosome. An example is 5 -GCCGCCRCCAUGG-3 , where red color indicates that these nucleotides are fixed, AUG is the start codon encoding for methionine, R stands for A or G, and the function of GCC is not well defined.

Component
Nucleotides Function mRNA is composed of nucleotide building blocks consisting of a selection of four different nucleobases, adenine, cystine, guanine, and uracil, coupled to ribose which contains a phosphate group as the linker to the next nucleotide. mRNA transfers the genetic information for the synthesis of a specific protein from the genes via the DNA to the protein production facility: the ribosome. The nucleotides constitute the alphabet of the genetic code. Three consecutive nucleotides, for example, AUG, function as codons. Accordingly, by using four letters of the alphabet, A, C, G, and U, 64 different codons are possible. In general, codons define the amino acid which will be recruited next in the protein synthesis. On top of that, a single codon or groups of codons have specific functions, for example, the codon AUG codes for the amino acid methionine and operates as the start codon for the coding part of the mRNA. Table 1 provides an overview of the functionality of codons.
The task mRNA has to fulfil involves the following steps. It starts with the production of a single-strand copy of the genetic code laid down in the DNA, which is performed by RNA polymerase. This process is called transcription and leads to precursor mRNA (pre-mRNA), which-via splicing procedures-finally becomes mature mRNA. In parallel to transcription, the 5 cap is added to the molecule and the poly(A) tail is attached at the end of the chain. After that, the mRNA is transported from the nucleus of the cell to the cytosol. Various mechanisms can be utilized for this step involving different proteins, such as CBP20 and CBP80. Subsequently, ribosomes take over, read the code, and start the protein synthesis according to the codons provided in the coding section of the mRNA. The synthesis is terminated as soon as a stop codon is reached. Stop codons are UAG (amber), UAA (ochre), and UGA (opal). Thereafter, the mRNA is degraded by ribonucleases. mRNA activates the innate immune response through various RNA sensors such as toll-like receptors (TLRs), retinoic acid-inducible gene I (RIG-I), and protein kinase R (PKR) [5][6][7][8].

Lessons Learned from COVID-19
The unprecedented speed of the global spread of the COVID-19 pandemic caused by the coronavirus, SARS-CoV2, resulted in an extremely rapid development of mRNA vaccines [9]. In January 2020, the genetic sequence of SARS-CoV-2 was available; in March 2020, the first Phase I trial (NCT04283461) started with an mRNA vaccine (BNT162b2) and a rolling submission procedure was initiated, in which data were provided to the authorities as soon as they were available. In December 2020, BNT162b2 received a temporary emergency use authorization in the UK followed by a conditional marketing authorization (CMA) by the EMA based on data from a global Phase I/II/III study (NCT04368728).
Although SARS viruses are common in humans, vaccines had not been developed since the course of the infection normally was very mild. The SARS outbreak in early 2000 triggered DNA vaccine development [10] but was stopped very soon as the disease vanished on its own. Together with research on MERS-COV, the antigen target rapidly emerged [11]. The spike protein is the major surface protein for most coronaviruses. The virus enters the host cell by endocytosis after binding to angiotensin-converting enzyme 2 [12], followed by the release of the viral DNA. An antibody developed against the spike protein would then inhibit cell entry and thereby neutralize the virus [13]. Furthermore, it could be shown in rhesus monkeys that SARS-CoV-2 infection protects against re-infection [14], mostly via neutralizing antibodies but not via T cell responses [15]. Intramuscularly injected mRNA vaccines mostly induce IgG-type antibodies and less IgA antibodies [16], which provide disease prevention or attenuation but not sterilizing immunity [9]. The preclinical data necessary for the development of a COVID-19 mRNA vaccine were essentially available from the previous SARS and MERS experiments and so saved a lot of development time. In addition to mRNA vaccines, other functional types have been and are in development for the treatment of COVID-19 including DNA and virus vaccines. The latter range from live attenuated virus vaccines via inactivated virus vaccines to recombinant protein vaccines. These types will not be addressed further in this review. For the mRNA-based treatment of COVID-19, various positions of the mRNA molecule had been modified as listed in Table 2.   [20,[37][38][39][40] mRNA capping determines the stability and maturity of the mRNA molecule [41,42]. A eukaryotic translation initiation factor (eIF4E of the eIF4F complex) couples to the 5 cap and subsequently initiates the translation process [20]. Targets of modification include the 5 cap [43,44], 5 -and 3 -UTRs, the coding region, and the poly(A) tail [25]. Modification of the 5 -and 3 -UTRs improves the translation and increases the half-life of in vitro transcription mRNA. The 5 -UTR can be rather short containing only three nucleotides [45] or up to several hundreds. The coding sequence is responsible for the target protein that is produced in the ribosomes and determines the rate of mRNA translation and the stability of mRNA in the cytosol [46,47]. The 3 -UTR is vital for the recognition by proteins in the cytosol and also determines the mRNA stability [25,34]. Again, most details are proprietary and have not been disclosed. The addition of a suitable-length poly(A) tail at the 3 end of mRNA also plays an important role in its successful translation and stability containing ideally more than 90 nucleotides. In some cases, uridine is replaced by pseudouridine, denoted as Ψ. The poly(A) tail can be added to in vitro transcription (IVT) mRNA either through a template vector or by recombinant poly(A) polymerase after the transcription process has occurred [24,25]. mRNA which encodes the protein of interest, in the case of SARS-CoV-2, the spike protein, can be administered as such or as self-replicating or self-amplifying mRNA (saRNA) [48,49]. Antigen expression is proportional to the number of conventional mRNA transcripts successfully delivered during vaccination [50], which might result in the need for relatively high vaccine doses. saRNA vaccines are derived from alphaviruses such as Sindbis and Semliki Forest viruses [51,52]. The viral genome is divided into two open reading frames (ORFs): the first ORF encodes for the RNA-dependent RNA polymerase (replicase), and the second ORF encodes the antigen (spike protein) [53,54]. Unlike saRNA, ordinary mRNA is small due to its simpler structure and is characterized by only one ORF. According to Bidram [41], there are three types of saRNA available: plasmid-based DNA saRNA, virus-like particle delivery saRNA, and IVT saRNA [41,55]. Figure 2 provides an illustration of the saRNA design. Due to self-replication, considerably lower doses of vaccines are needed. Cancer mRNA vaccines are mostly non-replicating [56,57]. Another alternative is trans-amplifying RNA (taRNA) [58]. In this model, the vector cassette encoding the vaccine antigen originates from the saRNA, from which the replication was deleted to form a trans-replicon. Replicase activity is provided in taRNA by a second molecule, either by a standard saRNA or an optimized non-replicating mRNA (nrRNA), which results in a 10-to 100-fold increase in transcription expression. mRNA vaccines are different from classic viral vaccines by providing, instead of the antigen itself, the genetic information for producing the antigen. Now, it is up to the host to start protein production. This is a huge benefit compared to administering the antigen to the host. No longer are cell lines required with their potential to produce multiple impurities and the consequence to establish tedious purification and quality control procedures. This is the same issue for DNA vaccines. mRNA vaccines have another significant advantage over DNA vaccines. They have one major hurdle less. DNA needs to enter the cell nucleus [59], while for mRNA vaccines, it is sufficient to reach the cytosol of the cell, which itself is already rather cumbersome.
Vaccine mRNA manufacturing is rapid and cheap by using an IVT from a DNA template with T7 RNA polymerase [60,61]. Since mRNA and appropriate delivery systems are self-adjuvant, they result in strong and long-lasting adaptive immune responses. This is not the case for protein or peptide-based vaccines, which need the addition of adjuvants [62]. Another benefit of mRNA vaccines is that mRNA is much less likely integrated into the host DNA genome than is the case for DNA vaccines [63]. The major reason is that for mRNA manufacturing bacterial fermentation with all its sequelae of isolation and purification is not necessary. Following the expression of antigens, the activity of mRNA is short-lived due to its clearance by RNases, thereby lowering the burden to the host homeostasis.
mRNA is a large hydrophilic, negatively charged molecule with secondary and tertiary structure formations. The cell membranes are negatively charged as well. Additionally, ion pumps and ion channels maintain a negative potential (−40 to −80 mV) across the cell membrane, keeping the cytosol negatively charged by controlling the balance of most of the essential metal ions (for example, K + , Na + , Ca 2+ , and Mg 2+ ). Naked mRNA is therefore not able to pass this barrier [44]. It was, however, hypothesized that uptake of naked mRNA might occur by endocytosis using cells as mediators. For dendritic cells, this pathway has been described by various groups [64][65][66].
Although the mRNA molecule itself is chemically very stable in the dry state, it is rather unstable in the solution, and, moreover, is rapidly degraded by extracellular and intracellular exo-and endo-ribonucleases [23, 67,68]. As a consequence, and in order to mask the negative charge, mRNA needs packaging to be able to pass the cell membrane barrier and to protect itself inside the cell cytosol against degradation before it reaches its target, the ribosome. On the other hand, the package needs to be such that the ribosome is still able to detect and then to process the mRNA [69].
Most mRNA vaccines use lipid nanoparticles (LNPs) as the carrier with a particle size between 1 and 100 nm. LNP formulations are composed of an ionizable or cationic lipid, a helper phospholipid, cholesterol or a cholesterol derivative, and a polyethylene glycol (PEG)-modified lipid. The purpose of the ionizable or cationic lipid is an interaction with the negatively charged mRNA. Examples are DLin-MC3-DMA (MC3) [70] or DOPE [71]. The helper phospholipid, for example, 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine or DSPC, stabilizes the bilayer structure of the LNP. The pegylated lipid contributes to the stability of the LNP and prevents opsonization followed by uptake in the liver [3,71,72]. However, the PEG group in the LNP formulation is considered to be a possible allergen for anaphylaxis due to approximately 72% of people having some antibodies against PEGs [73]. LNPs are taken up by apolipoprotein or albumin receptor-mediated endocytosis and can electrostatically attach and fuse with the cell membrane using inverted non-bilayer lipid phases [74] or by non-specific pinocytosis [75]. LNPs modified with mannose target DCs through the mannose receptor CD206 [76]. Once inside the cell, LNPs are routed into early endosomes, followed by late endosomes, and finally the lysosomes where the mRNA contents are enzymatically degraded [77,78].
The approved SARS-CoV-2 mRNA vaccines use intramuscular (i.m.) administration. During their development, alternative injection routes have been evaluated, for example, intradermal application. A general consensus on what is the best route has not yet been achieved [3]. Although the i.m. administration route induces strong IgG responses that are thought to protect the lower respiratory tract, unlike natural infection it does not drive the secretory IgA responses that are thought to protect the upper respiratory tract [9]. Most vaccines will protect only against infection of the lower respiratory tract and might not induce sterilizing immunity in the upper respiratory tract, which might still enable transmission of the virus. Live attenuated vaccines or viral vectors that can be applied intranasally would probably also lead to a strong mucosal immune response as well as an IgG response. Currently, 68 clinical studies on intranasal administration are listed at www.clintrials.gov. An overview can be found at [87].
Another important issue of mRNA vaccines, which had to be addressed carefully, is manufacturing. A major lesson from the COVID-19 pandemic is that synthesis of the active ingredient is simple because it is cell-free, scalable, and cost effective. Large-scale production of mRNA vaccines consists of a 1-or 2-step in vitro reaction followed by a purification platform with multiple steps that can include DNase digestion, precipitation, chromatography, or tangential flow filtration [88]. A facility dedicated to mRNA production is able to rapidly manufacture vaccines against multiple targets, with minimal adaptation to processes and formulation [61]. Expression may be possible for complex proteins including monoclonal antibodies that are difficult or impossible to generate with conventional expression systems [89]. The manufacturing process comprises the following steps: generation of a plasmid DNA with an RNA polymerase promoter, e.g., T7, and the mRNA sequence of interest, followed by DNA linearization and transcription of the RNA template and by degradation of the DNA. The 5 cap and the poly(A) tail can be added either during or after transcription. Purification of the mRNA is an important aspect that needs careful attention. For example, various pollutants such as dsRNA in mRNA may activate pattern recognition receptors. HPLC is a suitable method to remove impurities [90,91]. Alternative methods have been described by Shivalingam et al. [92]. Whereas current methods use enzymatic ligation for nucleic acid assembly, their approach is based on the formation of urea and squaramide artificial backbones from minimally modified, commercially available 3 -and 5 -amino oligonucleotides, which provide a one-pot linkage that can be modified on demand for use with stable pre-activated precursor oligonucleotides under reagent-free, mild conditions. Table 3. Active ingredients and formulations of mRNA and viral-based COVID-19 vaccines approved in the European Union. Data have been retrieved from the European Public Assessment Report (EPAR). The structural modifications provided in Table 3 do not reflect all changes that have been introduced into the mRNA molecules; major portions are proprietary and have not been disclosed publicly.

Product
Active A comparative evaluation of the efficacy in clinical trials of a total of 19 COVID-19 vaccines including mRNA and viral versions has been performed by Fiolet [93]. High efficacy was observed for all the vaccines against SARS-CoV-2. BNT162b2, mRNA-1273, and Sputnik V were superior (>90%) compared to the other vaccines. AZD1222, and CoronaVac were effective in preventing symptomatic COVID-19 and severe infections against the Alpha, Beta, Gamma, or Delta variants. Real-life data revealed effectiveness against the Alpha and Beta variants and reduced efficacy against Delta. A decline was observed for BNT162b2 and AZD1222 after six months, indicating a need for booster vaccinations. In another study on the efficacy of different COVID-19 vaccines, Krammer [9] described the following ranking of neutralizing antibodies elicited by the vaccine candidates: inactivated and AdV5 vaccine candidates < ChAdOx1 nCoV-19 ≈ mRNA vaccines < recombinant protein vaccine candidate [9]. Tolerability was excellent. The inflammatory activity of mRNA vaccines can result in local and systemic inflammation and more autoimmune responses. Serious adverse event rates were rare including myocarditis and pericarditis, cytokine release syndrome, and cerebral venous thrombosis [94,95]. Anaphylaxis was found in 2.5-4.7 cases per million doses, and myocarditis in 3.5 cases per million doses [63].
The lessons to learn from the COVID-19 pandemic regarding mRNA vaccines are the following. The pandemic was the strongest driver possible for speeding up the development, production, and distribution of vaccines by providing more than necessary funding options and an immense public pressure to find solutions. Furthermore, mRNA is the ideal candidate for a vaccine since the target antigen is specific for the virus and-so far-has been rather stable regarding evading mutations. If the antigen should slip away, the development of new vaccines is rather straightforward and quick to establish from modifying the mRNA to manufacturing the final product. A still-unresolved issue is the storage temperature, which means freezing, in the case of Comirnaty at −90 • C to −60 • C and for Spikevax at −50 • C to −15 • C during storage and transport, which makes distribution rather cumbersome, in particular in hot areas such as Africa. Freeze-drying of the formulation has been tried as alternative; however, there is a danger of decreasing stability and losing activity [96]. The addition of cryoprotectants, for example, mannitol or sucrose, does preserve the stability and might be a way out of the dilemma [97].

mRNA-Based Cancer Immunotherapy
The objective of cancer immunotherapy is to manipulate the immune system to effectively eliminate cancer cells [98][99][100]. Immunotherapy mainly targets immune cells. It activates the body's immune system by inhibiting negative immune regulatory factors and enhancing the ability of immune cells to recognize tumor cell surface antigens to eliminate tumor cells [101][102][103]. Cellular immune responses are mediated by T cells; in particular, CD8 + T cells can eliminate tumor cells. Humoral immune activity is mediated by antibodies, which induce clearance by phagocytic cells. On the other hand, the suppressive tumor microenvironment characterized by acidity, hypoxia, and an overexpression of enzymes results in a low immunogenicity of tumor cells and subsequent immunosuppression [104] by preventing T cell infiltration into cancers and causing T cell exhaustion [3]. According to Liu [105], immunosuppressive cells in the microenvironment of tumors include myeloidderived suppressor cells (MDSCs) [106], tumor-associated macrophages [107], T regulatory cells [108], pro-tumor N2 neutrophils, and cancer-associated fibroblasts. Remodeling the microenvironment, promoting immune cell infiltration, as well as inhibiting tumor angiogenesis and tumor metastasis are imperative for immunotherapy to be effective [109][110][111].
The first marketed immunotherapies for cancer were recombinant versions of the cytokine interferon-α (IFNα), which were approved by the FDA in 1986 for hairy cell leukemia [112,113]. Other FDA-approved immunotherapy drugs are monoclonal antibodies functioning as checkpoint inhibitors such as ipilimumab (melanoma), pembrolizumab (various tumors), and nivolumab (various tumors); bispecific antibodies such as blinatumomab (ALL), which is directed against CD19 and CD3; cytokines such as Intron A which is recombinant INF α 2a (hairy cell leukemia, melanoma, follicular lymphoma, and Kaposi sarcoma); engineered T cell therapies, such as tisagenlecleucel (ALL and NHL), which are CD19-specific CAR T cells; oncolytic viruses such as talimogene laherparepvec (melanoma), which is genetically modified HSV type 1 designed to replicate within tumors and produce GM-CSF; and last but not least, the cancer vaccine Sipuleucel-T (prostate cancer), obtained from a strain of Mycobacterium tuberculosis [114]. However, the therapeutic effect was far from satisfying [115]. Potential reasons are the low specificity of TAAs, immune escape of cancer cells, and immune suppression in the tumor microenvironment [116]. An overview of therapies is provided in Figure 3. Cancer immunotherapeutics may be divided into four categories: use of recombinant viruses, tumor and immune cell (mainly dendritic cells)-constructed immunizations, and peptide and nucleic acid-based (DNA or RNA) vaccines [41,98] as illustrated in Figure 3. The term "vaccine" is often used instead of immunotherapeutics. Vaccines can be prophylactic or therapeutic. In the treatment of cancer, both prophylactic and therapeutic treatments have been used. Examples for prophylaxis are HPV (Gardasil-9) [117] leading to cervical cancer and HBV (HEPLISAV-B) leading to hepatocellular carcinoma. The topic of this review is focused on mRNA vaccines for the treatment of cancer. All the other therapies will not be covered, as they have been addressed in various review articles [63,[118][119][120][121].
mRNA represents the minimal genetic vector and contains only the elements directly required for the expression of the encoded protein [56]. mRNA vaccines constitute an excellent platform for immunotherapy for a number of reasons [122][123][124]. A major point is the possibility for the simultaneous injection of more than one antigen, resulting in both immune and cell-mediated immunity, thereby increasing the likelihood of tumor tissue eradication [125][126][127].
Antigen selection is the most difficult task in mRNA-based immunotherapy. In the COVID-19 case, the opposite was the case. The choice was easy and straightforward, at least at the beginning when the selected antigen, the spike protein of SARS-CoV-2, was rather stable and more or less exempt from mutations. This changed as soon as the Delta and Gamma variants appeared. In cancer immunotherapy, choices for antigens are tumorassociated antigens (TAAs) [128]. The problem is, however, that these structures are also shared by healthy tissues-although at lower expression levels [129]-and therefore likewise become a target of the mRNA administration leading to autoimmunity and resistance. The ideal antigen should be highly immunogenic, expressed only in cancer cells and not or only at very reduced levels in normal tissue, and it should be essential for cancer cell survival [130]. Tissue-specific antigens (TSAs), therefore, are preferable but much less available. Examples are PSA and HER2. However, immunotherapies based on these TSAs were only of limited success [131]. More preferable options are neoantigens, which originate from non-synonymous mutations in tumor cells and are absent in normal cells [132][133][134][135]. Neoantigens are presented by major histocompatibility complex (MHC) molecules.
Furthermore, mRNA vaccines are not constrained by the patient's HLA class. mRNA vaccines are preferable over DNA vaccines for several reasons. DNA has to reach the nucleus of the tumor cells in contrast to mRNA, for which it is sufficient to reach the cytosol. For mRNA, there is no danger of splice mutations [64,136]. Modifications addressing efficacy, half-life, etc. are easier to achieve by either modifying the molecule itself, the route of administration, or the formulation [69,[137][138][139][140][141].
The mRNA modifications described in Section 2 to improve stability and efficacy apply to cancer vaccines as well. Since uridine-rich sequences activate Toll-like receptors [142], which suppress RNA recognition [32], uridine should be replaced by N1-methylpseudouridine (1 mΨ Despite all these efforts, the tumor is trying to avoid the effect of the treatment by various methods. Tumor escape mechanisms down-regulate the tumor cell surface antigens which means reducing the immunogenicity and thereby the efficacy of the mRNA vaccine [147]. Another action is the up-regulation of immune checkpoint expression on cell surfaces, for example, PD-L1, which inhibits T lymphocyte activity and induces immune evasion [148]. Furthermore, immunosuppressive cells, MDSCs, and Tregs can be recruited into the tumor microenvironment and cytokines can be secreted which inhibit the immune response [149]. The route of administration of mRNA-based vaccines plays an important role [150]. In addition to intramuscular injection, subcutaneous, intradermal, intranodal [151], and intratumoral routes have been used [152]. Adjuvants such as protamine, granulocytemacrophage colony-stimulating factor (GM-CSF), and interleukin 2 (IL2) cause strong activation of the innate immune system, which leads to a potent adaptive immune response [153]. mRNA vaccines are not immunogenic, so that multiple administrations are possible [154][155][156]. However, as already mentioned above, there is one caveat. Formulations containing pegylated moieties might still run into the problem of pre-existing antibodies [71,156]. This issue had first been detected in animal models leading to modifications of the formulations [157][158][159] and verification in animals and in patients [160].
According to Barbier [3], the following objectives should be taken into account in the development of an mRNA cancer vaccine. A strong cytotoxic CD8 + T cell response is needed to eradicate cancer cells. The antigens should be selected such that they are able to induce highly tumor-specific immune responses. Potential targets are TAAs, TSAs, and neoantigens.
The major obstacle to this objective is the high variability of antigens across different individuals [161] or even within the same patient. The heterogeneity of tumors-not only between patients but also within the same patient when looking at the time axis-is one of the highest hurdles to treatment success. Most likely, there are not two identical tumors in the whole patient population of a cancer type. This issue is illustrated in Figure 4. A possible way out of this dilemma is the selection of more than one antigen within the mRNA vaccine.  Neoantigens should be the major target whenever possible [135,162]. During carcinogenesis, malignant cells acquire somatic mutations that lead to the production of protein sequences not expressed by normal cells [132,163]. These proteins are called neoantigens. They may be either common across various patients or specific for each patient, which then can be used to develop personalized treatment [164][165][166]. To that end, the tumor is biopsied, and the neoantigens are analyzed by sequencing, encoded in mRNA, and injected into the patient [3,[167][168][169]. Common or shared neoantigens include, for example, BRAF and NRAS mutations, which are observed in approx. 50% and 15-25% of melanoma patients, respectively [170,171]. The advantage of this approach is manifold. First, it is an individual treatment, which is a major point taking into consideration the heterogeneity of tumors. Second, mRNA cannot only encode one whole neoantigen but also as many as there are found in the individual patient [3]. Sahin et al. [172] reported that the first injectable mRNA cancer vaccine encoding neoantigens for advanced melanoma patients through intranodal injection achieved potent T cell responses against multiple neoantigens in all patients after vaccination. However, although personalized cancer vaccines based on neoantigens have shown encouraging results, a large number of predicted neoantigens tend to trigger very few actual anti-tumor responses [173]. Table 4 provides a selection of neoantigens.  Neoantigens should be the major target whenever possible [135,162]. During carcinogenesis, malignant cells acquire somatic mutations that lead to the production of protein sequences not expressed by normal cells [132,163]. These proteins are called neoantigens. They may be either common across various patients or specific for each patient, which then can be used to develop personalized treatment [164][165][166]. To that end, the tumor is biopsied, and the neoantigens are analyzed by sequencing, encoded in mRNA, and injected into the patient [3,[167][168][169]. Common or shared neoantigens include, for example, BRAF and NRAS mutations, which are observed in approx. 50% and 15-25% of melanoma patients, respectively [170,171]. The advantage of this approach is manifold. First, it is an individual treatment, which is a major point taking into consideration the heterogeneity of tumors. Second, mRNA cannot only encode one whole neoantigen but also as many as there are found in the individual patient [3]. Sahin et al. [172] reported that the first injectable mRNA cancer vaccine encoding neoantigens for advanced melanoma patients through intranodal injection achieved potent T cell responses against multiple neoantigens in all patients after vaccination. However, although personalized cancer vaccines based on neoantigens have shown encouraging results, a large number of predicted neoantigens tend to trigger very few actual anti-tumor responses [173]. Table 4 provides a selection of neoantigens.  Neoantigens should be the major target whenever possible [135,162]. During carcinogenesis, malignant cells acquire somatic mutations that lead to the production of protein sequences not expressed by normal cells [132,163]. These proteins are called neoantigens. They may be either common across various patients or specific for each patient, which then can be used to develop personalized treatment [164][165][166]. To that end, the tumor is biopsied, and the neoantigens are analyzed by sequencing, encoded in mRNA, and injected into the patient [3,[167][168][169]. Common or shared neoantigens include, for example, BRAF and NRAS mutations, which are observed in approx. 50% and 15-25% of melanoma patients, respectively [170,171]. The advantage of this approach is manifold. First, it is an individual treatment, which is a major point taking into consideration the heterogeneity of tumors. Second, mRNA cannot only encode one whole neoantigen but also as many as there are found in the individual patient [3]. Sahin et al. [172] reported that the first injectable mRNA cancer vaccine encoding neoantigens for advanced melanoma patients through intranodal injection achieved potent T cell responses against multiple neoantigens in all patients after vaccination. However, although personalized cancer vaccines based on neoantigens have shown encouraging results, a large number of predicted neoantigens tend to trigger very few actual anti-tumor responses [173]. Table 4 provides a selection of neoantigens.  Neoantigens should be the major target whenever possible [135,162]. During carcinogenesis, malignant cells acquire somatic mutations that lead to the production of protein sequences not expressed by normal cells [132,163]. These proteins are called neoantigens. They may be either common across various patients or specific for each patient, which then can be used to develop personalized treatment [164][165][166]. To that end, the tumor is biopsied, and the neoantigens are analyzed by sequencing, encoded in mRNA, and injected into the patient [3,[167][168][169]. Common or shared neoantigens include, for example, BRAF and NRAS mutations, which are observed in approx. 50% and 15-25% of melanoma patients, respectively [170,171]. The advantage of this approach is manifold. First, it is an individual treatment, which is a major point taking into consideration the heterogeneity of tumors. Second, mRNA cannot only encode one whole neoantigen but also as many as there are found in the individual patient [3]. Sahin et al. [172] reported that the first injectable mRNA cancer vaccine encoding neoantigens for advanced melanoma patients through intranodal injection achieved potent T cell responses against multiple neoantigens in all patients after vaccination. However, although personalized cancer vaccines based on neoantigens have shown encouraging results, a large number of predicted neoantigens tend to trigger very few actual anti-tumor responses [173]. Table 4 provides a selection of neoantigens. . Neoantigens should be the major target whenever possible [135,162]. During carcinogenesis, malignant cells acquire somatic mutations that lead to the production of protein sequences not expressed by normal cells [132,163]. These proteins are called neoantigens. They may be either common across various patients or specific for each patient, which then can be used to develop personalized treatment [164][165][166]. To that end, the tumor is biopsied, and the neoantigens are analyzed by sequencing, encoded in mRNA, and injected into the patient [3,[167][168][169]. Common or shared neoantigens include, for example, BRAF and NRAS mutations, which are observed in approx. 50% and 15-25% of melanoma patients, respectively [170,171]. The advantage of this approach is manifold. First, it is an individual treatment, which is a major point taking into consideration the heterogeneity of tumors. Second, mRNA cannot only encode one whole neoantigen but also as many as there are found in the individual patient [3]. Sahin et al. [172] reported that the first injectable mRNA cancer vaccine encoding neoantigens for advanced melanoma patients through intranodal injection achieved potent T cell responses against multiple neoantigens in all patients after vaccination. However, although personalized cancer vaccines based on neoantigens have shown encouraging results, a large number of predicted neoantigens tend to trigger very few actual anti-tumor responses [173]. Table 4 provides a selection of neoantigens. Dosing of mRNA can be achieved by titrating up or down, depending on the need, weight, and disease state of the patient. The duration of action is intrinsically limited by mRNA degradation, reducing the likelihood of irreversible side effects, and enabling the treatment of acute indications [184,185].
Formulations for use in cancer treatment are similar to those described for COVID-19 administration. However, as opposed to the delivery systems of mRNA vaccines for pathogen infections, therapeutic mRNA vaccines for cancer treatments are required to generate both robust CD8 + and CD4 + T cell responses [186]. The activation of type I interferon (IFN) proved important in developing a cytotoxic T cell response [187]. Early on, naked mRNA dissolved in Ringer's solution at a concentration of 1.0 mg mL −1 was injected into separate inguinal lymph nodes in thirteen melanoma patients [172]. However, since naked mRNA cannot enter cell membranes freely but only by cell-mediated endocytic pathways [56,[64][65][66], preferred formulations not only for the treatment of COVID-19 but also for cancer therapy are LNPs which can substantially affect intracellular delivery efficiency, determine cell specificity of delivery, and modulate immunogenicity [3]. Ionizable lipid components of LNPs play a key role in multiple aspects of mRNA delivery, including particle formation, cellular uptake, and endosomal escape [44,57,188]. Their self-adjuvant activity, resulting in the stimulation of specific parts of the immune system, is an important aspect of mRNA LNPs. Examples are the stimulator of IFN-γ (STING) pathway and the TLR-RIG-I-like receptor (RLR)-independent mediator of innate immune responses [3,189]. Other formulations that have been developed are hybrid lipopolymer shell mRNA nanoparticles or lipoplexes [190]; nanocapsules with flexible polysaccharide shells and hollow cores, termed a sugar-capsule composed of mannan carrying mRNA [191]; and a nanoparticle platform, called mRNA Galsomes [192]. Reviews on formulations are available from various groups [193][194][195][196]. Large-scale production has been established extremely rapidly [88]. Table 5 provides a selection of mRNA vaccines in development.   As can be seen from Table 5, the range of indications is rather broad reaching from head and neck, breast, lung, pancreas, prostate, gastric, and rectal cancer to melanoma. BNT111 encoding the antigens, tyrosinase, NY-ESO-1, MAGE A3, and TPTE, is currently being investigated in melanoma. A Phase I trial (Lipo-MERIT) using liposomal mRNA, injected intravenously, in combination with or without the checkpoint inhibitor PD1, showed in an interim analysis induction of strong CD4 and CD8 T cell immunity against the vaccine antigens [197]. A Phase II trial with BNT111 ± cemiplimab is ongoing. Preliminary data from a Phase I/II trial of liposomal BNT112 encoding the five prostate cancer-specific antigens, kallikrein-2, kallikrein-3, acid phosphatase prostate, HOXB13, and NK3 homeobox 1, as monotherapy or in combination with cemiplimab, in metastatic castration resistant prostate cancer (mCRPC) showed that all five antigens were immunogenic and responses to each antigen were observed in at least two patients [198]. BNT121 was studied in melanoma. Thirteen metastatic patients received repeat administrations in inguinal lymph nodes showing clear immunological responses and some evidence of clinical activity [3]. Data from a Phase I trial in solid tumors with nine i.v. injections of BNT122 (RO7198457) encoding twenty patient-specific antigens have been reported by Braiteh [199]. BNT122 induced the pulsatile release of pro-inflammatory cytokines with each dose, consistent with the innate immune agonist activity of the RNA. Neoantigen-specific T cell responses were observed in peripheral blood in 14/16 patients (87%). Overall, 1 out of 26 patients had a complete response and 11 had a stable disease. mRNA-2416 is encoding the immune checkpoint modulator, OX40L, and is administered intratumorally. In a first trial with 41 patients with different malignancies, the compound did not meet the response criteria for solid tumors. A Phase II study in ovarian cancer in combination with durvalumab (NCT03323398) is ongoing [3]. mRNA-2752 encoding OX40L, IL-23, and IL-36γ was investigated in colorectal cancer (NCT03739931). However, in 17 patients there were no responses [198]. mRNA-4157 is a personalized mRNA encoding up to 34 neoantigens. In a Phase II solid tumor study, either as monotherapy (N = 16) or in combination with pembrolizumab (N = 63), multiple disease-free patients were observed during the study [199]. The safety and immunogenicity of mRNA-4650 encoding-defined neoantigens, mutations in driver genes, and HLA-I-predicted epitopes were determined in patients with metastatic gastrointestinal cancer (NCT03480152). CD8 + and CD4 + neoantigen-specific T cells elicited by the vaccine could be detected. However, since in 3 of 4 patients, no clinical response was observed, Phase II was not initiated [205]. TriMix, a mixture of monocyte-derived dendritic cells electroporated with mRNA encoding CD70, CD40 ligand, and constitutively active TLR4, as well as the tumor-associated antigens tyrosinase, gp100, MAGE-A3, or MAGE-C2 were administered together with ipilimumab in patients with advanced melanoma [206]. Enzyme-linked immunospot assay responses detected after in vitro T cell stimulation were shown in 12/15 patients. Vaccination in combination with ipilimumab resulted in robust CD8 + T cell responses in a meaningful portion of late-stage melanoma patients, and obviously in patients with a clinical response. TriMixDC-MEL, autologous monocyte-derived mRNA co-electroporated dendritic cells with mRNA encoding CD40L, CD70, and caTLR4, was administered i.v. to 21 late-stage melanoma patients. The control group without treatment comprised 20 patients. One year after randomization, 71% of patients in the study arm were alive and free of disease compared to 35% in the control arm [208]. The median time to non-salvageable recurrence was superior in the TriMixDC-MEL arm (median 8 months (range 1-6) vs. not reached. CV9103 (RNActive ® ) is based on four prostate-specific antigens, PSA, PSMA, PSCA, and STEAP and was administered intradermally in a Phase I trial in prostate cancer patients [210]. In a subsequent Phase II study, antigen-specific T cells were detected in around 80% of prostate carcinoma patients independent of their HLA background. A majority of immune responders, around 58%, reacted against multiple antigens, and the responses were detected against all antigens independent of their cellular localization. Individual patients were showing prolonged stabilization of PSA levels after initial rises. One patient had a greater than 85% drop in his PSA level [211]. CV9104 is a mixture of six mRNAs, each encoding one antigen, PSA, PSCA, PSMA, STEAP1, PAP, or MUC1. No significant difference in OS was found between the vaccine and control arm. There were also no significant differences in the rPFS endpoints and time to symptom progression [212]. CV9201 is an RNActive ® -based cancer immunotherapy encoding five NSCLC antigens, NY-ESO-1, MAGE C1, MAGE C2, survivin, and TBG. In Phase IIa, antigen-specific immune responses against ≥1 antigen were detected in 63% of evaluable patients after treatment. The frequency of activated IgD CD38 B cells increased >2-fold. In total, 31% of evaluable patients in Phase IIa had a stable disease and 69% had a progressive disease [213]. MEDI1191, which is encoding IL-12, showed in a Phase I study in combination with durvalumab two partial responses in patients who had received prior immunotherapy and also had PD-L1-negative tumors [217]. The direct injection in 15 melanoma patients of naked mRNA encoding various antigens has been described by Weide et al. [218]. An increase in the anti-tumor humoral immune response was seen in some patients. However, a demonstration of the clinical effectiveness of the direct injection of copy mRNA for anti-tumor immunotherapy was not shown. Another Phase I trial in 30 renal cell cancer patients using naked mRNA coding for MUC1, CEA, Her-2/neu, telomerase, survivin, and MAGE-A1 has been published by Rittig et al. [219].
The induction of CD4 + and CD8 + T cell responses was shown for several TAAs. In summary, the clinical efficacy of mRNA vaccines looks like a mixed bag of different outcomes ranging from no effect to a rather good response, which means that the route to sufficient efficacy still is rather steep but, nevertheless, nourishes high hopes for the future.

Conclusions
The tremendous success and unbelievable speed in developing and bringing to the market COVID-19 mRNA vaccines has raised high expectations for duplicating this performance in other indications, in particular in cancer therapy. mRNA vaccines are easy to design and can rapidly be modified if there is a need for change, exhibiting an extremely broad versatility of building blocks, structural elements, and formulations of the synthetic mRNA including the targeting of defined cells, duration of expression, and immunological effects. Dosing can be adjusted according to the individual needs of the patients, and there is the possibility to individualize treatment by adjusting the selection of antigen(s) to the specific tumor type of each patient. However, the hurdles in cancer immunotherapy are much higher than in fighting coronaviruses. In particular, the following issues need to be addressed: neoantigen use, LNP modifications, combination treatments, immune escape of cancer, and the therapeutic vs. prophylactic use of mRNA vaccines.
If available, neoantigen-specific vaccines, which are currently considered a top priority in cancer immunotherapy, result in a potential killing of all cells exhibiting the neoantigen epitope but leave any other cancer cells unharmed (which might then start or continue multiplying). As a consequence, the cancer genotype and phenotype changes, and the tumor will continue growing [220]. Targeting multiple neoantigens within a single vaccine and/or combining the vaccine with non-mRNA tumor therapies should be targeted to reduce immune evasion and effectively eliminate tumors. Similarly, the general mechanisms for the immune escape of cancer, which are much more complex than those for breakthrough infections of pathogens [221], need to be addressed. Combinations of mRNA vaccines with agents that can reverse immunosuppression such as immune agonists or cytokines and compounds that block immune checkpoints have been shown to be more potent than a single administration of vaccine therapy [41, 64,214,222,223]. However, not all patients are responsive to these treatments [224]. Nevertheless, this approach, and the combination of one or more non-mRNA agents fighting immune escape with multitarget mRNAs, should be continued even more extensively in the future.
LNPs are state-of-the-art for mRNA vaccines. However, they bear several pitfalls. Since they are particles, they can be taken up by macrophages of the liver or the spleenirrespective of their administration route-thereby reducing the activity at the intended site of action, the tumor, and/or leading to side effects at off-target accumulation sites. The side effects might include toxicity and/or immunogenicity [150,225] and inflammation exacerbation [226]. The effects depend on various factors including the constituents of the vaccine carrier and the antigen. For example, cationic/ionizable lipids can induce inflammation by activating TLR pathways and lead to cell toxicity. On the other hand, this inflammatory effect is a useful contributor to the adjuvant activity of the LNP. Accordingly, the positive adjuvant activity and the negative inflammatory action need to be carefully balanced by selecting an appropriate cationic/ionizable lipid [227]. Other strategies have been described including the effort to make the particles invisible to macrophages by modifying their surface with, for example, PEG derivatives. The caveat for this solution has been illustrated above: the potential immunogenicity of PEG compounds [160]. Yet another possibility has been published by Dirisala [228], regarding the in situ stealth coating of liver sinusoids using linear or two-armed PEG-conjugated oligo(L-lysine). Another approach could be the reduction in the size of the particles. Small particles are no longer taken up by the liver or the spleen; they keep circulating in the blood, and therefore prolong the activity, as has been described for blood-pool contrast agents [229]. In summary, the mRNA vaccine carrier for cancer immunotherapy, in particular LNPs, needs further consideration to optimize activity and tolerability.
A further issue that needs to be addressed is the observation that vaccine treatment may not successfully reach the tumor site, in particular solid tumors [175]. Probably, cancer vaccine treatments are more suitable for patients with an undisturbed immune system, a relatively small tumor load, and a greater risk of recurrence.
The focus in cancer immunotherapy should also be directed towards prophylactic vaccination, which has already been very successful in HPV and HBV leading to cervical cancer and hepatocellular carcinoma. An extension to other tumor types seems mandatory.